Planet HantsLUG

Syndicate content
Planet HantsLUG - http://hantslug.org.uk/planet/
Updated: 2 min 19 sec ago

Andy Smith: Giving Cinema Paradiso a try

Sun, 24/09/2017 - 00:38
Farewell, LoveFiLM

I’ve been a customer of LoveFiLM for something like 12 years—since before they were owned by Amazon. In their original incarnation they were great: very cheap, and titles very often arrived in exactly the order you specified, i.e. they often managed to send the thing from the very top of the list.

In 2011 they got bought by Amazon and I was initially a bit concerned, but to be honest Amazon have run it well. The single list disappeared and was replaced by three priority lists; high, normal and low, and then a list of things that haven’t yet been released. New rentals were supposed to almost always come from the high priority list (as long as you had enough titles on there) but in a completely unpredictable order. Though of course they would keep multi-disc box sets together, and send lower-numbered seasons before later seasons.

Amazon have now announced that they’re shutting LoveFiLM by Post down at the end of October which I think is a shame, as it was a service I still enjoy.

It was inevitable I suppose due to the increasing popularity of streaming and downloads, and although I’m perfectly able to do the streaming and download thing, receiving discs by post still works for me.

I am used to receiving mockery for consuming some of my entertainment on little plastic discs that a human being has to physically transport to my residence, but LoveFiLM’s service was still cheap, the selection was very good, things could be rented as soon as they were available on disc, and the passive nature of just making a list and having the things sent to me worked well for me.

Cinema Paradiso

My first thought was that that was it for the disc-by-post rental model in the UK. That progress had left it behind. But very quickly people pointed me to Cinema Paradiso. After a quick look around I’ve decided to give it a try and so here are my initial thoughts.

Pricing

At a casual glance the pricing is slightly worse than LoveFiLM’s. I was paying £6.99 a month for 2 discs at home, unlimited rental per month. £6.98 at Cinema Paradiso gets you 2 discs at home but only 4 rentals per month.

I went back through my LoveFiLM rental history for the last year and found there were only 2 months where I managed to rent more than 4 discs, and those times I rented 5 and 6 discs respectively. Realistically it doesn’t seem like 4 discs per month will be much of a restriction to me.

Annoyingly, Cinema Paradiso have a 2 week trial period but only if you sign up to the £9.98 subscription (6 discs a month). You’d have to remember to downgrade to the cheaper subscriptions after 2 weeks, if that’s all you wanted.

Selection

I was pleasantly surprised at how good the selection is at Cinema Paradiso. Not only did they have every title that is currently on my LoveFiLM rental list (96 titles), but they also had a few things that LoveFiLM thinks haven’t been released yet.

I’m not going to claim that my tastes are particularly niche, but there are a few foreign language films and some anime in there, and release dates range from the 70s to 2017.

Manual approval

It seems that new Cinema Paradiso signups need to be manually approved, and this happens only on week days between 8am and mid day. I’ve signed up on a Saturday evening so nothing will get sent out until Monday I suppose.

It’s probably not a big deal as we’re talking about the postal service here so even with LoveFiLM nothing would get posted out until Monday anyway. It is a little jarring after moving away from the behemoth that is Amazon though, and serves as a reminder that Cinema Paradiso is a much smaller company.

Searching for titles

The search feature is okay. It provides suggestions as you type but if your title is obscure then it may not appear in the list of suggestions at all you and need to submit the search box and look through the longer list that appears.

A slight niggle is that if you have moused over any of the initial suggestions it replaces your text with that, so if your title isn’t amongst the suggestions you now have to re-type it.

I like that it shows a rating from Rotten Tomatoes as well as from their own site’s users. LoveFiLM shows IMDB ratings which I don’t trust very much, and also Amazon ratings, which I don’t trust at all for movies or TV. Seeing some of the shockingly-low Rotten Tomatoes scores for some of my LoveFiLM titles resulted in my Cinema Paradiso list shrinking to 83 titles!

Rental list mechanics

It’s hard to tell for sure at this stage because I haven’t yet got my account approved and had any rentals, but it looks to me like the rental list mechanics are a bit clunky compared to LoveFiLM’s.

At LoveFiLM at the point of adding a new title you would choose which of the three “buckets” to put a rental in; high priority, normal priority, or low priority. Every title in those buckets were of equal priority to every other item in the same bucket. So, when adding a new title all you had to consider was whether it was high, medium or low.

Cinema Paradiso has a single big list of rentals. In some ways this might appeal because you can fine-tune what order you would like things in. But I would suggest that very few people want to put that much effort into ordering their list. Personally, when I add a new title I can cope with:

  • “I want to see this soon”
  • “I want to see this some time”
  • “I want to see this, but I’m not bothered when”

Cinema Paradiso appears to want me to say:

  • “Put this at the top, I want it immediately!”
  • “This belongs at #11, just after the 6th season of American Horror Story, but before Capitalism: A love Story
  • “Just stick it at the end”

I can’t find any explanation anywhere on their site as to how the selection actually works, so the logical assumption is that they go down your list from top to bottom until they find a title that you want that they have available right now. Without the three buckets to put titles in, it seems to me then that every addition will have to involve some list management unless I either want to see that title really soon, or probably never.

I’ll have to give it a go but this mechanism seems a bit more awkward than LoveFiLM’s approach and needlessly so, because LoveFiLM’s way doesn’t make any promises about which of the titles in each bucket will come next either, nor even that it will be anything from the high priority bucket at all. Although I cannot remember a time when something has come that wasn’t from the high priority bucket.

Cinema Paradiso does let you have more than one list, and you can divide your disc allocation between lists, but I don’t think I could emulate the high/normal/low with that. Having a 2 disc allocation I’d always be getting one disc from the “high” list and one disc from the “normal” priority, which isn’t how I’d want that to work.

Let’s see how it goes.

Referral

I did not know when I signed up that there was a referral scheme which is a shame because I do know some people already using Cinema Paradiso. If you’re going to sign up then please use my referral link. I will get a ⅙ reduction in rental fees for each person that does.

Categories: LUG Community Blogs

Andy Smith: Tricky issues when upgrading to the GoCardless “Pro” API

Thu, 21/09/2017 - 22:06
Background

Since 2012 BitFolk has been using GoCardless as a Direct Debit payment provider. On the whole it has been a pleasant experience:

  • Their API is a pleasure to integrate against, having excellent documentation
  • Their support is responsive and knowledgeable
  • Really good sandbox environment with plenty of testing tools
  • The fees, being 1% capped at £2.00, are pretty good for any kind of payment provider (much less than PayPal, Stripe, etc.)

Of course, if I was submitting Direct Debits myself there would be no charge at all, but BitFolk is too small and my bank (Barclays) are not interested in talking to me about that.

The “Pro” API

In September 2014 GoCardless came out with a new version of their API called the “Pro API”. It made a few things nicer but didn’t come with any real new features applicable to BitFolk, and also added a minimum fee of £0.20.

The original API I’d integrated against has a 1% fee capped at £2.00, and as BitFolk’s smallest plan is £10.79 including VAT the fee would generally be £0.11. Having a £0.20 fee on these payments would represent nearly a doubling of fees for many of my payments.

So, no compelling reason to use the Pro API.

Over the years, GoCardless made more noise about their Pro API and started calling their original API the “legacy API”. I could see the way things were going. Sure enough, eventually they announced that the legacy API would be disabled on 31 October 2017. No choice but to move to the Pro API now.

Payment caps

There aren’t normally any limits on Direct Debit payments. When you let your energy supplier or council or whatever do a Direct Debit, they can empty your bank account if they like.

The Direct Debit Guarantee has very strong provisions in it for protecting the payee and essentially if you dispute anything, any time, you get your money back without question and the supplier has to pursue you for the money by other means if they still think the charge was correct. A company that repeatedly gets Direct Debit chargebacks is going to be kicked off the service by their bank or payment provider.

The original GoCardless API had the ability to set caps on the mandate which would be enforced their side. A simple “X amount per Y time period”. I thought that this would provide some comfort to customers who may not be otherwise familiar with authorising Direct Debits from small companies like BitFolk, so I made use of that feature by default.

This turned out to be a bad decision.

The main problem with this was that there was no way to change the cap. If a customer upgraded their service then I’d have to cancel their Direct Debit mandate and ask them to authorise a new one because it would cease being possible to charge them the correct amount. Authorising a new mandate was not difficult—about the same amount of work as making any sort of online payment—but asking people to do things is always a pain point.

There was a long-standing feature request with GoCardless to implement some sort of “follow this link to authorise the change” feature, but it never happened.

Payment caps and the new API

The Pro API does not support mandates with a capped amount per interval. Given that I’d already established that it was a mistake to do that, I wasn’t too bothered about that.

I’ve since discovered however that the Pro API not only does not support setting the caps, it does not have any way to query them either. This is bad because I need to use the Pro API with mandates that were created in the legacy API. And all of those have caps.

Here’s the flow I had using the legacy API.
Legacy payment process

This way if the charge was coming a little too early, I could give some latitude and let it wait a couple of days until it could be charged. I’d also know if the problem was that the cap was too low. In that case there would be no choice but to cancel the customer’s mandate and ask them to authorise another one, but at least I would know exactly what the problem was.

With the Pro API, there is no way to check timings and charge caps. All I can do is make the charge, and then if it’s too soon or too much I get the same error message:

“Validation failed / exceeds mandate cap”

That’s it. It doesn’t tell me what the cap is, it doesn’t tell me if it’s because I’m charging too soon, nor if I’m charging too much. There is no way to distinguish between those situations.

Backwards compatible – sort of

GoCardless talk about the Pro API being backwards compatible to the legacy API, so that once switched I would still be able to create payments against mandates that were created using the legacy API. I would not need to get customers to re-authorise.

This is true to a point, but my use of caps per interval in the legacy API has severely restricted how compatible things are, and that’s something I wasn’t aware of. Sure, their “Guide to upgrading” does briefly mention that caps would continue to be enforced:

“Pre-authorisation mandates are not restricted, but the maximum amount and interval that you originally specified will still apply.”

That is the only mention of this issue in that entire document, and that statement would be fine by me, if there would have continued to be a way to tell which failure mode would be encountered.

Thinking that I was just misunderstanding, I asked GoCardless support about this. Their reply:

Thanks for emailing.

I’m afraid the limits aren’t exposed within the new API. The only solution as you suggest, is to try a payment and check for failure.

Apologies for the inconvenience caused here and if you have any further queries please don’t hesitate to let us know.

What now?

I am not yet sure of the best way to handle this.

The nuclear option would be to cancel all mandates and ask customers to authorise them again. I would like to avoid this if possible.

I am thinking that most customers continue to be fine on the “amount per interval” legacy mandates as long as they don’t upgrade, so I can leave them as they are until that happens. If they upgrade, or if a DD payment ever fails with “exceeds mandate cap” then I will have to cancel their mandate and ask them to authorise again. I can see if their mandate was created before ~today and advise them on the web site to cancel it and authorise it again.

Conclusion

I’m a little disappointed that GoCardless didn’t think that there would need to be a way to query mandate caps even though creating new mandates with those limits is no longer possible.

I can’t really accept that there is a good level of backwards compatibility here if there is a feature that you can’t even tell is in use until it causes a payment to fail, and even then you can’t tell which details of that feature cause the failure.

I understand why they haven’t just stopped honouring the caps: it wouldn’t be in line with the consumer-focused spirit of the Direct Debit Guarantee to alter things against customer expectations, and even sending out a notification to the customer might not be enough. I think they should have gone the other way and allowed querying of things that they are going to continue to enforce, though.

Could I have tested for this? Well, the difficulty there is that the GoCardless sandbox environment for the Pro API starts off clean with no access to any of your legacy activity neither from live nor from legacy sandbox. So I couldn’t do something like the following:

  1. Create legacy mandate in legacy sandbox, with amount per interval caps
  2. Try to charge against the legacy mandate from the Pro API sandbox, exceeding the cap
  3. Observe that it fails but with no way to tell why

I did note that there didn’t seem to be attributes of the mandate endpoint that would let me know when it could be charged and what the amount left to charge was, but it didn’t set off any alarm bells. Perhaps it should have.

Also I will admit I’ve had years to switch to Pro API and am only doing it now when forced. Perhaps if I had made a start on this years ago, I’d have noted what I consider to be a deficiency, asked them to remedy it and they might have had time to do so. I don’t actually think it’s likely they would bump the API version for that though. In my defence, as I mentioned, there is nothing attractive about the Pro API for my use, and it does cost more, so no surprise I’ve been reluctant to explore it.

So, if you are scrambling to update your GoCardless integration before 31 October, do check that you are prepared for payments against capped mandates to fail.

Categories: LUG Community Blogs

Andy Smith: When is a 64-bit counter not a 64-bit counter?

Sun, 03/09/2017 - 21:17

…when you run a Xen device backend (commonly dom0) on a kernel version earlier than 4.10, e.g. Debian stable.

TL;DR

Xen netback devices used 32-bit counters until that bug was fixed and released in kernel version 4.10.

On a kernel with that bug you will see counter wraps much sooner than you would expect, and if the interface is doing enough traffic for there to be multiple wraps in 5 minutes, your monitoring will no longer be accurate.

The problem

A high-bandwidth VPS customer reported that the bandwidth figures presented by BitFolk’s monitoring bore no resemblance to their own statistics gathered from inside their VPS. Their figures were a lot higher.

About octet counters

The Linux kernel maintains byte/octet counters for its network interfaces. You can view them in /sys/class/net/interface>/statistics/*_bytes.

They’re a simple count of bytes transferred, and so the count always goes up. Typically these are 64-bit unsigned integers so their maximum value would be 18,446,744,073,709,551,615 (264-1).

When you’re monitoring bandwidth use the monitoring system records the value and the timestamp. The difference in value over a known period allows the monitoring system to work out the rate.

Wrapping

Monitoring of network devices is often done using SNMP. SNMP has 32-bit and 64-bit counters.

The maximum value that can be held in a 32-bit counter is 4,294,967,295. As that is a byte count, that represents 34,359,738,368 bits or 34,359.74 megabits. Divide that by 300 (seconds in 5 minutes) and you get 114.5. Therefore if the average bandwidth is above 114.5Mbit/s for 5 minutes, you will overflow a 32-bit counter. When the counter overflows it wraps back through zero.

Wrapping a counter once is fine. We have to expect that a counter will wrap eventually, and as counters never decrease, if a new value is smaller than the previous one then we know it has wrapped and can still work out what the rate should be.

The problem comes when the counter wraps more than once. There is no way to tell how many times it has wrapped so the monitoring system will have to assume the answer is once. Once traffic reaches ~229Mbit/s the counters will be wrapping at least twice in 5 minutes and the statistics become meaningless.

64-bit counters to the rescue

For that reason, network traffic is normally monitored using 64-bit counters. You would have to have a traffic rate of almost 492 Petabit/s to wrap a 64-bit byte counter in 5 minutes.

The thing is, I was already using 64-bit SNMP counters.

Examining the sysfs files

I decided to remove SNMP from the equation by going to the source of the data that SNMP uses: the kernel on the device being monitored.

As mentioned, the kernel’s interface byte counters are exposed in sysfs at /sys/class/net/interface>/statistics/*_bytes. I dumped out those values every 10 seconds and watched them scroll in a terminal session.

What I observed was that these counters, for that particular customer, were wrapping every couple of minutes. I never observed a value greater than 8,469,862,875. That’s larger than a 32-bit counter would hold, but very close to what a 33 bit counter would hold (8,589,934,591).

64-bit counters not to the rescue

Once I realised that the kernel’s own counters were wrapping every couple of minutes inside the kernel it became clear that using 64-bit counters in SNMP was not going to help at all, and multiple wraps would be seen in 5 minutes.

What a difference a minute makes

To test the hypothesis I switched to 1-minute polling. Here’s what 12 hours of real data looks like under both 5- and 1-minute polling.

As you can see that is a pretty dramatic difference.

The bug

By this point, I’d realised that there must be a bug in Xen’s netback driver (the thing that makes virtual network interfaces in dom0).

I went searching through the source of the kernel and found that the counters had changed from an unsigned long in kernel version 4.9 to a u64 in kernel version 4.10.

Of course, once I knew what to search for it was easy to unearth a previous bug report. If I’d found that at the time of the initial report that would have saved 2 days of investigation!

Even so, the fix for this was only committed in February of this year so, unfortunately, is not present in the kernel in use by the current Debian stable. Nor in many other current distributions.

For Xen set-ups on Debian the bug could be avoided by using a backports kernel or packaging an upstream kernel.

Or you could do 1-minute polling as that would only wrap one time at an average bandwidth of ~572Mbit/s and should be safe from multiple wraps up to ~1.1Gbit/s.

Inside the VPS the counters are 64-bit so it isn’t an issue for guest administrators.

Categories: LUG Community Blogs

Debian Bits: Work on Debian for mobile devices continues

Thu, 17/08/2017 - 14:39

Work on Debian for mobile devices, i.e. telephones, tablets, and handheld computers, continues. During the recent DebConf17 in Montréal, Canada, more than 50 people had a meeting to reconsider opportunities and challenges for Debian on mobile devices.

A number of devices were shown at DebConf:

  • PocketCHIP: A very small handheld computer with keyboard, Wi-Fi, USB, and Bluetooth, running Debian 8 (Jessie) or 9 (Stretch).
  • Pyra: A modular handheld computer with a touchscreen, gaming controls, Wi-Fi, keyboard, multiple USB ports and SD card slots, and an optional modem for either Europe or the USA. It will come preinstalled with Debian.
  • Samsung Galaxy S Relay 4G: An Android smartphone featuring a physical keyboard, which can already run portions of Debian userspace on the Android kernel. Kernel upstreaming is on the way.
  • ZeroPhone: An open-source smartphone based on Raspberry Pi Zero, with a small screen, classic telephone keypad and hardware switches for telephony, Wi-Fi, and the microphone. It is running Debian-based Raspbian OS.

The photo (click to enlarge) shows all four devices, together with a Nokia N900, which was the first Linux-based smartphone by Nokia, running Debian-based Maemo and a completely unrelated Gnuk cryptographic token, which just sneaked into the setting.

If you like to participate, please

Categories: LUG Community Blogs

Debian Bits: Debian turns 24!

Wed, 16/08/2017 - 16:50

Today is Debian's 24th anniversary. If you are close to any of the cities celebrating Debian Day 2017, you're very welcome to join the party!

If not, there's still time for you to organize a little celebration or contribution to Debian. For example, spread the word about Debian Day with this nice piece of artwork created by Debian Developer Daniel Lenharo de Souza and Valessio Brito, taking inspiration from the desktop themes Lines and softWaves by Juliette Belin:

If you also like graphics design, or design in general, have a look at https://wiki.debian.org/Design and join the team! Or you can visit the general list of Debian Teams for many other opportunities to participate in Debian development.

Thanks to everybody who has contributed to develop our beloved operating system in these 24 years, and happy birthday Debian!

Categories: LUG Community Blogs

Debian Bits: DebConf17 closes in Montreal and DebConf18 dates announced

Sat, 12/08/2017 - 22:59

Today, Saturday 12 August 2017, the annual Debian Developers and Contributors Conference came to a close. With over 405 people attending from all over the world, and 169 events including 89 talks, 61 discussion sessions or BoFs, 6 workshops and 13 other activities, DebConf17 has been hailed as a success.

Highlights included DebCamp with 117 participants, the Open Day,
where events of interest to a broader audience were offered, talks from invited speakers (Deb Nicholson, Matthew Garrett and Katheryn Sutter), the traditional Bits from the DPL, lightning talks and live demos and the announcement of next year's DebConf (DebConf18 in Hsinchu, Taiwan).

The schedule has been updated every day, including 32 ad-hoc new activities, planned
by attendees during the whole conference.

For those not able to attend, talks and sessions were recorded and live streamed, and videos are being made available at the Debian meetings archive website. Many sessions also facilitated remote participation via IRC or a collaborative pad.

The DebConf17 website will remain active for archive purposes, and will continue to offer links to the presentations and videos of talks and events.

Next year, DebConf18 will be held in Hsinchu, Taiwan, from 29 July 2018 until 5 August 2018. It will be the first DebConf held in Asia. For the days before DebConf the local organisers will again set up DebCamp (21 July - 27 July), a session for some intense work on improving the distribution, and organise the Open Day on 28 July 2018, aimed at the general public.

DebConf is committed to a safe and welcome environment for all participants. See the DebConf Code of Conduct and the Debian Code of Conduct for more details on this.

Debian thanks the commitment of numerous sponsors to support DebConf17, particularly our Platinum Sponsors Savoir-Faire Linux, Hewlett Packard Enterprise, and Google.

About Savoir-faire Linux

Savoir-faire Linux is a Montreal-based Free/Open-Source Software company with offices in Quebec City, Toronto, Paris and Lyon. It offers Linux and Free Software integration solutions in order to provide performance, flexibility and independence for its clients. The company actively contributes to many free software projects, and provides mirrors of Debian, Ubuntu, Linux and others.

About Hewlett Packard Enterprise

Hewlett Packard Enterprise (HPE) is one of the largest computer companies in the world, providing a wide range of products and services, such as servers, storage, networking, consulting and support, software, and financial services.

HPE is also a development partner of Debian, and provides hardware for port development, Debian mirrors, and other Debian services.

About Google

Google is one of the largest technology companies in the world, providing a wide range of Internet-related services and products as online advertising technologies, search, cloud computing, software, and hardware.

Google has been supporting Debian by sponsoring DebConf since more than ten years, at gold level since DebConf12, and at platinum level for this DebConf17.

Categories: LUG Community Blogs

Steve Kemp: A day in the life of Steve

Sat, 12/08/2017 - 22:00

I used to think I was a programmer who did "sysadmin-stuff". Nowadays I interact with too many real programmers to believe that.

Or rather I can code/program/develop, but I'm not often as good as I could be. These days I'm getting more consistent with writing tests, and I like it when things are thoroughly planned and developed. But too often if I'm busy, or distracted, I think to myself "Hrm .. compiles? Probably done. Oops. Bug, you say?"

I was going to write about working with golang today. The go language is minimal and quite neat. I like the toolset:

  • go fmt
    • Making everything consistent.
  • go test

Instead I think today I'm going to write about something else. Since having a child a lot of my life is different. Routine becomes something that is essential, as is planning and scheduling.

So an average week-day goes something like this:

  • 6:00AM
    • Wake up (naturally).
  • 7:00AM
    • Wake up Oiva and play with him for 45 minutes.
  • 7:45AM
    • Prepare breakfast for my wife, and wake her up, then play with Oiva for another 15 minutes while she eats.
  • 8:00AM
    • Take tram to office.
  • 8:30AM
    • Make coffee, make a rough plan for the day.
  • 9:00AM
    • Work, until lunchtime which might be 1pm, 2pm, or even 3pm.
  • 5:00PM
    • Leave work, and take bus home.
    • Yes I go to work via tram, but come back via bus. There are reasons.
  • 5:40PM
    • Arrive home, and relax in peace for 20 minutes.
  • 6:00PM-7:00PM
    • Take Oiva for a walk, stop en route to relax in a hammock for 30 minutes reading a book.
  • 7:00-7:20PM
    • Feed Oiva his evening meal.
  • 7:30PM
    • Give Oiva his bath, then pass him over to my wife to put him to bed.
  • 7:30PM - 8:00pm
    • Relax
  • 8:00PM - 10:00PM
    • Deal with Oiva waking up, making noises, or being unsettled.
    • Try to spend quality time with my wife, watch TV, read a book, do some coding, etc.
  • 10:00PM ~ 11:30PM
    • Go to bed.

In short I'm responsible for Oiva from 6ish-8ish in the morning, then from 6PM-10PM (with a little break while he's put to bed.) There are some exceptions to this routine - for example I work from home on Monday/Friday afternoons, and Monday evenings he goes to his swimming classes. But most working-days are the same.

Weekends are a bit different. There I tend to take him 6AM-8AM, then 1PM-10PM with a few breaks for tea, and bed. At the moment we're starting to reach the peak-party time of year, which means weekends often involve negotiation(s) about which parent is having a party, and which parent is either leaving early, or not going out at all.

Today I have him all day, and it's awesome. He's just learned to say "Daddy" which makes any stress, angst or unpleasantness utterly worthwhile.

Categories: LUG Community Blogs

Debian Bits: DebConf17 starts today in Montreal

Sun, 06/08/2017 - 15:00

DebConf17, the 18th annual Debian Conference, is taking place in Montreal, Canada from August 6 to August 12, 2017.

Debian contributors from all over the world have come together at Collège Maisonneuve during the preceding week for DebCamp (focused on individual work and team sprints for in-person collaboration developing Debian), and the Open Day on August 5th (with presentations and workshops of interest to a wide audience).

Today the main conference starts with nearly 400 attendants and over 120 activities scheduled, including 45- and 20-minute talks and team meetings, workshops, a job fair, talks from invited speakers, as well as a variety of other events.

The full schedule at https://debconf17.debconf.org/schedule/ is updated every day, including activities planned ad-hoc by attendees during the whole conference.

If you want to engage remotely, you can follow the video streaming of the events happening in the three talk rooms: Buzz (the main auditorium), Rex, and Bo, or join the conversation about what is happening in the talk rooms: #debconf17-buzz, #debconf17-rex and #debconf17-bo, and the BoF (discussions) rooms: #debconf17-potato and #debconf17-woody (all those channels in the OFTC IRC network).

DebConf is committed to a safe and welcome environment for all participants. See the DebConf Code of Conduct and the Debian Code of Conduct for more details on this.

Debian thanks the commitment of numerous sponsors to support DebConf17, particularly our Platinum Sponsors Savoir-Faire Linux, Hewlett Packard Enterprise, and Google.

Categories: LUG Community Blogs

Debian Bits: Google Platinum Sponsor of DebConf17

Sat, 05/08/2017 - 22:30

We are very pleased to announce that Google has committed support to DebConf17 as a Platinum sponsor.

Google is one of the largest technology companies in the world, providing a wide range of Internet-related services and products as online advertising technologies, search, cloud computing, software, and hardware.

Google has been supporting Debian by sponsoring DebConf since more than ten years, and at gold level since DebConf12.

With this additional commitment as Platinum Sponsor for DebConf17, Google contributes to make possible our annual conference, and directly supports the progress of Debian and Free Software helping to strengthen the community that continues to collaborate on Debian projects throughout the rest of the year.

Thank you very much Google, for your support of DebConf17!

DebConf17 is starting!

Many Debian contributors are already taking advantage of DebCamp and the Open Day to work individually or in groups developing and improving Debian. DebConf17 will officially start on August 6, 2017. Visit the DebConf17 website at https://debconf17.debconf.org to know the schedule, live streaming and other details.

Categories: LUG Community Blogs

Debian Bits: DebConf17 Open Day

Sat, 05/08/2017 - 15:00

Today, the day preceeding the official start of the annual Debian Conference, is the Open Day at DebConf17, at Collège Maisonneuve in Montreal (Canada).

This day is open to the public with events of interest to a wide audience.

The schedule of today's events include, among others:

  • A Newbie's Newbie Guide to Debian
  • Ask Anything About Debian
  • Debian Packaging 101
  • Debian InstallFest
  • Presentations or workshops related to free software projects and local organizations.

Everyone is welcome to attend! It is a great possibility for interested users to meet our community and for Debian to widen our community.

See the full schedule for today's events at https://debconf17.debconf.org/schedule/open-day/.

If you want to engage remotely, you can watch the video streaming of the Open Day events happening in the "Rex" room, or join the conversation in the channels #debconf17-rex, #debconf17-potato and #debconf17-woody in the OFTC IRC network.

DebConf is committed to a safe and welcome environment for all participants. See the DebConf Code of Conduct and the Debian Code of Conduct for more details on this.

Debian thanks the commitment of numerous sponsors to support DebConf17, particularly our Platinum Sponsors Savoir-Faire Linux, Hewlett Packard Enterprise, and Google.

Categories: LUG Community Blogs

Andy Smith: A slightly more realistic look at lvmcache

Thu, 03/08/2017 - 14:59
Recap And then…

I decided to perform some slightly more realistic benchmarks against lvmcache.

The problem with the initial benchmark was that it only covered 4GiB of data with a 4GiB cache device. Naturally once lvmcache was working correctly its performance was awesome – the entire dataset was in the cache. But clearly if you have enough fast block device available to fit all your data then you don’t need to cache it at all and may as well just use the fast device directly.

I decided to perform some fio tests with varying data sizes, some of which were larger than the cache device.

Test methodology

Once again I used a Zipf distribution with a factor of 1.2, which should have caused about 90% of the hits to come from just 10% of the data. I kept the cache device at 4GiB but varied the data size. The following data sizes were tested:

  • 1GiB
  • 2GiB
  • 4GiB
  • 8GiB
  • 16GiB
  • 32GiB
  • 48GiB

With the 48GiB test I expected to see lvmcache struggling, as the hot 10% (~4.8GiB) would no longer fit within the 4GiB cache device.

A similar fio job spec to those from the earlier articles was used:

[cachesize-1g] size=512m ioengine=libaio direct=1 iodepth=8 numjobs=2 readwrite=randread random_distribution=zipf:1.2 bs=4k unlink=1 runtime=30m time_based=1 per_job_logs=1 log_avg_msec=500 write_iops_log=/var/tmp/fio-${FIOTEST}

…the only difference being that several different job files were used each with a different size= directive. Note that as there are two jobs, the size= is half the desired total data size: each job lays out a data file of the specified size.

For each data size I took care to fill the cache with data first before doing a test run, as unreproducible performance is still seen against a completely empty cache device. This produced IOPS logs and a completion latency histogram. Test were also run against SSD and HDD to provide baseline figures.

Results IOPS graphs All-in-one

Immediately we can see that for data sizes 4GiB and below performance converges quite quickly to near-SSD levels. That is very much what we would expect when the cache device is 4GiB, so big enough to completely cache everything.

Let’s just have a look at the lower-performing configurations.

Low-end performers

For 8, 16 and 32GiB data sizes performance clearly gets progressively worse, but it is still much better than baseline HDD. The 10% of hot data still fits within the cache device, so plenty of acceleration is still happening.

For the 48GiB data size it is a little bit of a different story. Performance is still better (on average) than baseline HDD, but there are periodic dips back down to roughly HDD figures. This is because not all of the 10% hot data fits into the cache device any more. Cache misses cause reads from HDD and consequently end up with HDD levels of performance for those reads.

The results no longer look quite so impressive, with even the 8GiB data set achieving only a few thousand IOPS on average. Are things as bad as they seem? Well no, I don’t think they are, and to see why we will have to look at the completion latency histograms.

Completion latency histograms

The above graphs are generated by fitting a Bezier curve to a scatter of data points each of which represents a 500ms average of IOPS achieved. The problem there is the word average.

It’s important to understand what effect averaging the figures gives. We’ve already seen that HDDs are really slow. Even if only a few percent of IOs end up missing cache and going to HDD, the massive latency of those requests will pull the average for the whole 500ms window way down.

Presumably we have a cache because we suspect we have hot spots of data, and we’ve been trying to evaluate that by doing most of the reads from only 10% of the data. Do we care what the average performance is then? Well it’s a useful metric but it’s not going to say much about the performance of reads from the hot data.

The histogram of completion latencies can be more useful. This shows how long it took between issuing the IO and completing the read for a certain percentage of issued IOs. Below I have focused on the 50% to 99% latency buckets, with the times for each bucket averaged between the two jobs. In the interests of being able to see anything at all I’ve had to double the height of the graph and still cut off the y axis for the three worst performers.

A couple of observations:

  • Somewhere between 70% and 80% of IOs complete with a latency that’s so close to SSD performance as to be near-indistinguishable, no matter what the data size. So what I think I am proving is that:

    you can cache a 48GiB slow backing device with 4GiB of fast SSD and if you have 10% hot data then you can expect it to be served up at near-SSD latencies 70%–80% of the time. If your hot spots are larger (not so hot) then you won’t achieve that. If your fast device is larger than 1/12th the backing device then you should do better than 70%–80%.

  • If the cache were perfect then we should expect the 90th percentile to be near SSD performance even for the 32GiB data set, as the 10% hot spot of ~3.2GiB fits inside the 4GiB cache. For whatever reason this is not achieved, but for that data size the 90th percentile latency is still about half that of HDD.
  • When the backing device is many times larger (32GiB+) than the cache device, the 99th percentile latencies can be slightly worse than for baseline HDD.

    I hesitate to suggest there is a problem here as there are going to be very few samples in the top 1%, so it could just be showing close to HDD performance.

Conclusion

Assuming you are okay with using a 4.12..x kernel, and assuming you are already comfortable using LVM, then at the moment it looks fairly harmless to deploy lvmcache.

Getting a decent performance boost out of it though will require you to check that your data really does have hot spots and size your cache appropriately.

Measuring your existing workload with something like blktrace is probably advisable, and these days you can feed the output of blktrace back into fio to see what performance might be like in a difference configuration.

Full test output

You probably want to stop reading here unless the complete output of all the fio runs is of interest to you.

Baseline SSD $ cd /srv/ssd/fio && FIOTEST=ssd-1g fio ~/cachesize-1g.fio cachesize-1g: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes cachesize-1g: Laying out IO file(s) (1 file(s) / 512MB) cachesize-1g: Laying out IO file(s) (1 file(s) / 512MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [385.7MB/0KB/0KB /s] [98.8K/0/0 iops] [eta 00m:00s] cachesize-1g: (groupid=0, jobs=1): err= 0: pid=15206: Thu Aug 3 03:59:51 2017 read : io=346130MB, bw=196910KB/s, iops=49227, runt=1800001msec slat (usec): min=0, max=4188, avg= 4.76, stdev= 4.10 clat (usec): min=17, max=7093, avg=155.55, stdev=59.28 lat (usec): min=91, max=7098, avg=160.77, stdev=59.32 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 108], 10.00th=[ 112], 20.00th=[ 119], | 30.00th=[ 124], 40.00th=[ 131], 50.00th=[ 137], 60.00th=[ 147], | 70.00th=[ 159], 80.00th=[ 181], 90.00th=[ 225], 95.00th=[ 274], | 99.00th=[ 366], 99.50th=[ 402], 99.90th=[ 474], 99.95th=[ 506], | 99.99th=[ 924] lat (usec) : 20=0.01%, 50=0.01%, 100=0.57%, 250=92.21%, 500=7.16% lat (usec) : 750=0.04%, 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01% cpu : usr=14.54%, sys=42.95%, ctx=50708834, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=88609406/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 cachesize-1g: (groupid=0, jobs=1): err= 0: pid=15207: Thu Aug 3 03:59:51 2017 read : io=346960MB, bw=197381KB/s, iops=49345, runt=1800001msec slat (usec): min=0, max=4180, avg= 4.76, stdev= 3.93 clat (usec): min=16, max=7097, avg=155.16, stdev=59.32 lat (usec): min=60, max=7101, avg=160.38, stdev=59.36 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 108], 10.00th=[ 113], 20.00th=[ 119], | 30.00th=[ 124], 40.00th=[ 131], 50.00th=[ 137], 60.00th=[ 145], | 70.00th=[ 157], 80.00th=[ 179], 90.00th=[ 225], 95.00th=[ 274], | 99.00th=[ 370], 99.50th=[ 406], 99.90th=[ 474], 99.95th=[ 506], | 99.99th=[ 892] lat (usec) : 20=0.01%, 50=0.01%, 100=0.59%, 250=92.13%, 500=7.22% lat (usec) : 750=0.05%, 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01% cpu : usr=14.75%, sys=43.14%, ctx=50816422, majf=0, minf=14 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=88821698/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=693090MB, aggrb=394291KB/s, minb=196909KB/s, maxb=197381KB/s, mint=1800001msec, maxt=1800001msec Disk stats (read/write): xvdf: ios=177420039/9, merge=0/6, ticks=27109524/0, in_queue=27169920, util=100.00% Baseline HDD $ cd /srv/slow/fio && FIOTEST=hdd-1g fio ~/cachesize-1g.fio cachesize-1g: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes cachesize-1g: Laying out IO file(s) (1 file(s) / 512MB) cachesize-1g: Laying out IO file(s) (1 file(s) / 512MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [2650KB/0KB/0KB /s] [662/0/0 iops] [eta 00m:00s] cachesize-1g: (groupid=0, jobs=1): err= 0: pid=18951: Thu Aug 3 04:30:08 2017 read : io=2307.8MB, bw=1312.5KB/s, iops=328, runt=1800099msec slat (usec): min=0, max=101, avg=14.31, stdev= 9.09 clat (usec): min=58, max=943511, avg=24362.98, stdev=47600.55 lat (usec): min=63, max=943517, avg=24378.15, stdev=47600.62 clat percentiles (usec): | 1.00th=[ 71], 5.00th=[ 82], 10.00th=[ 96], 20.00th=[ 127], | 30.00th=[ 181], 40.00th=[ 596], 50.00th=[ 1432], 60.00th=[ 9152], | 70.00th=[21632], 80.00th=[38656], 90.00th=[73216], 95.00th=[115200], | 99.00th=[228352], 99.50th=[280576], 99.90th=[415744], 99.95th=[468992], | 99.99th=[593920] lat (usec) : 100=11.17%, 250=23.28%, 500=4.34%, 750=2.55%, 1000=2.09% lat (msec) : 2=8.86%, 4=1.99%, 10=6.48%, 20=8.26%, 50=15.33% lat (msec) : 100=9.26%, 250=5.65%, 500=0.72%, 750=0.03%, 1000=0.01% cpu : usr=0.30%, sys=0.85%, ctx=572788, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=590611/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 cachesize-1g: (groupid=0, jobs=1): err= 0: pid=18952: Thu Aug 3 04:30:08 2017 read : io=2233.4MB, bw=1270.5KB/s, iops=317, runt=1800071msec slat (usec): min=0, max=113, avg=14.63, stdev= 9.18 clat (usec): min=56, max=1153.7K, avg=25167.62, stdev=48500.18 lat (usec): min=64, max=1153.8K, avg=25183.12, stdev=48500.29 clat percentiles (usec): | 1.00th=[ 72], 5.00th=[ 84], 10.00th=[ 99], 20.00th=[ 135], | 30.00th=[ 199], 40.00th=[ 892], 50.00th=[ 2640], 60.00th=[12224], | 70.00th=[21888], 80.00th=[39168], 90.00th=[74240], 95.00th=[117248], | 99.00th=[234496], 99.50th=[288768], 99.90th=[428032], 99.95th=[485376], | 99.99th=[602112] lat (usec) : 100=10.22%, 250=22.53%, 500=4.15%, 750=2.20%, 1000=1.83% lat (msec) : 2=8.43%, 4=1.66%, 10=6.40%, 20=11.41%, 50=15.37% lat (msec) : 100=9.26%, 250=5.69%, 500=0.79%, 750=0.04%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.32%, sys=0.82%, ctx=551097, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=571725/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=4540.4MB, aggrb=2582KB/s, minb=1270KB/s, maxb=1312KB/s, mint=1800071msec, maxt=1800099msec Disk stats (read/write): xvde: ios=1162334/4, merge=0/1, ticks=28774452/692, in_queue=28775512, util=100.00% 1GiB data $ (cd /srv/cache/fio && FIOTEST=cachesize-1g-full fio ~/cachesize-1g.fio) cachesize-1g: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes cachesize-1g: Laying out IO file(s) (1 file(s) / 512MB) cachesize-1g: Laying out IO file(s) (1 file(s) / 512MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [381.4MB/0KB/0KB /s] [97.7K/0/0 iops] [eta 00m:00s] cachesize-1g: (groupid=0, jobs=1): err= 0: pid=31905: Thu Aug 3 01:50:57 2017 read : io=340715MB, bw=193829KB/s, iops=48457, runt=1800001msec slat (usec): min=0, max=4500, avg= 7.09, stdev= 4.48 clat (usec): min=27, max=7408, avg=155.45, stdev=48.00 lat (usec): min=94, max=7415, avg=162.99, stdev=48.04 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 112], 10.00th=[ 118], 20.00th=[ 124], | 30.00th=[ 131], 40.00th=[ 137], 50.00th=[ 143], 60.00th=[ 151], | 70.00th=[ 163], 80.00th=[ 179], 90.00th=[ 211], 95.00th=[ 245], | 99.00th=[ 322], 99.50th=[ 350], 99.90th=[ 414], 99.95th=[ 450], | 99.99th=[ 644] lat (usec) : 50=0.01%, 100=0.21%, 250=95.28%, 500=4.49%, 750=0.01% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01% cpu : usr=14.89%, sys=52.95%, ctx=39879588, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=87222941/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 cachesize-1g: (groupid=0, jobs=1): err= 0: pid=31906: Thu Aug 3 01:50:57 2017 read : io=347850MB, bw=197888KB/s, iops=49472, runt=1800001msec slat (usec): min=0, max=4535, avg= 7.06, stdev= 4.60 clat (usec): min=25, max=7437, avg=152.11, stdev=47.15 lat (usec): min=62, max=7443, avg=159.62, stdev=47.23 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 111], 10.00th=[ 116], 20.00th=[ 123], | 30.00th=[ 129], 40.00th=[ 133], 50.00th=[ 139], 60.00th=[ 147], | 70.00th=[ 157], 80.00th=[ 175], 90.00th=[ 203], 95.00th=[ 239], | 99.00th=[ 314], 99.50th=[ 342], 99.90th=[ 410], 99.95th=[ 446], | 99.99th=[ 724] lat (usec) : 50=0.01%, 100=0.26%, 250=95.76%, 500=3.97%, 750=0.01% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01% cpu : usr=15.41%, sys=53.33%, ctx=39452819, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=89049700/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=688565MB, aggrb=391716KB/s, minb=193828KB/s, maxb=197888KB/s, mint=1800001msec, maxt=1800001msec Disk stats (read/write): dm-2: ios=176255221/16, merge=0/0, ticks=26526400/152, in_queue=26612320, util=100.00%, aggrios=58763012/5471, aggrmerge=0/0, aggrticks=8781482/98230, aggrin_queue=8904092, aggrutil=100.00% dm-1: ios=0/0, merge=0/0, ticks=0/0, in_queue=0, util=0.00%, aggrios=176289036/11, aggrmerge=0/5, aggrticks=26394628/0, aggrin_queue=26430140, aggrutil=100.00% xvdc: ios=176289036/11, merge=0/5, ticks=26394628/0, in_queue=26430140, util=100.00% dm-0: ios=176289036/16, merge=0/0, ticks=26344448/0, in_queue=26417584, util=100.00% dm-3: ios=0/16397, merge=0/0, ticks=0/294692, in_queue=294692, util=1.07%, aggrios=0/16397, aggrmerge=0/0, aggrticks=0/294580, aggrin_queue=294580, aggrutil=1.07% xvdd: ios=0/16397, merge=0/0, ticks=0/294580, in_queue=294580, util=1.07% 2GiB data $ (cd /srv/cache/fio && FIOTEST=cachesize-2g-full fio ~/cachesize-2g.fio) cachesize-2g: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes cachesize-2g: Laying out IO file(s) (1 file(s) / 1024MB) cachesize-2g: Laying out IO file(s) (1 file(s) / 1024MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [382.1MB/0KB/0KB /s] [98.4K/0/0 iops] [eta 00m:00s] cachesize-2g: (groupid=0, jobs=1): err= 0: pid=3134: Wed Aug 2 17:32:04 2017 read : io=343242MB, bw=195266KB/s, iops=48816, runt=1800001msec slat (usec): min=0, max=4825, avg= 6.98, stdev= 4.42 clat (usec): min=26, max=6629, avg=154.48, stdev=47.36 lat (usec): min=93, max=6634, avg=161.93, stdev=47.41 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 110], 10.00th=[ 115], 20.00th=[ 123], | 30.00th=[ 129], 40.00th=[ 137], 50.00th=[ 143], 60.00th=[ 151], | 70.00th=[ 163], 80.00th=[ 179], 90.00th=[ 207], 95.00th=[ 237], | 99.00th=[ 314], 99.50th=[ 342], 99.90th=[ 418], 99.95th=[ 458], | 99.99th=[ 1020] lat (usec) : 50=0.01%, 100=0.32%, 250=95.80%, 500=3.84%, 750=0.02% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01% cpu : usr=14.32%, sys=52.25%, ctx=39483832, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=87869839/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 cachesize-2g: (groupid=0, jobs=1): err= 0: pid=3135: Wed Aug 2 17:32:04 2017 read : io=338592MB, bw=192621KB/s, iops=48155, runt=1800001msec slat (usec): min=0, max=4485, avg= 7.00, stdev= 4.45 clat (usec): min=16, max=365626, avg=156.69, stdev=493.84 lat (usec): min=66, max=365646, avg=164.17, stdev=493.95 clat percentiles (usec): | 1.00th=[ 102], 5.00th=[ 109], 10.00th=[ 114], 20.00th=[ 122], | 30.00th=[ 127], 40.00th=[ 133], 50.00th=[ 141], 60.00th=[ 149], | 70.00th=[ 161], 80.00th=[ 177], 90.00th=[ 205], 95.00th=[ 241], | 99.00th=[ 318], 99.50th=[ 346], 99.90th=[ 426], 99.95th=[ 474], | 99.99th=[12352] lat (usec) : 20=0.01%, 50=0.01%, 100=0.40%, 250=95.46%, 500=4.10% lat (usec) : 750=0.01%, 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01% cpu : usr=14.08%, sys=51.79%, ctx=38995321, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=86679580/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=681834MB, aggrb=387887KB/s, minb=192621KB/s, maxb=195266KB/s, mint=1800001msec, maxt=1800001msec Disk stats (read/write): dm-2: ios=174542882/20, merge=0/0, ticks=26569204/1076, in_queue=26658352, util=100.00%, aggrios=58194070/26591, aggrmerge=0/0, aggrticks=8834468/168569, aggrin_queue=9026514, aggrutil=100.00% dm-1: ios=15/46976, merge=0/0, ticks=0/8080, in_queue=8084, util=0.35%, aggrios=174561550/46008, aggrmerge=1/6963, aggrticks=26076884/8636, aggrin_queue=26119484, aggrutil=100.00% xvdc: ios=174561550/46008, merge=1/6963, ticks=26076884/8636, in_queue=26119484, util=100.00% dm-0: ios=174561536/6745, merge=0/0, ticks=26039312/1316, in_queue=26111052, util=100.00% dm-3: ios=20660/26054, merge=0/0, ticks=464092/496312, in_queue=960408, util=5.44%, aggrios=20660/26051, aggrmerge=0/3, aggrticks=464080/495444, aggrin_queue=959516, aggrutil=5.44% xvdd: ios=20660/26051, merge=0/3, ticks=464080/495444, in_queue=959516, util=5.44% 4GiB data $ (cd /srv/cache/fio && FIOTEST=cachesize-4g-full fio ~/cachesize-4g.fio) cachesize-4g: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes cachesize-4g: Laying out IO file(s) (1 file(s) / 2048MB) cachesize-4g: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [375.8MB/0KB/0KB /s] [96.2K/0/0 iops] [eta 00m:00s] cachesize-4g: (groupid=0, jobs=1): err= 0: pid=26795: Wed Aug 2 16:22:51 2017 read : io=249614MB, bw=142002KB/s, iops=35500, runt=1800001msec slat (usec): min=0, max=4469, avg= 7.00, stdev= 4.50 clat (usec): min=9, max=1363.9K, avg=215.84, stdev=3099.68 lat (usec): min=65, max=1363.9K, avg=223.29, stdev=3099.98 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 109], 10.00th=[ 114], 20.00th=[ 121], | 30.00th=[ 127], 40.00th=[ 133], 50.00th=[ 139], 60.00th=[ 147], | 70.00th=[ 159], 80.00th=[ 177], 90.00th=[ 211], 95.00th=[ 255], | 99.00th=[ 354], 99.50th=[ 402], 99.90th=[ 4384], 99.95th=[35072], | 99.99th=[142336] lat (usec) : 10=0.01%, 50=0.01%, 100=0.30%, 250=94.32%, 500=5.21% lat (usec) : 750=0.06%, 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.02%, 50=0.03% lat (msec) : 100=0.02%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=10.92%, sys=38.72%, ctx=29995970, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=63901149/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 cachesize-4g: (groupid=0, jobs=1): err= 0: pid=26796: Wed Aug 2 16:22:51 2017 read : io=239913MB, bw=136484KB/s, iops=34120, runt=1800001msec slat (usec): min=0, max=4613, avg= 7.08, stdev= 4.64 clat (usec): min=20, max=1315.4K, avg=224.86, stdev=3340.78 lat (usec): min=67, max=1315.4K, avg=232.39, stdev=3341.07 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 110], 10.00th=[ 116], 20.00th=[ 122], | 30.00th=[ 127], 40.00th=[ 133], 50.00th=[ 139], 60.00th=[ 147], | 70.00th=[ 157], 80.00th=[ 177], 90.00th=[ 219], 95.00th=[ 274], | 99.00th=[ 390], 99.50th=[ 442], 99.90th=[10304], 99.95th=[39680], | 99.99th=[154624] lat (usec) : 50=0.01%, 100=0.27%, 250=92.97%, 500=6.52%, 750=0.12% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.02%, 50=0.04% lat (msec) : 100=0.02%, 250=0.02%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=10.28%, sys=37.64%, ctx=29036049, majf=0, minf=14 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=61417783/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=489527MB, aggrb=278486KB/s, minb=136483KB/s, maxb=142002KB/s, mint=1800001msec, maxt=1800001msec Disk stats (read/write): dm-2: ios=125296525/29, merge=0/0, ticks=27228000/6004, in_queue=27295900, util=100.00%, aggrios=41794883/193226, aggrmerge=0/0, aggrticks=9825693/38738, aggrin_queue=9881421, aggrutil=84.96% dm-1: ios=86/514016, merge=0/0, ticks=16/61332, in_queue=61376, util=2.08%, aggrios=125186978/561183, aggrmerge=0/17273, aggrticks=18938212/71188, aggrin_queue=19034916, aggrutil=85.91% xvdc: ios=125186978/561183, merge=0/17273, ticks=18938212/71188, in_queue=19034916, util=85.91% dm-0: ios=125186892/64571, merge=0/0, ticks=18904024/11972, in_queue=18966848, util=84.96% dm-3: ios=197671/1091, merge=0/0, ticks=10573040/42912, in_queue=10616040, util=33.53%, aggrios=197671/1084, aggrmerge=0/7, aggrticks=10573020/39132, aggrin_queue=10612148, aggrutil=33.53% xvdd: ios=197671/1084, merge=0/7, ticks=10573020/39132, in_queue=10612148, util=33.53% 8GiB data $ (cd /srv/cache/fio && FIOTEST=cachesize-8g-full fio ~/cachesize-8g.fio) cachesize-8g: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes cachesize-8g: Laying out IO file(s) (1 file(s) / 4096MB) cachesize-8g: Laying out IO file(s) (1 file(s) / 4096MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [18048KB/0KB/0KB /s] [4512/0/0 iops] [eta 00m:00s] cachesize-8g: (groupid=0, jobs=1): err= 0: pid=15599: Wed Aug 2 14:53:25 2017 read : io=13260MB, bw=7542.7KB/s, iops=1885, runt=1800146msec slat (usec): min=0, max=385, avg= 8.25, stdev= 5.69 clat (usec): min=9, max=1516.6K, avg=4231.66, stdev=26655.31 lat (usec): min=69, max=1516.6K, avg=4240.38, stdev=26657.41 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 104], 10.00th=[ 105], 20.00th=[ 107], | 30.00th=[ 109], 40.00th=[ 116], 50.00th=[ 119], 60.00th=[ 121], | 70.00th=[ 122], 80.00th=[ 124], 90.00th=[ 159], 95.00th=[16064], | 99.00th=[113152], 99.50th=[175104], 99.90th=[358400], 99.95th=[456704], | 99.99th=[700416] lat (usec) : 10=0.01%, 100=0.45%, 250=92.98%, 500=0.41%, 750=0.05% lat (usec) : 1000=0.04% lat (msec) : 2=0.08%, 4=0.01%, 10=0.22%, 20=1.29%, 50=2.04% lat (msec) : 100=1.25%, 250=0.93%, 500=0.21%, 750=0.03%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.95%, sys=2.94%, ctx=3521460, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=3394445/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 cachesize-8g: (groupid=0, jobs=1): err= 0: pid=15600: Wed Aug 2 14:53:25 2017 read : io=13886MB, bw=7899.2KB/s, iops=1974, runt=1800112msec slat (usec): min=0, max=2743, avg= 8.26, stdev= 5.84 clat (usec): min=2, max=1515.1K, avg=4040.21, stdev=24950.67 lat (usec): min=69, max=1515.1K, avg=4048.94, stdev=24952.83 clat percentiles (usec): | 1.00th=[ 100], 5.00th=[ 103], 10.00th=[ 105], 20.00th=[ 107], | 30.00th=[ 108], 40.00th=[ 110], 50.00th=[ 118], 60.00th=[ 120], | 70.00th=[ 121], 80.00th=[ 123], 90.00th=[ 157], 95.00th=[15552], | 99.00th=[109056], 99.50th=[166912], 99.90th=[333824], 99.95th=[419840], | 99.99th=[634880] lat (usec) : 4=0.01%, 100=0.56%, 250=92.98%, 500=0.31%, 750=0.05% lat (usec) : 1000=0.05% lat (msec) : 2=0.07%, 4=0.01%, 10=0.24%, 20=1.33%, 50=2.04% lat (msec) : 100=1.25%, 250=0.91%, 500=0.19%, 750=0.02%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.94%, sys=3.09%, ctx=3684550, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=3554777/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=27145MB, aggrb=15441KB/s, minb=7542KB/s, maxb=7899KB/s, mint=1800112msec, maxt=1800146msec Disk stats (read/write): dm-2: ios=6948764/45, merge=0/0, ticks=28718292/11380, in_queue=28731160, util=100.00%, aggrios=2383761/603042, aggrmerge=0/0, aggrticks=12202621/76005, aggrin_queue=12278754, aggrutil=100.00% dm-1: ios=48/1607066, merge=0/0, ticks=4/178612, in_queue=178652, util=5.79%, aggrios=6527499/1756743, aggrmerge=0/51937, aggrticks=781204/210664, aggrin_queue=991672, aggrutil=41.93% xvdc: ios=6527499/1756743, merge=0/51937, ticks=781204/210664, in_queue=991672, util=41.93% dm-0: ios=6527451/202019, merge=0/0, ticks=780348/38028, in_queue=818476, util=38.13% dm-3: ios=623784/41, merge=0/0, ticks=35827512/11376, in_queue=35839136, util=100.00%, aggrios=623784/10, aggrmerge=0/31, aggrticks=35827612/2744, aggrin_queue=35830320, aggrutil=100.00% xvdd: ios=623784/10, merge=0/31, ticks=35827612/2744, in_queue=35830320, util=100.00% 16GiB data $ (cd /srv/cache/fio && FIOTEST=cachesize-16g-full fio ~/cachesize-16g.fio) cachesize-16g: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes cachesize-16g: Laying out IO file(s) (1 file(s) / 8192MB) cachesize-16g: Laying out IO file(s) (1 file(s) / 8192MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [8288KB/0KB/0KB /s] [2072/0/0 iops] [eta 00m:00s] cachesize-16g: (groupid=0, jobs=1): err= 0: pid=2957: Wed Aug 2 13:13:01 2017 read : io=5417.5MB, bw=3081.9KB/s, iops=770, runt=1800060msec slat (usec): min=0, max=264, avg=10.14, stdev= 6.63 clat (usec): min=64, max=1972.9K, avg=10370.01, stdev=35718.97 lat (usec): min=71, max=1972.9K, avg=10380.72, stdev=35719.88 clat percentiles (usec): | 1.00th=[ 102], 5.00th=[ 105], 10.00th=[ 107], 20.00th=[ 116], | 30.00th=[ 119], 40.00th=[ 121], 50.00th=[ 122], 60.00th=[ 131], | 70.00th=[ 143], 80.00th=[ 165], 90.00th=[31104], 95.00th=[65280], | 99.00th=[173056], 99.50th=[228352], 99.90th=[374784], 99.95th=[440320], | 99.99th=[618496] lat (usec) : 100=0.28%, 250=82.96%, 500=0.10%, 750=0.04%, 1000=0.03% lat (msec) : 2=0.06%, 4=0.02%, 10=0.38%, 20=3.05%, 50=6.44% lat (msec) : 100=3.82%, 250=2.43%, 500=0.36%, 750=0.02%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.40%, sys=1.39%, ctx=1393813, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=1386875/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 cachesize-16g: (groupid=0, jobs=1): err= 0: pid=2958: Wed Aug 2 13:13:01 2017 read : io=6208.7MB, bw=3531.9KB/s, iops=882, runt=1800089msec slat (usec): min=0, max=437, avg=10.21, stdev= 6.74 clat (usec): min=1, max=1988.1K, avg=9046.94, stdev=31703.98 lat (usec): min=71, max=1989.2K, avg=9057.72, stdev=31704.93 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 104], 10.00th=[ 105], 20.00th=[ 107], | 30.00th=[ 112], 40.00th=[ 119], 50.00th=[ 121], 60.00th=[ 125], | 70.00th=[ 137], 80.00th=[ 157], 90.00th=[27776], 95.00th=[57088], | 99.00th=[152576], 99.50th=[201728], 99.90th=[329728], 99.95th=[395264], | 99.99th=[577536] lat (usec) : 2=0.01%, 50=0.01%, 100=0.41%, 250=83.62%, 500=0.09% lat (usec) : 750=0.04%, 1000=0.03% lat (msec) : 2=0.05%, 4=0.01%, 10=0.39%, 20=3.06%, 50=6.51% lat (msec) : 100=3.52%, 250=2.00%, 500=0.25%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.53%, sys=1.56%, ctx=1597691, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=1589415/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=11626MB, aggrb=6613KB/s, minb=3081KB/s, maxb=3531KB/s, mint=1800060msec, maxt=1800089msec Disk stats (read/write): dm-2: ios=2976172/77, merge=0/0, ticks=28759760/45044, in_queue=28805924, util=100.00%, aggrios=998518/56115, aggrmerge=0/0, aggrticks=9860773/22645, aggrin_queue=9883437, aggrutil=100.00% dm-1: ios=88/149089, merge=0/0, ticks=4/18444, in_queue=18452, util=0.58%, aggrios=2489913/161345, aggrmerge=0/6821, aggrticks=314936/21908, aggrin_queue=336800, aggrutil=17.38% xvdc: ios=2489913/161345, merge=0/6821, ticks=314936/21908, in_queue=336800, util=17.38% dm-0: ios=2489825/19183, merge=0/0, ticks=314816/4448, in_queue=319308, util=16.88% dm-3: ios=505643/74, merge=0/0, ticks=29267500/45044, in_queue=29312552, util=100.00%, aggrios=505643/12, aggrmerge=0/62, aggrticks=29267500/10580, aggrin_queue=29278060, aggrutil=100.00% xvdd: ios=505643/12, merge=0/62, ticks=29267500/10580, in_queue=29278060, util=100.00% 32GiB data $ (cd /srv/cache/fio && FIOTEST=cachesize-32g-full fio ~/cachesize-32g.fio) cachesize-32g: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes cachesize-32g: Laying out IO file(s) (1 file(s) / 16384MB) cachesize-32g: Laying out IO file(s) (1 file(s) / 16384MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [4624KB/0KB/0KB /s] [1156/0/0 iops] [eta 00m:00s] cachesize-32g: (groupid=0, jobs=1): err= 0: pid=23915: Wed Aug 2 11:42:12 2017 read : io=3378.9MB, bw=1922.5KB/s, iops=480, runt=1800108msec slat (usec): min=0, max=152, avg=11.84, stdev= 9.33 clat (usec): min=61, max=1523.8K, avg=16633.48, stdev=52554.00 lat (usec): min=68, max=1523.8K, avg=16645.89, stdev=52556.11 clat percentiles (usec): | 1.00th=[ 100], 5.00th=[ 104], 10.00th=[ 106], 20.00th=[ 108], | 30.00th=[ 117], 40.00th=[ 121], 50.00th=[ 123], 60.00th=[ 133], | 70.00th=[ 147], 80.00th=[ 8768], 90.00th=[49920], 95.00th=[104960], | 99.00th=[257024], 99.50th=[333824], 99.90th=[536576], 99.95th=[626688], | 99.99th=[880640] lat (usec) : 100=0.57%, 250=78.79%, 500=0.19%, 750=0.06%, 1000=0.05% lat (msec) : 2=0.09%, 4=0.02%, 10=0.47%, 20=2.96%, 50=6.82% lat (msec) : 100=4.70%, 250=4.22%, 500=0.94%, 750=0.11%, 1000=0.02% lat (msec) : 2000=0.01% cpu : usr=0.34%, sys=0.87%, ctx=916902, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=864974/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 cachesize-32g: (groupid=0, jobs=1): err= 0: pid=23916: Wed Aug 2 11:42:12 2017 read : io=5022.5MB, bw=2857.5KB/s, iops=714, runt=1800090msec slat (usec): min=0, max=173, avg=11.35, stdev= 9.15 clat (usec): min=62, max=1768.2K, avg=11185.81, stdev=48075.30 lat (usec): min=70, max=1768.2K, avg=11197.69, stdev=48077.75 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 104], 10.00th=[ 106], 20.00th=[ 108], | 30.00th=[ 115], 40.00th=[ 120], 50.00th=[ 122], 60.00th=[ 125], | 70.00th=[ 137], 80.00th=[ 153], 90.00th=[26240], 95.00th=[63744], | 99.00th=[220160], 99.50th=[309248], 99.90th=[585728], 99.95th=[733184], | 99.99th=[1122304] lat (usec) : 100=0.46%, 250=85.40%, 500=0.22%, 750=0.06%, 1000=0.04% lat (msec) : 2=0.07%, 4=0.02%, 10=0.25%, 20=2.04%, 50=5.27% lat (msec) : 100=2.99%, 250=2.39%, 500=0.62%, 750=0.12%, 1000=0.03% lat (msec) : 2000=0.02% cpu : usr=0.50%, sys=1.27%, ctx=1372044, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=1285733/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=8401.3MB, aggrb=4779KB/s, minb=1922KB/s, maxb=2857KB/s, mint=1800090msec, maxt=1800108msec Disk stats (read/write): dm-2: ios=2150707/143, merge=0/0, ticks=28767744/34836, in_queue=28804592, util=100.00%, aggrios=751272/286183, aggrmerge=0/0, aggrticks=11166794/101613, aggrin_queue=11268450, aggrutil=100.00% dm-1: ios=62/755357, merge=0/0, ticks=12/85084, in_queue=85108, util=2.67%, aggrios=1798588/823298, aggrmerge=494/28038, aggrticks=232996/102144, aggrin_queue=335068, aggrutil=15.62% xvdc: ios=1798588/823298, merge=494/28038, ticks=232996/102144, in_queue=335068, util=15.62% dm-0: ios=1799020/96389, merge=0/0, ticks=234124/20548, in_queue=254704, util=13.23% dm-3: ios=454734/6803, merge=0/0, ticks=33266248/199208, in_queue=33465540, util=100.00%, aggrios=454734/6184, aggrmerge=0/619, aggrticks=33266152/97928, aggrin_queue=33364124, aggrutil=100.00% xvdd: ios=454734/6184, merge=0/619, ticks=33266152/97928, in_queue=33364124, util=100.00% 48GiB data $ (cd /srv/cache/fio && FIOTEST=cachesize-48g-full fio ~/cachesize-48g.fio) cachesize-48g: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes cachesize-48g: Laying out IO file(s) (1 file(s) / 24576MB) cachesize-48g: Laying out IO file(s) (1 file(s) / 24576MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [4084KB/0KB/0KB /s] [1021/0/0 iops] [eta 00m:00s] cachesize-48g: (groupid=0, jobs=1): err= 0: pid=11765: Tue Aug 1 12:02:44 2017 read : io=3108.7MB, bw=1768.4KB/s, iops=442, runt=1800134msec slat (usec): min=0, max=75057, avg=12.49, stdev=100.40 clat (usec): min=61, max=7717.4K, avg=18080.36, stdev=79780.10 lat (usec): min=70, max=7717.4K, avg=18093.41, stdev=79783.33 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 104], 10.00th=[ 106], 20.00th=[ 108], | 30.00th=[ 115], 40.00th=[ 120], 50.00th=[ 124], 60.00th=[ 133], | 70.00th=[ 145], 80.00th=[ 197], 90.00th=[44288], 95.00th=[103936], | 99.00th=[296960], 99.50th=[419840], 99.90th=[1028096], 99.95th=[1269760], | 99.99th=[2179072] lat (usec) : 100=0.49%, 250=81.20%, 500=0.22%, 750=0.05%, 1000=0.04% lat (msec) : 2=0.07%, 4=0.03%, 10=0.34%, 20=2.26%, 50=6.08% lat (msec) : 100=4.04%, 250=3.79%, 500=1.03%, 750=0.19%, 1000=0.07% lat (msec) : 2000=0.09%, >=2000=0.01% cpu : usr=0.37%, sys=0.78%, ctx=846462, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=795801/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 cachesize-48g: (groupid=0, jobs=1): err= 0: pid=11766: Tue Aug 1 12:02:44 2017 read : io=3181.9MB, bw=1809.1KB/s, iops=452, runt=1800123msec slat (usec): min=0, max=177, avg=12.61, stdev=10.17 clat (usec): min=51, max=4963.9K, avg=17663.67, stdev=73795.84 lat (usec): min=68, max=4963.1K, avg=17676.84, stdev=73797.60 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 105], 10.00th=[ 107], 20.00th=[ 110], | 30.00th=[ 118], 40.00th=[ 122], 50.00th=[ 129], 60.00th=[ 137], | 70.00th=[ 151], 80.00th=[ 9664], 90.00th=[48384], 95.00th=[99840], | 99.00th=[259072], 99.50th=[354304], 99.90th=[978944], 99.95th=[1269760], | 99.99th=[2179072] lat (usec) : 100=0.47%, 250=78.72%, 500=0.21%, 750=0.06%, 1000=0.05% lat (msec) : 2=0.09%, 4=0.03%, 10=0.43%, 20=2.92%, 50=7.31% lat (msec) : 100=4.70%, 250=3.92%, 500=0.82%, 750=0.12%, 1000=0.05% lat (msec) : 2000=0.08%, >=2000=0.02% cpu : usr=0.40%, sys=0.80%, ctx=867597, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=814542/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=6290.5MB, aggrb=3578KB/s, minb=1768KB/s, maxb=1809KB/s, mint=1800123msec, maxt=1800134msec Disk stats (read/write): dm-2: ios=1610343/207, merge=0/0, ticks=28775436/83448, in_queue=28860328, util=100.00%, aggrios=562278/204739, aggrmerge=0/0, aggrticks=11044750/380021, aggrin_queue=11424792, aggrutil=100.00% dm-1: ios=53/537570, merge=0/0, ticks=4/63980, in_queue=64000, util=1.97%, aggrios=1303106/589473, aggrmerge=993/17280, aggrticks=184484/78092, aggrin_queue=262524, aggrutil=11.76% xvdc: ios=1303106/589473, merge=993/17280, ticks=184484/78092, in_queue=262524, util=11.76% dm-0: ios=1304046/69309, merge=0/0, ticks=188420/16684, in_queue=205128, util=9.97% dm-3: ios=382737/7340, merge=0/0, ticks=32945828/1059400, in_queue=34005248, util=100.00%, aggrios=382737/6163, aggrmerge=0/1177, aggrticks=32945132/641880, aggrin_queue=33587056, aggrutil=100.00% xvdd: ios=382737/6163, merge=0/1177, ticks=32945132/641880, in_queue=33587056, util=100.00%
Categories: LUG Community Blogs

Steve Kemp: So I did a thing, then another thing.

Wed, 02/08/2017 - 22:00

So I did start a project, to write a puppet-dashboard, it is functionally complete, but the next step is to allow me to raise alerts based on failing runs of puppet - in real-time.

(i.e. Now that I have a dashboard I wish to not use it. I want to be alerted to failures, without having to remember to go look for them. Something puppet-dashboard can't do ..)

In other news a while back I slipped in a casual note about having a brain scan done, here in sunny Helsinki.

One of the cool things about that experience, in addition to being told I wasn't going to drop dead that particular day, was that the radiologist told me that I could pay €25 to get a copy of my brain data in DICOM format.

I've not yet played with this very much, but I couldn't resist a brief animation:

  • See my brain.
    • Not the best quality, or the best detail, but damn. It is my brain.
    • I shall do better with more experimentation I think.
    • After I posted it my wife, a doctor, corrected me: That wasn't a gif of my brain, instead it was a gif of my skull. D'oh!
Categories: LUG Community Blogs

Andy Smith: Tracking down the lvmcache fix

Sun, 30/07/2017 - 07:00
Background

In the previous article I covered how, in order to get decent performance out of lvmcache with a packaged Debian kernel, you’d have to use the 4.12.2-1~exp1 kernel from experimental. The kernels packaged in sid, testing (buster) and stable (stretch) aren’t new enough.

I decided to bisect the Linux kernel upstream git repository to find out exactly which commit(s) fixed things.

Results

Here’s a graph showing the IOPS over time for baseline SSD and lvmcache with a full cache under several different kernel versions. As in previous articles, the lines are actually Bezier curves fitted to the data which is scattered all over the place from 500ms averages.

What we can see here is that performance starts to improve with commit 4d44ec5ab751 authored by Joe Thornber:

dm cache policy smq: put newly promoted entries at the top of the multiqueue

This stops entries bouncing in and out of the cache quickly.

This is part of a set of commits authored by Joe Thornber on the drivers/md/dm-cache-policy-smq.c

file and committed on 2017-05-14. By the time we reach commit 6cf4cc8f8b3b we have the long-term good performance that we were looking for.

The first of Joe Thornber’s commits on that day in the dm-cache area was 072792dcdfc8 and stepping back to the commit immediately prior to that one (2ea659a9ef48) we get a kernel representing the moment that Linus designated the v4.12-rc1 tag. Joe’s commits went into -rc1, and without them the performance of lvmcache under these test conditions isn’t much better than baseline HDD.

It seems like some of Joe’s changes helped a lot and then the last one really provided the long term performance.

git bisect procedure

Normally when you do a git bisect you’re starting with something that works and you’re looking for the commit that introduced a bug. In this case I was starting off with a known-good state and was interested in which commit(s) got me there. The normal bisect key words of “good” and “bad” in this case would be backwards to what I wanted. Dominic gave me the tip that I could alias the terms in order to reduce my confusion:

$ git bisect start --term-old broken --term-new fixed

From here on, when I encountered a test run that produced poor results I would issue:

$ git bisect broken

and when I had a test run with good results I would issue:

$ git bisect fixed

As I knew that the tag v4.13-rc1 produced a good run and v4.11 was bad, I could start off with:

$ git bisect reset v4.13-rc1 $ git bisect fixed $ git bisect broken v4.11

git would then keep bisecting the search space of commits until I would find the one(s) that resulted in the high performance I was looking for.

Good and bad?

As before I’m using fio to conduct the testing, with the same job specification:

ioengine=libaio direct=1 iodepth=8 numjobs=2 readwrite=randread random_distribution=zipf:1.2 bs=4k size=2g unlink=1 runtime=15m time_based=1 per_job_logs=1 log_avg_msec=500 write_iops_log=/var/tmp/fio-${FIOTEST}

The only difference from the other articles was that the run time was reduced to 15 minutes as all of the interesting behaviour happened within the first 11 minutes.

To recap, this fio job specification lays out two 2GiB files of random data and then starts two processes that perform 4kiB-sized reads against the files. Direct IO is used, in order to bypass the page cache.

A Zipfian distribution with a factor of 1.2 is used; this gives a 90/10 split where about 90% of the reads should come from about 10% of the data. The purpose of this is to simulate the hot spots of popular data that occur in real life. If the access pattern were to be perfectly and uniformly random then caching would not be effective.

In previous tests we had observed that dramatically different performance would be seen on the first run against an empty cache device compared to all other subsequent runs against what would be a full cache device. In the tests using kernels with the fix present the IOPS achieved would converge towards baseline SSD performance, whereas in kernels without the fix the performance would remain down near the level of baseline HDD. Therefore the fio tests were carried out twice.

Where to next?

I think I am going to see what happens when the cache device is pretty small in comparison to the working data.

All of the tests so far have used a 4GiB cache with 4GiB of data, so if everything got promoted it would entirely fit in cache. Not only that but the Zipf distribution makes most of the hits come from 10% of the data, so it’s actually just ~400MiB of hot data. I think it would be interesting to see what happens when the hot 10% is bigger than the cache device.

git bisect progress and test output

Unless you are particularly interested in the fio output and why I considered each one to be either fixed or broken, you probably want to stop reading now.

4.12.0-rc1+

I forgot to record the commit revision for this test so all I know is that it’s from somewhere in the first v4.12 release candidate.

The second job against an empty cache shows poor performance (8,210 average IOPS), and both jobs against a full cache show poor performance (2,141, 730 average IOPS), so this was marked broken.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc1+ fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [206.7MB/0KB/0KB /s] [52.9K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1275: Fri Jul 28 18:23:26 2017 read : io=126101MB, bw=143465KB/s, iops=35866, runt=900065msec slat (usec): min=3, max=25001, avg= 6.12, stdev= 4.73 clat (usec): min=19, max=808324, avg=214.83, stdev=2941.63 lat (usec): min=68, max=808334, avg=221.36, stdev=2942.04 clat percentiles (usec): | 1.00th=[ 102], 5.00th=[ 105], 10.00th=[ 106], 20.00th=[ 109], | 30.00th=[ 112], 40.00th=[ 116], 50.00th=[ 120], 60.00th=[ 123], | 70.00th=[ 129], 80.00th=[ 137], 90.00th=[ 161], 95.00th=[ 207], | 99.00th=[ 362], 99.50th=[ 422], 99.90th=[27008], 99.95th=[50944], | 99.99th=[138240] lat (usec) : 20=0.01%, 50=0.01%, 100=0.23%, 250=96.86%, 500=2.52% lat (usec) : 750=0.07%, 1000=0.05% lat (msec) : 2=0.11%, 4=0.01%, 10=0.01%, 20=0.03%, 50=0.08% lat (msec) : 100=0.03%, 250=0.02%, 500=0.01%, 750=0.01%, 1000=0.01% cpu : usr=9.28%, sys=32.19%, ctx=19968665, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=32281968/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1276: Fri Jul 28 18:23:26 2017 read : io=28868MB, bw=32843KB/s, iops=8210, runt=900062msec slat (usec): min=3, max=238, avg= 7.37, stdev= 3.18 clat (usec): min=1, max=718073, avg=964.43, stdev=8565.89 lat (usec): min=67, max=718080, avg=972.27, stdev=8566.73 clat percentiles (usec): | 1.00th=[ 100], 5.00th=[ 104], 10.00th=[ 106], 20.00th=[ 109], | 30.00th=[ 113], 40.00th=[ 117], 50.00th=[ 120], 60.00th=[ 122], | 70.00th=[ 126], 80.00th=[ 133], 90.00th=[ 157], 95.00th=[ 237], | 99.00th=[27264], 99.50th=[51456], 99.90th=[127488], 99.95th=[164864], | 99.99th=[261120] lat (usec) : 2=0.01%, 4=0.01%, 20=0.01%, 50=0.01%, 100=0.75% lat (usec) : 250=94.77%, 500=2.20%, 750=0.11%, 1000=0.08% lat (msec) : 2=0.15%, 4=0.02%, 10=0.13%, 20=0.52%, 50=0.75% lat (msec) : 100=0.35%, 250=0.16%, 500=0.01%, 750=0.01% cpu : usr=2.57%, sys=8.79%, ctx=6706909, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=7390152/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=154969MB, aggrb=176307KB/s, minb=32842KB/s, maxb=143465KB/s, mint=900062msec, maxt=900065msec Disk stats (read/write): dm-2: ios=39667388/28, merge=0/0, ticks=14142912/5308, in_queue=14158728, util=100.00%, aggrios=13263605/322030, aggrmerge=0/0, aggrticks=5479672/69964, aggrin_queue=5551697, aggrutil=100.00% dm-1: ios=2/847365, merge=0/0, ticks=0/131084, in_queue=131108, util=10.32%, aggrios=39457066/908046, aggrmerge=1462/48403, aggrticks=5249084/146536, aggrin_queue=5395788, aggrutil=99.89% xvdc: ios=39457066/908046, merge=1462/48403, ticks=5249084/146536, in_queue=5395788, util=99.89% dm-0: ios=39458525/110207, merge=0/0, ticks=5253332/20832, in_queue=5280728, util=99.89% dm-3: ios=332289/8519, merge=0/0, ticks=11185684/57976, in_queue=11243256, util=100.00%, aggrios=332289/6962, aggrmerge=0/1563, aggrticks=11185604/38544, aggrin_queue=11223832, aggrutil=100.00% xvdd: ios=332289/6962, merge=0/1563, ticks=11185604/38544, in_queue=11223832, util=100.00% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc1+ fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [14150KB/0KB/0KB /s] [3537/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1674: Fri Jul 28 18:51:41 2017 read : io=7529.9MB, bw=8566.5KB/s, iops=2141, runt=900093msec slat (usec): min=3, max=190, avg= 7.84, stdev= 5.41 clat (usec): min=34, max=1514.5K, avg=3725.26, stdev=24447.20 lat (usec): min=72, max=1514.6K, avg=3733.56, stdev=24449.04 clat percentiles (usec): | 1.00th=[ 100], 5.00th=[ 103], 10.00th=[ 104], 20.00th=[ 106], | 30.00th=[ 107], 40.00th=[ 108], 50.00th=[ 109], 60.00th=[ 118], | 70.00th=[ 121], 80.00th=[ 123], 90.00th=[ 155], 95.00th=[12992], | 99.00th=[100864], 99.50th=[158720], 99.90th=[325632], 99.95th=[419840], | 99.99th=[675840] lat (usec) : 50=0.01%, 100=0.39%, 250=93.41%, 500=0.29%, 750=0.08% lat (usec) : 1000=0.06% lat (msec) : 2=0.12%, 4=0.01%, 10=0.29%, 20=1.36%, 50=1.88% lat (msec) : 100=1.12%, 250=0.82%, 500=0.17%, 750=0.02%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.73%, sys=2.52%, ctx=1993978, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=1927642/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1675: Fri Jul 28 18:51:41 2017 read : io=2568.1MB, bw=2921.6KB/s, iops=730, runt=900113msec slat (usec): min=3, max=146, avg= 8.82, stdev= 5.83 clat (usec): min=64, max=1522.3K, avg=10941.33, stdev=44467.18 lat (usec): min=70, max=1522.3K, avg=10950.66, stdev=44468.17 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 104], 10.00th=[ 106], 20.00th=[ 108], | 30.00th=[ 117], 40.00th=[ 119], 50.00th=[ 121], 60.00th=[ 123], | 70.00th=[ 135], 80.00th=[ 161], 90.00th=[26240], 95.00th=[62720], | 99.00th=[205824], 99.50th=[288768], 99.90th=[528384], 99.95th=[659456], | 99.99th=[1036288] lat (usec) : 100=0.30%, 250=83.22%, 500=0.41%, 750=0.11%, 1000=0.08% lat (msec) : 2=0.15%, 4=0.02%, 10=0.62%, 20=3.47%, 50=5.45% lat (msec) : 100=3.16%, 250=2.31%, 500=0.57%, 750=0.08%, 1000=0.02% lat (msec) : 2000=0.01% cpu : usr=0.35%, sys=0.94%, ctx=673091, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=657432/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=10098MB, aggrb=11487KB/s, minb=2921KB/s, maxb=8566KB/s, mint=900093msec, maxt=900113msec Disk stats (read/write): dm-2: ios=2585032/29, merge=0/0, ticks=14380160/2084, in_queue=14383828, util=100.00%, aggrios=897708/321791, aggrmerge=0/0, aggrticks=6187020/40796, aggrin_queue=6227841, aggrutil=100.00% dm-1: ios=34/857325, merge=0/0, ticks=4/100092, in_queue=100120, util=6.30%, aggrios=2366278/926181, aggrmerge=0/38684, aggrticks=290092/115464, aggrin_queue=405484, aggrutil=35.21% xvdc: ios=2366278/926181, merge=0/38684, ticks=290092/115464, in_queue=405484, util=35.21% dm-0: ios=2366244/108039, merge=0/0, ticks=290336/20212, in_queue=310600, util=30.59% dm-3: ios=326848/10, merge=0/0, ticks=18270720/2084, in_queue=18272804, util=100.00%, aggrios=326848/6, aggrmerge=0/4, aggrticks=18270688/1332, aggrin_queue=18272160, aggrutil=100.00% xvdd: ios=326848/6, merge=0/4, ticks=18270688/1332, in_queue=18272160, util=100.00% 4.12.0+5518b69b7668 (2017-07-05 12:31:59 -0700)

Over 20k average IOPS on the run against a full cache means this was marked fixed.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0+ fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [401.3MB/0KB/0KB /s] [103K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1126: Fri Jul 28 21:06:01 2017 read : io=178623MB, bw=203233KB/s, iops=50808, runt=900001msec slat (usec): min=1, max=3059, avg= 7.04, stdev= 3.00 clat (usec): min=37, max=1175.7K, avg=147.92, stdev=585.27 lat (usec): min=66, max=1175.7K, avg=155.43, stdev=585.32 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 110], 10.00th=[ 114], 20.00th=[ 121], | 30.00th=[ 126], 40.00th=[ 131], 50.00th=[ 137], 60.00th=[ 143], | 70.00th=[ 153], 80.00th=[ 163], 90.00th=[ 185], 95.00th=[ 207], | 99.00th=[ 274], 99.50th=[ 314], 99.90th=[ 434], 99.95th=[ 498], | 99.99th=[ 1400] lat (usec) : 50=0.01%, 100=0.42%, 250=97.80%, 500=1.72%, 750=0.02% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=14.76%, sys=51.47%, ctx=19663307, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=45727501/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1127: Fri Jul 28 21:06:01 2017 read : io=170314MB, bw=193779KB/s, iops=48444, runt=900001msec slat (usec): min=0, max=4199, avg= 7.13, stdev= 3.11 clat (usec): min=29, max=493434, avg=155.48, stdev=688.81 lat (usec): min=64, max=493446, avg=163.07, stdev=688.95 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 111], 10.00th=[ 116], 20.00th=[ 122], | 30.00th=[ 129], 40.00th=[ 133], 50.00th=[ 139], 60.00th=[ 145], | 70.00th=[ 153], 80.00th=[ 167], 90.00th=[ 189], 95.00th=[ 215], | 99.00th=[ 290], 99.50th=[ 326], 99.90th=[ 446], 99.95th=[ 588], | 99.99th=[27008] lat (usec) : 50=0.01%, 100=0.32%, 250=97.35%, 500=2.27%, 750=0.02% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01% cpu : usr=13.65%, sys=50.13%, ctx=18808830, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=43600368/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=348937MB, aggrb=397012KB/s, minb=193779KB/s, maxb=203233KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=89308932/28, merge=0/0, ticks=13329748/144, in_queue=13365712, util=100.00%, aggrios=29798461/41330, aggrmerge=0/0, aggrticks=4447461/538024, aggrin_queue=4995029, aggrutil=100.00% dm-1: ios=54/56500, merge=0/0, ticks=4/10576, in_queue=10580, util=0.86%, aggrios=89368372/54448, aggrmerge=3/8823, aggrticks=12815140/10948, aggrin_queue=12836708, aggrutil=100.00% xvdc: ios=89368372/54448, merge=3/8823, ticks=12815140/10948, in_queue=12836708, util=100.00% dm-0: ios=89368321/7982, merge=0/0, ticks=12772360/1672, in_queue=12802664, util=100.00% dm-3: ios=27010/59510, merge=0/0, ticks=570020/1601824, in_queue=2171844, util=22.47%, aggrios=27010/59506, aggrmerge=0/4, aggrticks=570012/1588432, aggrin_queue=2158444, aggrutil=22.47% xvdd: ios=27010/59506, merge=0/4, ticks=570012/1588432, in_queue=2158444, util=22.47% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0+ fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=0): [f(2)] [100.0% done] [398.2MB/0KB/0KB /s] [102K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1701: Fri Jul 28 21:31:14 2017 read : io=81664MB, bw=92916KB/s, iops=23228, runt=900001msec slat (usec): min=1, max=236, avg= 7.03, stdev= 2.96 clat (usec): min=20, max=1436.8K, avg=334.90, stdev=5455.47 lat (usec): min=65, max=1436.8K, avg=342.37, stdev=5455.94 clat percentiles (usec): | 1.00th=[ 98], 5.00th=[ 106], 10.00th=[ 108], 20.00th=[ 116], | 30.00th=[ 122], 40.00th=[ 127], 50.00th=[ 135], 60.00th=[ 143], | 70.00th=[ 155], 80.00th=[ 171], 90.00th=[ 197], 95.00th=[ 225], | 99.00th=[ 322], 99.50th=[ 406], 99.90th=[57600], 99.95th=[102912], | 99.99th=[244736] lat (usec) : 50=0.01%, 100=1.15%, 250=95.96%, 500=2.51%, 750=0.04% lat (usec) : 1000=0.02% lat (msec) : 2=0.02%, 4=0.01%, 10=0.01%, 20=0.07%, 50=0.11% lat (msec) : 100=0.06%, 250=0.04%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=6.80%, sys=24.11%, ctx=10402905, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=20906091/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1702: Fri Jul 28 21:31:14 2017 read : io=70584MB, bw=80309KB/s, iops=20077, runt=900001msec slat (usec): min=0, max=1925, avg= 7.09, stdev= 3.08 clat (usec): min=28, max=1511.1K, avg=388.89, stdev=6127.50 lat (usec): min=66, max=1511.1K, avg=396.42, stdev=6127.96 clat percentiles (usec): | 1.00th=[ 100], 5.00th=[ 108], 10.00th=[ 114], 20.00th=[ 121], | 30.00th=[ 127], 40.00th=[ 135], 50.00th=[ 145], 60.00th=[ 155], | 70.00th=[ 169], 80.00th=[ 187], 90.00th=[ 223], 95.00th=[ 266], | 99.00th=[ 382], 99.50th=[ 470], 99.90th=[69120], 99.95th=[120320], | 99.99th=[272384] lat (usec) : 50=0.01%, 100=0.87%, 250=92.83%, 500=5.85%, 750=0.06% lat (usec) : 1000=0.01% lat (msec) : 2=0.02%, 4=0.01%, 10=0.02%, 20=0.08%, 50=0.12% lat (msec) : 100=0.07%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.70%, sys=21.56%, ctx=9637603, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=18069579/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=152249MB, aggrb=173225KB/s, minb=80309KB/s, maxb=92915KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=38952573/29, merge=0/0, ticks=13950768/2216, in_queue=13969108, util=100.00%, aggrios=13013773/192960, aggrmerge=0/0, aggrticks=5424846/37388, aggrin_queue=5466722, aggrutil=70.38% dm-1: ios=50/513251, merge=0/0, ticks=0/63836, in_queue=63844, util=4.22%, aggrios=38843781/560322, aggrmerge=0/17355, aggrticks=5786824/73580, aggrin_queue=5863840, aggrutil=72.24% xvdc: ios=38843781/560322, merge=0/17355, ticks=5786824/73580, in_queue=5863840, util=72.24% dm-0: ios=38843731/64559, merge=0/0, ticks=5772804/12088, in_queue=5798340, util=70.38% dm-3: ios=197538/1071, merge=0/0, ticks=10501736/36240, in_queue=10537984, util=66.09%, aggrios=197538/1068, aggrmerge=0/3, aggrticks=10501736/35036, aggrin_queue=10536764, aggrutil=66.09% xvdd: ios=197538/1068, merge=0/3, ticks=10501736/35036, in_queue=10536764, util=66.09% 4.12.0-rc7+458bc30cec26 (2017-07-04 22:35:16 +0100)

Over 20k average IOPS on the run against a full cache means this was marked fixed.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc7+ fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [407.1MB/0KB/0KB /s] [104K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1154: Sat Jul 29 00:38:19 2017 read : io=183121MB, bw=208350KB/s, iops=52087, runt=900001msec slat (usec): min=2, max=7756, avg= 6.96, stdev= 2.90 clat (usec): min=44, max=1200.4K, avg=144.14, stdev=472.71 lat (usec): min=71, max=1200.4K, avg=151.59, stdev=472.73 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 110], 10.00th=[ 114], 20.00th=[ 120], | 30.00th=[ 125], 40.00th=[ 131], 50.00th=[ 135], 60.00th=[ 143], | 70.00th=[ 151], 80.00th=[ 161], 90.00th=[ 181], 95.00th=[ 201], | 99.00th=[ 262], 99.50th=[ 298], 99.90th=[ 418], 99.95th=[ 466], | 99.99th=[ 1064] lat (usec) : 50=0.01%, 100=0.11%, 250=98.52%, 500=1.34%, 750=0.02% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=12.82%, sys=49.46%, ctx=19495852, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=46878856/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1155: Sat Jul 29 00:38:19 2017 read : io=173527MB, bw=197435KB/s, iops=49358, runt=900001msec slat (usec): min=2, max=4008, avg= 6.91, stdev= 2.74 clat (usec): min=1, max=1231.4K, avg=152.67, stdev=790.02 lat (usec): min=70, max=1231.4K, avg=160.07, stdev=790.13 clat percentiles (usec): | 1.00th=[ 105], 5.00th=[ 110], 10.00th=[ 115], 20.00th=[ 121], | 30.00th=[ 126], 40.00th=[ 133], 50.00th=[ 137], 60.00th=[ 143], | 70.00th=[ 153], 80.00th=[ 163], 90.00th=[ 183], 95.00th=[ 205], | 99.00th=[ 266], 99.50th=[ 298], 99.90th=[ 430], 99.95th=[ 548], | 99.99th=[24192] lat (usec) : 2=0.01%, 50=0.01%, 100=0.10%, 250=98.42%, 500=1.42% lat (usec) : 750=0.02%, 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=12.64%, sys=46.41%, ctx=18940551, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=44422812/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=356647MB, aggrb=405784KB/s, minb=197434KB/s, maxb=208350KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=91291408/28, merge=0/0, ticks=13391864/1088, in_queue=13429116, util=100.00%, aggrios=30455976/37704, aggrmerge=0/0, aggrticks=4468269/553084, aggrin_queue=5032300, aggrutil=100.00% dm-1: ios=20/46840, merge=0/0, ticks=4/8444, in_queue=8452, util=0.72%, aggrios=91341450/45704, aggrmerge=35/6871, aggrticks=12941984/9032, aggrin_queue=12962160, aggrutil=100.00% xvdc: ios=91341450/45704, merge=35/6871, ticks=12941984/9032, in_queue=12962160, util=100.00% dm-0: ios=91341465/6637, merge=0/0, ticks=12928860/1388, in_queue=12963072, util=100.00% dm-3: ios=26445/59635, merge=0/0, ticks=475944/1649420, in_queue=2125376, util=22.10%, aggrios=26445/59598, aggrmerge=0/37, aggrticks=475956/1637996, aggrin_queue=2113944, aggrutil=22.10% xvdd: ios=26445/59598, merge=0/37, ticks=475956/1637996, in_queue=2113944, util=22.10% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc7+ fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=0): [f(2)] [100.0% done] [395.1MB/0KB/0KB /s] [101K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1692: Sat Jul 29 01:03:52 2017 read : io=80360MB, bw=91432KB/s, iops=22857, runt=900001msec slat (usec): min=2, max=2817, avg= 6.86, stdev= 2.92 clat (usec): min=29, max=1517.4K, avg=340.64, stdev=5499.29 lat (usec): min=70, max=1517.4K, avg=347.99, stdev=5499.77 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 108], 10.00th=[ 112], 20.00th=[ 120], | 30.00th=[ 124], 40.00th=[ 129], 50.00th=[ 135], 60.00th=[ 143], | 70.00th=[ 153], 80.00th=[ 171], 90.00th=[ 205], 95.00th=[ 247], | 99.00th=[ 346], 99.50th=[ 414], 99.90th=[57600], 99.95th=[104960], | 99.99th=[248832] lat (usec) : 50=0.01%, 100=0.11%, 250=95.21%, 500=4.30%, 750=0.03% lat (usec) : 1000=0.02% lat (msec) : 2=0.03%, 4=0.01%, 10=0.01%, 20=0.07%, 50=0.11% lat (msec) : 100=0.06%, 250=0.04%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.73%, sys=21.81%, ctx=10535581, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=20572128/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1693: Sat Jul 29 01:03:52 2017 read : io=78767MB, bw=89620KB/s, iops=22404, runt=900001msec slat (usec): min=2, max=4079, avg= 6.86, stdev= 3.00 clat (usec): min=37, max=1305.7K, avg=347.72, stdev=5796.35 lat (usec): min=70, max=1305.7K, avg=355.07, stdev=5796.74 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 108], 10.00th=[ 112], 20.00th=[ 119], | 30.00th=[ 124], 40.00th=[ 129], 50.00th=[ 135], 60.00th=[ 141], | 70.00th=[ 149], 80.00th=[ 161], 90.00th=[ 181], 95.00th=[ 203], | 99.00th=[ 270], 99.50th=[ 350], 99.90th=[62208], 99.95th=[111104], | 99.99th=[264192] lat (usec) : 50=0.01%, 100=0.14%, 250=98.33%, 500=1.15%, 750=0.02% lat (usec) : 1000=0.01% lat (msec) : 2=0.02%, 4=0.01%, 10=0.02%, 20=0.07%, 50=0.11% lat (msec) : 100=0.06%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.40%, sys=21.50%, ctx=9823341, majf=0, minf=14 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=20164463/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=159127MB, aggrb=181051KB/s, minb=89619KB/s, maxb=91431KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=40711294/29, merge=0/0, ticks=13985492/1604, in_queue=14002044, util=100.00%, aggrios=13600731/192513, aggrmerge=0/0, aggrticks=5441234/38712, aggrin_queue=5484525, aggrutil=70.85% dm-1: ios=26/511933, merge=0/0, ticks=4/63176, in_queue=63180, util=4.35%, aggrios=40605636/559333, aggrmerge=1/16864, aggrticks=5853336/73380, aggrin_queue=5930420, aggrutil=72.72% xvdc: ios=40605636/559333, merge=1/16864, ticks=5853336/73380, in_queue=5930420, util=72.72% dm-0: ios=40605611/64403, merge=0/0, ticks=5848084/12288, in_queue=5874032, util=70.85% dm-3: ios=196556/1204, merge=0/0, ticks=10475616/40672, in_queue=10516364, util=65.70%, aggrios=196556/1201, aggrmerge=0/3, aggrticks=10475564/39108, aggrin_queue=10514664, aggrutil=65.70% xvdd: ios=196556/1201, merge=0/3, ticks=10475564/39108, in_queue=10514664, util=65.70% 4.12.0-rc4+458bc30cec26 (2017-07-04 22:35:16 +0100)

Over 20k average IOPS on the run against a full cache means this was marked fixed.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc4+ fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [395.4MB/0KB/0KB /s] [101K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1087: Sat Jul 29 02:19:19 2017 read : io=175192MB, bw=199329KB/s, iops=49832, runt=900001msec slat (usec): min=0, max=3864, avg= 6.57, stdev= 3.22 clat (usec): min=30, max=267928, avg=151.65, stdev=87.18 lat (usec): min=69, max=267942, avg=158.71, stdev=87.17 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 110], 10.00th=[ 114], 20.00th=[ 121], | 30.00th=[ 127], 40.00th=[ 133], 50.00th=[ 139], 60.00th=[ 145], | 70.00th=[ 155], 80.00th=[ 171], 90.00th=[ 203], 95.00th=[ 245], | 99.00th=[ 334], 99.50th=[ 374], 99.90th=[ 462], 99.95th=[ 516], | 99.99th=[ 1096] lat (usec) : 50=0.01%, 100=0.13%, 250=95.23%, 500=4.58%, 750=0.04% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01% cpu : usr=11.80%, sys=44.52%, ctx=20627701, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=44849083/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1088: Sat Jul 29 02:19:19 2017 read : io=169484MB, bw=192835KB/s, iops=48208, runt=900001msec slat (usec): min=0, max=920, avg= 6.65, stdev= 3.13 clat (usec): min=27, max=395221, avg=156.93, stdev=712.11 lat (usec): min=64, max=395249, avg=164.07, stdev=712.24 clat percentiles (usec): | 1.00th=[ 92], 5.00th=[ 108], 10.00th=[ 113], 20.00th=[ 120], | 30.00th=[ 125], 40.00th=[ 131], 50.00th=[ 137], 60.00th=[ 143], | 70.00th=[ 153], 80.00th=[ 167], 90.00th=[ 197], 95.00th=[ 235], | 99.00th=[ 322], 99.50th=[ 358], 99.90th=[ 466], 99.95th=[ 636], | 99.99th=[29056] lat (usec) : 50=0.01%, 100=1.35%, 250=94.81%, 500=3.76%, 750=0.03% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01% cpu : usr=11.83%, sys=43.37%, ctx=19677405, majf=0, minf=14 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=43387901/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=344676MB, aggrb=392163KB/s, minb=192834KB/s, maxb=199329KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=88216016/26, merge=0/0, ticks=13519572/1384, in_queue=13556832, util=100.00%, aggrios=29434288/40523, aggrmerge=0/0, aggrticks=4526756/395165, aggrin_queue=4932334, aggrutil=100.00% dm-1: ios=16/55677, merge=0/0, ticks=0/10000, in_queue=10004, util=0.85%, aggrios=88278149/52300, aggrmerge=0/9623, aggrticks=13075608/10432, aggrin_queue=13103668, aggrutil=100.00% xvdc: ios=88278149/52300, merge=0/9623, ticks=13075608/10432, in_queue=13103668, util=100.00% dm-0: ios=88278133/7958, merge=0/0, ticks=13062252/1688, in_queue=13095112, util=100.00% dm-3: ios=24715/57934, merge=0/0, ticks=518016/1173808, in_queue=1691888, util=19.21%, aggrios=24715/57931, aggrmerge=0/3, aggrticks=518016/1170636, aggrin_queue=1688652, aggrutil=19.21% xvdd: ios=24715/57931, merge=0/3, ticks=518016/1170636, in_queue=1688652, util=19.21% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc4+ fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=1): [f(2)] [100.0% done] [397.5MB/0KB/0KB /s] [102K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1441: Sat Jul 29 02:36:06 2017 read : io=87028MB, bw=99018KB/s, iops=24754, runt=900001msec slat (usec): min=0, max=2668, avg= 6.39, stdev= 3.49 clat (usec): min=4, max=1267.9K, avg=314.52, stdev=5200.82 lat (usec): min=70, max=1267.1K, avg=321.40, stdev=5201.22 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 108], 10.00th=[ 112], 20.00th=[ 118], | 30.00th=[ 123], 40.00th=[ 129], 50.00th=[ 133], 60.00th=[ 139], | 70.00th=[ 147], 80.00th=[ 157], 90.00th=[ 179], 95.00th=[ 201], | 99.00th=[ 274], 99.50th=[ 354], 99.90th=[52480], 99.95th=[95744], | 99.99th=[236544] lat (usec) : 10=0.01%, 50=0.01%, 100=0.15%, 250=98.28%, 500=1.22% lat (usec) : 750=0.02%, 1000=0.02% lat (msec) : 2=0.02%, 4=0.01%, 10=0.01%, 20=0.07%, 50=0.10% lat (msec) : 100=0.06%, 250=0.04%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.77%, sys=21.69%, ctx=10852657, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=22279060/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1442: Sat Jul 29 02:36:06 2017 read : io=75046MB, bw=85385KB/s, iops=21346, runt=900001msec slat (usec): min=0, max=2520, avg= 6.59, stdev= 3.40 clat (usec): min=18, max=1428.8K, avg=365.77, stdev=5975.46 lat (usec): min=70, max=1428.8K, avg=372.86, stdev=5975.77 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 111], 10.00th=[ 116], 20.00th=[ 122], | 30.00th=[ 127], 40.00th=[ 133], 50.00th=[ 139], 60.00th=[ 145], | 70.00th=[ 155], 80.00th=[ 173], 90.00th=[ 207], 95.00th=[ 249], | 99.00th=[ 350], 99.50th=[ 426], 99.90th=[63744], 99.95th=[112128], | 99.99th=[264192] lat (usec) : 20=0.01%, 50=0.01%, 100=0.11%, 250=94.94%, 500=4.55% lat (usec) : 750=0.03%, 1000=0.01% lat (msec) : 2=0.02%, 4=0.01%, 10=0.02%, 20=0.08%, 50=0.12% lat (msec) : 100=0.07%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.28%, sys=19.54%, ctx=9973368, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=19211686/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=162073MB, aggrb=184403KB/s, minb=85385KB/s, maxb=99017KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=41467013/28, merge=0/0, ticks=13982352/436, in_queue=14002096, util=100.00%, aggrios=13852123/192595, aggrmerge=0/0, aggrticks=5439580/41926, aggrin_queue=5486213, aggrutil=71.52% dm-1: ios=40/512174, merge=0/0, ticks=16/65512, in_queue=65540, util=4.37%, aggrios=41359847/559061, aggrmerge=1/17352, aggrticks=5953780/75680, aggrin_queue=6035140, aggrutil=73.41% xvdc: ios=41359847/559061, merge=1/17352, ticks=5953780/75680, in_queue=6035140, util=73.41% dm-0: ios=41359808/64396, merge=0/0, ticks=5949272/12428, in_queue=5975644, util=71.52% dm-3: ios=196521/1217, merge=0/0, ticks=10369452/47840, in_queue=10417456, util=65.41%, aggrios=196521/1215, aggrmerge=0/2, aggrticks=10369400/46792, aggrin_queue=10416188, aggrutil=65.41% xvdd: ios=196521/1215, merge=0/2, ticks=10369400/46792, in_queue=10416188, util=65.41% 4.12.0-rc2+b870aa901f4b (2017-05-31 19:29:48 -0400)

Over 20k average IOPS on the run against a full cache means this was marked fixed.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc2+ fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [387.1MB/0KB/0KB /s] [99.3K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1034: Sat Jul 29 03:42:15 2017 read : io=170714MB, bw=194235KB/s, iops=48558, runt=900001msec slat (usec): min=0, max=5188, avg= 7.17, stdev= 3.84 clat (usec): min=33, max=868253, avg=154.79, stdev=246.43 lat (usec): min=74, max=868258, avg=162.45, stdev=246.45 clat percentiles (usec): | 1.00th=[ 106], 5.00th=[ 112], 10.00th=[ 118], 20.00th=[ 125], | 30.00th=[ 131], 40.00th=[ 137], 50.00th=[ 143], 60.00th=[ 151], | 70.00th=[ 161], 80.00th=[ 175], 90.00th=[ 203], 95.00th=[ 239], | 99.00th=[ 326], 99.50th=[ 362], 99.90th=[ 454], 99.95th=[ 510], | 99.99th=[ 1176] lat (usec) : 50=0.01%, 100=0.07%, 250=95.81%, 500=4.06%, 750=0.03% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% cpu : usr=13.58%, sys=46.69%, ctx=17639239, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=43702872/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1035: Sat Jul 29 03:42:15 2017 read : io=166633MB, bw=189591KB/s, iops=47397, runt=900001msec slat (usec): min=0, max=5427, avg= 7.20, stdev= 3.86 clat (usec): min=25, max=1052.1K, avg=158.78, stdev=971.36 lat (usec): min=67, max=1052.1K, avg=166.47, stdev=971.50 clat percentiles (usec): | 1.00th=[ 105], 5.00th=[ 111], 10.00th=[ 116], 20.00th=[ 123], | 30.00th=[ 129], 40.00th=[ 135], 50.00th=[ 141], 60.00th=[ 147], | 70.00th=[ 157], 80.00th=[ 169], 90.00th=[ 191], 95.00th=[ 211], | 99.00th=[ 274], 99.50th=[ 310], 99.90th=[ 454], 99.95th=[ 740], | 99.99th=[31104] lat (usec) : 50=0.01%, 100=0.09%, 250=98.15%, 500=1.69%, 750=0.02% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01%, 2000=0.01% cpu : usr=13.55%, sys=45.80%, ctx=17358595, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=42657995/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=337347MB, aggrb=383825KB/s, minb=189590KB/s, maxb=194234KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=86351783/28, merge=0/0, ticks=13253620/1212, in_queue=13300668, util=100.00%, aggrios=28809668/39785, aggrmerge=0/0, aggrticks=4427297/502156, aggrin_queue=4941461, aggrutil=100.00% dm-1: ios=23/51212, merge=0/0, ticks=0/9628, in_queue=9632, util=0.83%, aggrios=86400786/48163, aggrmerge=36/10034, aggrticks=12739236/9992, aggrin_queue=12763444, aggrutil=100.00% xvdc: ios=86400786/48163, merge=36/10034, ticks=12739236/9992, in_queue=12763444, util=100.00% dm-0: ios=86400799/7291, merge=0/0, ticks=12710904/1508, in_queue=12748372, util=100.00% dm-3: ios=28182/60853, merge=0/0, ticks=570988/1495332, in_queue=2066380, util=21.95%, aggrios=28182/60807, aggrmerge=0/46, aggrticks=570992/1486948, aggrin_queue=2057924, aggrutil=21.95% xvdd: ios=28182/60807, merge=0/46, ticks=570992/1486948, in_queue=2057924, util=21.95% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc2+ fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [373.7MB/0KB/0KB /s] [95.7K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1530: Sat Jul 29 03:59:03 2017 read : io=76723MB, bw=87294KB/s, iops=21823, runt=900001msec slat (usec): min=0, max=2653, avg= 6.62, stdev= 4.06 clat (usec): min=30, max=1178.5K, avg=357.41, stdev=5505.55 lat (usec): min=70, max=1178.5K, avg=364.48, stdev=5505.64 clat percentiles (usec): | 1.00th=[ 105], 5.00th=[ 112], 10.00th=[ 117], 20.00th=[ 124], | 30.00th=[ 131], 40.00th=[ 135], 50.00th=[ 143], 60.00th=[ 151], | 70.00th=[ 163], 80.00th=[ 183], 90.00th=[ 217], 95.00th=[ 262], | 99.00th=[ 374], 99.50th=[ 458], 99.90th=[61696], 99.95th=[109056], | 99.99th=[246784] lat (usec) : 50=0.01%, 100=0.09%, 250=93.93%, 500=5.55%, 750=0.06% lat (usec) : 1000=0.02% lat (msec) : 2=0.03%, 4=0.01%, 10=0.01%, 20=0.08%, 50=0.11% lat (msec) : 100=0.07%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.69%, sys=19.74%, ctx=9908053, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=19641135/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1531: Sat Jul 29 03:59:03 2017 read : io=74239MB, bw=84468KB/s, iops=21116, runt=900001msec slat (usec): min=0, max=317, avg= 6.86, stdev= 3.89 clat (usec): min=24, max=1537.3K, avg=369.35, stdev=5950.92 lat (usec): min=71, max=1537.3K, avg=376.68, stdev=5950.98 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 109], 10.00th=[ 113], 20.00th=[ 120], | 30.00th=[ 126], 40.00th=[ 131], 50.00th=[ 137], 60.00th=[ 145], | 70.00th=[ 157], 80.00th=[ 173], 90.00th=[ 207], 95.00th=[ 247], | 99.00th=[ 350], 99.50th=[ 434], 99.90th=[66048], 99.95th=[116224], | 99.99th=[264192] lat (usec) : 50=0.01%, 100=0.16%, 250=95.11%, 500=4.32%, 750=0.04% lat (usec) : 1000=0.01% lat (msec) : 2=0.02%, 4=0.01%, 10=0.02%, 20=0.08%, 50=0.12% lat (msec) : 100=0.07%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.71%, sys=19.66%, ctx=9087432, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=19005271/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=150963MB, aggrb=171761KB/s, minb=84467KB/s, maxb=87293KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=38629806/29, merge=0/0, ticks=13959752/3564, in_queue=13980884, util=100.00%, aggrios=12904005/192734, aggrmerge=0/0, aggrticks=5435258/41146, aggrin_queue=5481312, aggrutil=71.34% dm-1: ios=57/512619, merge=0/0, ticks=4/65252, in_queue=65272, util=4.40%, aggrios=38515218/559828, aggrmerge=0/17125, aggrticks=5837752/75716, aggrin_queue=5921004, aggrutil=73.26% xvdc: ios=38515218/559828, merge=0/17125, ticks=5837752/75716, in_queue=5921004, util=73.26% dm-0: ios=38515161/64478, merge=0/0, ticks=5828092/12596, in_queue=5855388, util=71.34% dm-3: ios=196799/1107, merge=0/0, ticks=10477680/45592, in_queue=10523276, util=65.98%, aggrios=196799/1100, aggrmerge=0/7, aggrticks=10477632/43520, aggrin_queue=10521220, aggrutil=65.98% xvdd: ios=196799/1100, merge=0/7, ticks=10477632/43520, in_queue=10521220, util=65.98% 4.12.0-rc2+6ce478291137 (2017-05-26 10:51:18 -0700)

Over 20k average IOPS on the run against a full cache means this was marked fixed.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc2+6ce478291137 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [398.7MB/0KB/0KB /s] [102K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1324: Sat Jul 29 05:18:27 2017 read : io=181395MB, bw=206387KB/s, iops=51596, runt=900001msec slat (usec): min=0, max=3628, avg= 6.85, stdev= 3.18 clat (usec): min=31, max=650283, avg=145.71, stdev=401.45 lat (usec): min=70, max=650322, avg=153.04, stdev=401.53 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 110], 10.00th=[ 114], 20.00th=[ 120], | 30.00th=[ 125], 40.00th=[ 131], 50.00th=[ 135], 60.00th=[ 141], | 70.00th=[ 149], 80.00th=[ 161], 90.00th=[ 179], 95.00th=[ 201], | 99.00th=[ 266], 99.50th=[ 302], 99.90th=[ 430], 99.95th=[ 502], | 99.99th=[ 1704] lat (usec) : 50=0.01%, 100=0.17%, 250=98.42%, 500=1.36%, 750=0.02% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01% cpu : usr=15.23%, sys=53.43%, ctx=19704592, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=46437113/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1325: Sat Jul 29 05:18:27 2017 read : io=163235MB, bw=185725KB/s, iops=46431, runt=900001msec slat (usec): min=0, max=7249, avg= 7.00, stdev= 3.64 clat (usec): min=29, max=321686, avg=162.76, stdev=759.82 lat (usec): min=69, max=321717, avg=170.23, stdev=759.98 clat percentiles (usec): | 1.00th=[ 105], 5.00th=[ 112], 10.00th=[ 117], 20.00th=[ 124], | 30.00th=[ 129], 40.00th=[ 135], 50.00th=[ 141], 60.00th=[ 147], | 70.00th=[ 157], 80.00th=[ 173], 90.00th=[ 203], 95.00th=[ 243], | 99.00th=[ 330], 99.50th=[ 370], 99.90th=[ 474], 99.95th=[ 716], | 99.99th=[32128] lat (usec) : 50=0.01%, 100=0.12%, 250=95.44%, 500=4.36%, 750=0.03% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01% cpu : usr=13.56%, sys=49.51%, ctx=18629816, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=41788282/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=344630MB, aggrb=392112KB/s, minb=185725KB/s, maxb=206386KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=88203561/28, merge=0/0, ticks=13332148/1696, in_queue=13367996, util=100.00%, aggrios=29431008/44535, aggrmerge=0/0, aggrticks=4467504/392230, aggrin_queue=4869529, aggrutil=100.00% dm-1: ios=13/65959, merge=0/0, ticks=4/11628, in_queue=11636, util=0.98%, aggrios=88262702/62581, aggrmerge=0/10717, aggrticks=12756980/12036, aggrin_queue=12776352, aggrutil=100.00% xvdc: ios=88262702/62581, merge=0/10717, ticks=12756980/12036, in_queue=12776352, util=100.00% dm-0: ios=88262689/9237, merge=0/0, ticks=12731748/1756, in_queue=12762868, util=100.00% dm-3: ios=30323/58410, merge=0/0, ticks=670760/1163308, in_queue=1834084, util=20.15%, aggrios=30323/58407, aggrmerge=0/3, aggrticks=670772/1159656, aggrin_queue=1830424, aggrutil=20.15% xvdd: ios=30323/58407, merge=0/3, ticks=670772/1159656, in_queue=1830424, util=20.15% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc2+6ce478291137 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [364.8MB/0KB/0KB /s] [93.4K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1722: Sat Jul 29 05:41:43 2017 read : io=74359MB, bw=84603KB/s, iops=21150, runt=900001msec slat (usec): min=0, max=2436, avg= 6.76, stdev= 3.33 clat (usec): min=37, max=1305.6K, avg=369.04, stdev=5594.01 lat (usec): min=68, max=1305.9K, avg=376.26, stdev=5594.50 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 109], 10.00th=[ 113], 20.00th=[ 120], | 30.00th=[ 124], 40.00th=[ 129], 50.00th=[ 135], 60.00th=[ 143], | 70.00th=[ 155], 80.00th=[ 179], 90.00th=[ 258], 95.00th=[ 346], | 99.00th=[ 532], 99.50th=[ 636], 99.90th=[63232], 99.95th=[110080], | 99.99th=[250880] lat (usec) : 50=0.01%, 100=0.17%, 250=89.28%, 500=9.28%, 750=0.88% lat (usec) : 1000=0.03% lat (msec) : 2=0.03%, 4=0.01%, 10=0.02%, 20=0.08%, 50=0.12% lat (msec) : 100=0.07%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=6.42%, sys=22.97%, ctx=10466417, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=19035803/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1723: Sat Jul 29 05:41:43 2017 read : io=75299MB, bw=85674KB/s, iops=21418, runt=900001msec slat (usec): min=0, max=2420, avg= 6.72, stdev= 3.30 clat (usec): min=32, max=1379.2K, avg=364.38, stdev=5736.94 lat (usec): min=71, max=1379.2K, avg=371.55, stdev=5737.35 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 108], 10.00th=[ 111], 20.00th=[ 117], | 30.00th=[ 122], 40.00th=[ 127], 50.00th=[ 133], 60.00th=[ 141], | 70.00th=[ 149], 80.00th=[ 167], 90.00th=[ 205], 95.00th=[ 274], | 99.00th=[ 490], 99.50th=[ 612], 99.90th=[64768], 99.95th=[113152], | 99.99th=[259072] lat (usec) : 50=0.01%, 100=0.27%, 250=93.50%, 500=5.31%, 750=0.54% lat (usec) : 1000=0.02% lat (msec) : 2=0.02%, 4=0.01%, 10=0.02%, 20=0.08%, 50=0.12% lat (msec) : 100=0.07%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=6.67%, sys=22.54%, ctx=10228527, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=19276639/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=149658MB, aggrb=170277KB/s, minb=84603KB/s, maxb=85673KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=38294873/29, merge=0/0, ticks=13980376/4156, in_queue=13996968, util=100.00%, aggrios=12792699/192601, aggrmerge=0/0, aggrticks=5453128/37806, aggrin_queue=5494990, aggrutil=70.72% dm-1: ios=86/512203, merge=0/0, ticks=16/61356, in_queue=61388, util=4.27%, aggrios=38180502/560167, aggrmerge=1/16429, aggrticks=5907852/71764, aggrin_queue=5984152, aggrutil=72.60% xvdc: ios=38180502/560167, merge=1/16429, ticks=5907852/71764, in_queue=5984152, util=72.60% dm-0: ios=38180417/64538, merge=0/0, ticks=5896096/12280, in_queue=5920524, util=70.72% dm-3: ios=197595/1063, merge=0/0, ticks=10463272/39784, in_queue=10503060, util=65.39%, aggrios=197595/1056, aggrmerge=0/7, aggrticks=10463272/36948, aggrin_queue=10500280, aggrutil=65.39% xvdd: ios=197595/1056, merge=0/7, ticks=10463272/36948, in_queue=10500280, util=65.39% 4.12.0-rc1+331da109ec20 (2017-05-20 08:44:22 -0700)

Over 20k average IOPS on the run against a full cache means this was marked fixed.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc1+331da109ec20 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [378.6MB/0KB/0KB /s] [96.1K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1342: Sat Jul 29 06:54:00 2017 read : io=168144MB, bw=191310KB/s, iops=47827, runt=900001msec slat (usec): min=2, max=2850, avg= 6.93, stdev= 2.60 clat (usec): min=39, max=1150.2K, avg=157.80, stdev=403.97 lat (usec): min=70, max=1150.2K, avg=165.21, stdev=404.00 clat percentiles (usec): | 1.00th=[ 105], 5.00th=[ 111], 10.00th=[ 116], 20.00th=[ 124], | 30.00th=[ 131], 40.00th=[ 137], 50.00th=[ 145], 60.00th=[ 153], | 70.00th=[ 165], 80.00th=[ 181], 90.00th=[ 211], 95.00th=[ 247], | 99.00th=[ 330], 99.50th=[ 370], 99.90th=[ 466], 99.95th=[ 532], | 99.99th=[ 1336] lat (usec) : 50=0.01%, 100=0.08%, 250=95.27%, 500=4.58%, 750=0.04% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=12.05%, sys=45.56%, ctx=19112237, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=43044758/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1343: Sat Jul 29 06:54:00 2017 read : io=162387MB, bw=184760KB/s, iops=46190, runt=900001msec slat (usec): min=2, max=3771, avg= 6.95, stdev= 2.70 clat (usec): min=18, max=528413, avg=163.71, stdev=689.58 lat (usec): min=70, max=528456, avg=171.14, stdev=689.73 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 110], 10.00th=[ 115], 20.00th=[ 122], | 30.00th=[ 129], 40.00th=[ 135], 50.00th=[ 143], 60.00th=[ 151], | 70.00th=[ 163], 80.00th=[ 181], 90.00th=[ 211], 95.00th=[ 247], | 99.00th=[ 334], 99.50th=[ 374], 99.90th=[ 482], 99.95th=[ 708], | 99.99th=[27520] lat (usec) : 20=0.01%, 50=0.01%, 100=0.11%, 250=95.14%, 500=4.67% lat (usec) : 750=0.04%, 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01% cpu : usr=12.07%, sys=43.77%, ctx=18853321, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=41571072/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=330531MB, aggrb=376069KB/s, minb=184760KB/s, maxb=191309KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=84604087/28, merge=0/0, ticks=13465432/868, in_queue=13500164, util=100.00%, aggrios=28227541/40723, aggrmerge=0/0, aggrticks=4500645/520276, aggrin_queue=5032592, aggrutil=100.00% dm-1: ios=25/55370, merge=0/0, ticks=4/10480, in_queue=10484, util=0.88%, aggrios=84654863/53148, aggrmerge=27/9173, aggrticks=12992604/11056, aggrin_queue=13010688, aggrutil=100.00% xvdc: ios=84654863/53148, merge=27/9173, ticks=12992604/11056, in_queue=13010688, util=100.00% dm-0: ios=84654865/7809, merge=0/0, ticks=12987816/1676, in_queue=13024444, util=100.00% dm-3: ios=27734/58990, merge=0/0, ticks=514116/1548672, in_queue=2062848, util=22.31%, aggrios=27734/58961, aggrmerge=0/29, aggrticks=514100/1541080, aggrin_queue=2055176, aggrutil=22.31% xvdd: ios=27734/58961, merge=0/29, ticks=514100/1541080, in_queue=2055176, util=22.31% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc1+331da109ec20 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [409.7MB/0KB/0KB /s] [105K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1729: Sat Jul 29 07:10:43 2017 read : io=84440MB, bw=96074KB/s, iops=24018, runt=900001msec slat (usec): min=2, max=2584, avg= 6.82, stdev= 2.79 clat (usec): min=16, max=1503.2K, avg=323.80, stdev=5298.35 lat (usec): min=67, max=1503.2K, avg=331.09, stdev=5298.79 clat percentiles (usec): | 1.00th=[ 79], 5.00th=[ 98], 10.00th=[ 107], 20.00th=[ 115], | 30.00th=[ 122], 40.00th=[ 127], 50.00th=[ 135], 60.00th=[ 141], | 70.00th=[ 151], 80.00th=[ 165], 90.00th=[ 189], 95.00th=[ 211], | 99.00th=[ 282], 99.50th=[ 366], 99.90th=[56064], 99.95th=[100864], | 99.99th=[238592] lat (usec) : 20=0.01%, 50=0.01%, 100=5.36%, 250=92.78%, 500=1.51% lat (usec) : 750=0.03%, 1000=0.02% lat (msec) : 2=0.02%, 4=0.01%, 10=0.01%, 20=0.07%, 50=0.10% lat (msec) : 100=0.06%, 250=0.04%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=6.09%, sys=22.54%, ctx=10504259, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=21616707/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1730: Sat Jul 29 07:10:43 2017 read : io=80128MB, bw=91168KB/s, iops=22791, runt=900001msec slat (usec): min=2, max=2258, avg= 6.89, stdev= 2.81 clat (usec): min=21, max=1497.9K, avg=341.59, stdev=5610.12 lat (usec): min=66, max=1497.9K, avg=348.97, stdev=5610.53 clat percentiles (usec): | 1.00th=[ 76], 5.00th=[ 89], 10.00th=[ 104], 20.00th=[ 113], | 30.00th=[ 121], 40.00th=[ 127], 50.00th=[ 133], 60.00th=[ 141], | 70.00th=[ 151], 80.00th=[ 167], 90.00th=[ 191], 95.00th=[ 219], | 99.00th=[ 314], 99.50th=[ 394], 99.90th=[60672], 99.95th=[108032], | 99.99th=[252928] lat (usec) : 50=0.01%, 100=8.41%, 250=88.94%, 500=2.27%, 750=0.03% lat (usec) : 1000=0.02% lat (msec) : 2=0.02%, 4=0.01%, 10=0.02%, 20=0.07%, 50=0.11% lat (msec) : 100=0.06%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.90%, sys=21.52%, ctx=9740291, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=20512763/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=164568MB, aggrb=187241KB/s, minb=91167KB/s, maxb=96074KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=42117663/29, merge=0/0, ticks=13957468/4044, in_queue=13976576, util=100.00%, aggrios=14065022/191669, aggrmerge=0/0, aggrticks=5437961/43466, aggrin_queue=5485902, aggrutil=71.18% dm-1: ios=23/509401, merge=0/0, ticks=4/64012, in_queue=64032, util=4.37%, aggrios=41998935/556367, aggrmerge=0/16995, aggrticks=5877544/74016, aggrin_queue=5952992, aggrutil=73.08% xvdc: ios=41998935/556367, merge=0/16995, ticks=5877544/74016, in_queue=5952992, util=73.08% dm-0: ios=41998912/64099, merge=0/0, ticks=5872440/12180, in_queue=5897884, util=71.18% dm-3: ios=196133/1507, merge=0/0, ticks=10441440/54208, in_queue=10495792, util=65.28%, aggrios=196133/1501, aggrmerge=0/6, aggrticks=10441400/51888, aggrin_queue=10493276, aggrutil=65.28% xvdd: ios=196133/1501, merge=0/6, ticks=10441400/51888, in_queue=10493276, util=65.28% 4.12.0-rc1+d51aff16e821 (2017-05-20 06:00:49 +1000)

Over 20k average IOPS on the run against a full cache means this was marked fixed.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc1+d51aff16e821 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [406.3MB/0KB/0KB /s] [104K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1248: Sat Jul 29 08:51:19 2017 read : io=177645MB, bw=202121KB/s, iops=50530, runt=900001msec slat (usec): min=0, max=3003, avg= 7.12, stdev= 3.03 clat (usec): min=39, max=971679, avg=148.57, stdev=681.72 lat (usec): min=70, max=971686, avg=156.15, stdev=681.79 clat percentiles (usec): | 1.00th=[ 105], 5.00th=[ 110], 10.00th=[ 115], 20.00th=[ 122], | 30.00th=[ 127], 40.00th=[ 133], 50.00th=[ 137], 60.00th=[ 145], | 70.00th=[ 151], 80.00th=[ 163], 90.00th=[ 181], 95.00th=[ 201], | 99.00th=[ 262], 99.50th=[ 302], 99.90th=[ 438], 99.95th=[ 556], | 99.99th=[ 7264] lat (usec) : 50=0.01%, 100=0.11%, 250=98.53%, 500=1.30%, 750=0.03% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% cpu : usr=15.07%, sys=52.23%, ctx=18395569, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=45477192/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1249: Sat Jul 29 08:51:19 2017 read : io=169585MB, bw=192950KB/s, iops=48237, runt=900001msec slat (usec): min=0, max=2529, avg= 7.21, stdev= 3.13 clat (usec): min=27, max=1731.8K, avg=156.00, stdev=919.90 lat (usec): min=67, max=1731.8K, avg=163.66, stdev=920.02 clat percentiles (usec): | 1.00th=[ 105], 5.00th=[ 111], 10.00th=[ 116], 20.00th=[ 123], | 30.00th=[ 129], 40.00th=[ 133], 50.00th=[ 139], 60.00th=[ 145], | 70.00th=[ 153], 80.00th=[ 165], 90.00th=[ 185], 95.00th=[ 205], | 99.00th=[ 266], 99.50th=[ 302], 99.90th=[ 450], 99.95th=[ 724], | 99.99th=[31104] lat (usec) : 50=0.01%, 100=0.19%, 250=98.36%, 500=1.38%, 750=0.03% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=14.06%, sys=50.57%, ctx=17854020, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=43413691/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=347230MB, aggrb=395070KB/s, minb=192949KB/s, maxb=202120KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=88868766/28, merge=0/0, ticks=13255048/544, in_queue=13291156, util=100.00%, aggrios=29652308/43870, aggrmerge=0/0, aggrticks=4437944/678066, aggrin_queue=5124804, aggrutil=100.00% dm-1: ios=40/65580, merge=0/0, ticks=8/11168, in_queue=11184, util=0.96%, aggrios=88925731/62252, aggrmerge=5/10546, aggrticks=12610900/11908, aggrin_queue=12628964, aggrutil=100.00% xvdc: ios=88925731/62252, merge=5/10546, ticks=12610900/11908, in_queue=12628964, util=100.00% dm-0: ios=88925696/9160, merge=0/0, ticks=12587208/1880, in_queue=12615460, util=100.00% dm-3: ios=31188/56871, merge=0/0, ticks=726616/2021152, in_queue=2747768, util=27.31%, aggrios=31188/56865, aggrmerge=0/6, aggrticks=726608/2002592, aggrin_queue=2729192, aggrutil=27.32% xvdd: ios=31188/56865, merge=0/6, ticks=726608/2002592, in_queue=2729192, util=27.32% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc1+d51aff16e821 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [384.3MB/0KB/0KB /s] [98.4K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1535: Sat Jul 29 09:10:07 2017 read : io=81142MB, bw=92322KB/s, iops=23080, runt=900001msec slat (usec): min=0, max=2500, avg= 6.89, stdev= 3.06 clat (usec): min=27, max=1250.7K, avg=337.19, stdev=5337.52 lat (usec): min=66, max=1250.8K, avg=344.53, stdev=5338.02 clat percentiles (usec): | 1.00th=[ 89], 5.00th=[ 107], 10.00th=[ 110], 20.00th=[ 118], | 30.00th=[ 123], 40.00th=[ 129], 50.00th=[ 135], 60.00th=[ 143], | 70.00th=[ 153], 80.00th=[ 171], 90.00th=[ 203], 95.00th=[ 245], | 99.00th=[ 358], 99.50th=[ 434], 99.90th=[57600], 99.95th=[103936], | 99.99th=[240640] lat (usec) : 50=0.01%, 100=1.73%, 250=93.69%, 500=4.18%, 750=0.04% lat (usec) : 1000=0.02% lat (msec) : 2=0.03%, 4=0.01%, 10=0.01%, 20=0.07%, 50=0.11% lat (msec) : 100=0.06%, 250=0.04%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=6.95%, sys=23.93%, ctx=10312344, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=20772473/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1536: Sat Jul 29 09:10:07 2017 read : io=73428MB, bw=83544KB/s, iops=20886, runt=900001msec slat (usec): min=0, max=1453, avg= 6.96, stdev= 2.98 clat (usec): min=15, max=1524.7K, avg=373.51, stdev=6231.54 lat (usec): min=67, max=1524.7K, avg=380.92, stdev=6231.92 clat percentiles (usec): | 1.00th=[ 90], 5.00th=[ 107], 10.00th=[ 112], 20.00th=[ 120], | 30.00th=[ 125], 40.00th=[ 131], 50.00th=[ 139], 60.00th=[ 145], | 70.00th=[ 157], 80.00th=[ 173], 90.00th=[ 203], 95.00th=[ 241], | 99.00th=[ 350], 99.50th=[ 434], 99.90th=[66048], 99.95th=[117248], | 99.99th=[276480] lat (usec) : 20=0.01%, 50=0.01%, 100=1.63%, 250=94.04%, 500=3.91% lat (usec) : 750=0.04%, 1000=0.02% lat (msec) : 2=0.02%, 4=0.01%, 10=0.02%, 20=0.08%, 50=0.12% lat (msec) : 100=0.07%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=6.40%, sys=21.56%, ctx=9325698, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=18797466/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=154570MB, aggrb=175866KB/s, minb=83544KB/s, maxb=92321KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=39560967/29, merge=0/0, ticks=13949328/1084, in_queue=13964452, util=100.00%, aggrios=13211863/192921, aggrmerge=0/0, aggrticks=5416640/38728, aggrin_queue=5458597, aggrutil=70.47% dm-1: ios=47/513127, merge=0/0, ticks=0/64104, in_queue=64112, util=4.32%, aggrios=39438165/560401, aggrmerge=0/17140, aggrticks=5784740/74192, aggrin_queue=5861736, aggrutil=72.34% xvdc: ios=39438165/560401, merge=0/17140, ticks=5784740/74192, in_queue=5861736, util=72.34% dm-0: ios=39438118/64534, merge=0/0, ticks=5774760/12144, in_queue=5796584, util=70.47% dm-3: ios=197426/1102, merge=0/0, ticks=10475160/39936, in_queue=10515096, util=66.94%, aggrios=197426/1101, aggrmerge=0/1, aggrticks=10475168/39392, aggrin_queue=10514556, aggrutil=66.94% xvdd: ios=197426/1101, merge=0/1, ticks=10475168/39392, in_queue=10514556, util=66.94% 4.12.0-rc1+a58a260fd96b (2017-05-18 11:21:10 -0700)

Over 20k average IOPS on the run against a full cache means this was marked fixed.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc1+a58a260fd96b fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [403.7MB/0KB/0KB /s] [103K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1469: Sat Jul 29 10:38:52 2017 read : io=178006MB, bw=202531KB/s, iops=50632, runt=900001msec slat (usec): min=0, max=6867, avg= 7.12, stdev= 3.46 clat (usec): min=32, max=1286.2K, avg=148.28, stdev=540.32 lat (usec): min=67, max=1286.2K, avg=155.87, stdev=540.36 clat percentiles (usec): | 1.00th=[ 105], 5.00th=[ 111], 10.00th=[ 116], 20.00th=[ 123], | 30.00th=[ 129], 40.00th=[ 133], 50.00th=[ 139], 60.00th=[ 145], | 70.00th=[ 153], 80.00th=[ 165], 90.00th=[ 185], 95.00th=[ 207], | 99.00th=[ 270], 99.50th=[ 306], 99.90th=[ 430], 99.95th=[ 494], | 99.99th=[ 1352] lat (usec) : 50=0.01%, 100=0.12%, 250=98.28%, 500=1.55%, 750=0.02% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=15.17%, sys=54.06%, ctx=18845932, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=45569414/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1470: Sat Jul 29 10:38:52 2017 read : io=173743MB, bw=197681KB/s, iops=49420, runt=900001msec slat (usec): min=0, max=3071, avg= 7.07, stdev= 3.36 clat (usec): min=19, max=1304.0K, avg=152.23, stdev=680.26 lat (usec): min=69, max=1304.2K, avg=159.77, stdev=680.38 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 110], 10.00th=[ 115], 20.00th=[ 122], | 30.00th=[ 127], 40.00th=[ 133], 50.00th=[ 137], 60.00th=[ 143], | 70.00th=[ 153], 80.00th=[ 163], 90.00th=[ 185], 95.00th=[ 205], | 99.00th=[ 266], 99.50th=[ 302], 99.90th=[ 438], 99.95th=[ 588], | 99.99th=[22912] lat (usec) : 20=0.01%, 50=0.01%, 100=0.17%, 250=98.29%, 500=1.48% lat (usec) : 750=0.02%, 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=15.16%, sys=52.26%, ctx=18799543, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=44478196/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=351748MB, aggrb=400211KB/s, minb=197680KB/s, maxb=202530KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=90047147/28, merge=0/0, ticks=13242192/1464, in_queue=13298304, util=100.00%, aggrios=30038236/40110, aggrmerge=0/0, aggrticks=4422056/701994, aggrin_queue=5135569, aggrutil=100.00% dm-1: ios=45/53246, merge=0/0, ticks=8/9560, in_queue=9572, util=0.81%, aggrios=90087618/50539, aggrmerge=21/8883, aggrticks=12812936/9916, aggrin_queue=12830368, aggrutil=100.00% xvdc: ios=90087618/50539, merge=21/8883, ticks=12812936/9916, in_queue=12830368, util=100.00% dm-0: ios=90087594/7476, merge=0/0, ticks=12789572/1472, in_queue=12825580, util=100.00% dm-3: ios=27070/59608, merge=0/0, ticks=476588/2094952, in_queue=2571556, util=26.41%, aggrios=27070/59583, aggrmerge=0/25, aggrticks=476596/2078296, aggrin_queue=2554892, aggrutil=26.41% xvdd: ios=27070/59583, merge=0/25, ticks=476596/2078296, in_queue=2554892, util=26.41% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc1+a58a260fd96b fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [369.8MB/0KB/0KB /s] [94.7K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1970: Sat Jul 29 11:17:21 2017 read : io=77095MB, bw=87717KB/s, iops=21929, runt=900001msec slat (usec): min=0, max=2471, avg= 6.97, stdev= 3.30 clat (usec): min=4, max=1523.7K, avg=355.35, stdev=5578.26 lat (usec): min=64, max=1523.7K, avg=362.77, stdev=5578.75 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 108], 10.00th=[ 111], 20.00th=[ 120], | 30.00th=[ 125], 40.00th=[ 133], 50.00th=[ 143], 60.00th=[ 153], | 70.00th=[ 167], 80.00th=[ 185], 90.00th=[ 213], 95.00th=[ 247], | 99.00th=[ 346], 99.50th=[ 426], 99.90th=[61184], 99.95th=[108032], | 99.99th=[248832] lat (usec) : 10=0.01%, 50=0.01%, 100=0.83%, 250=94.43%, 500=4.33% lat (usec) : 750=0.04%, 1000=0.02% lat (msec) : 2=0.03%, 4=0.01%, 10=0.02%, 20=0.08%, 50=0.11% lat (msec) : 100=0.07%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=6.49%, sys=24.26%, ctx=10152879, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=19736266/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1971: Sat Jul 29 11:17:21 2017 read : io=70627MB, bw=80357KB/s, iops=20089, runt=900001msec slat (usec): min=0, max=1736, avg= 6.99, stdev= 3.37 clat (usec): min=17, max=1334.4K, avg=388.69, stdev=6080.75 lat (usec): min=62, max=1334.4K, avg=396.14, stdev=6081.19 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 108], 10.00th=[ 114], 20.00th=[ 122], | 30.00th=[ 129], 40.00th=[ 137], 50.00th=[ 147], 60.00th=[ 159], | 70.00th=[ 173], 80.00th=[ 191], 90.00th=[ 225], 95.00th=[ 262], | 99.00th=[ 366], 99.50th=[ 454], 99.90th=[69120], 99.95th=[119296], | 99.99th=[272384] lat (usec) : 20=0.01%, 50=0.01%, 100=0.76%, 250=92.96%, 500=5.83% lat (usec) : 750=0.04%, 1000=0.02% lat (msec) : 2=0.02%, 4=0.01%, 10=0.02%, 20=0.08%, 50=0.12% lat (msec) : 100=0.07%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.95%, sys=22.37%, ctx=9542001, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=18080386/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=147721MB, aggrb=168073KB/s, minb=80357KB/s, maxb=87716KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=37811927/29, merge=0/0, ticks=13950924/2420, in_queue=13976400, util=100.00%, aggrios=12627432/192884, aggrmerge=0/0, aggrticks=5423964/40908, aggrin_queue=5469672, aggrutil=70.48% dm-1: ios=41/513019, merge=0/0, ticks=8/61768, in_queue=61776, util=4.19%, aggrios=37684593/560591, aggrmerge=0/16798, aggrticks=5810780/72380, aggrin_queue=5885452, aggrutil=72.25% xvdc: ios=37684593/560591, merge=0/16798, ticks=5810780/72380, in_queue=5885452, util=72.25% dm-0: ios=37684552/64512, merge=0/0, ticks=5808316/12596, in_queue=5835296, util=70.48% dm-3: ios=197703/1122, merge=0/0, ticks=10463568/48360, in_queue=10511944, util=65.97%, aggrios=197703/1119, aggrmerge=0/3, aggrticks=10463584/46584, aggrin_queue=10510148, aggrutil=65.97% xvdd: ios=197703/1119, merge=0/3, ticks=10463584/46584, in_queue=10510148, util=65.97% 4.12.0-rc1+243bfd2c052d (2017-05-17 14:13:44 -0700)

just 2,130 and 738 average IOPS on the run against a full cache means this was marked broken.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc1+243bfd2c052d fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [223.5MB/0KB/0KB /s] [57.2K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1033: Sat Jul 29 13:21:34 2017 read : io=136455MB, bw=155242KB/s, iops=38810, runt=900078msec slat (usec): min=0, max=26662, avg= 4.82, stdev= 6.30 clat (usec): min=21, max=880182, avg=199.68, stdev=2322.99 lat (usec): min=70, max=880190, avg=204.84, stdev=2323.29 clat percentiles (usec): | 1.00th=[ 102], 5.00th=[ 106], 10.00th=[ 109], 20.00th=[ 114], | 30.00th=[ 119], 40.00th=[ 122], 50.00th=[ 126], 60.00th=[ 131], | 70.00th=[ 137], 80.00th=[ 149], 90.00th=[ 189], 95.00th=[ 245], | 99.00th=[ 382], 99.50th=[ 446], 99.90th=[17792], 99.95th=[36608], | 99.99th=[108032] lat (usec) : 50=0.01%, 100=0.24%, 250=95.07%, 500=4.31%, 750=0.09% lat (usec) : 1000=0.05% lat (msec) : 2=0.11%, 4=0.01%, 10=0.01%, 20=0.03%, 50=0.06% lat (msec) : 100=0.02%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% cpu : usr=7.38%, sys=26.41%, ctx=19119794, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=34932423/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1034: Sat Jul 29 13:21:34 2017 read : io=42407MB, bw=48247KB/s, iops=12061, runt=900066msec slat (usec): min=0, max=236, avg= 7.12, stdev= 4.37 clat (usec): min=2, max=1038.3K, avg=653.49, stdev=6208.11 lat (usec): min=71, max=1038.3K, avg=661.11, stdev=6209.00 clat percentiles (usec): | 1.00th=[ 100], 5.00th=[ 104], 10.00th=[ 106], 20.00th=[ 109], | 30.00th=[ 113], 40.00th=[ 117], 50.00th=[ 120], 60.00th=[ 124], | 70.00th=[ 129], 80.00th=[ 137], 90.00th=[ 163], 95.00th=[ 237], | 99.00th=[16512], 99.50th=[35072], 99.90th=[93696], 99.95th=[123392], | 99.99th=[195584] lat (usec) : 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01%, 100=0.64% lat (usec) : 250=94.91%, 500=2.76%, 750=0.11%, 1000=0.07% lat (msec) : 2=0.14%, 4=0.02%, 10=0.10%, 20=0.37%, 50=0.56% lat (msec) : 100=0.23%, 250=0.08%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=3.63%, sys=12.48%, ctx=9079177, majf=0, minf=14 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=10856295/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=178862MB, aggrb=203487KB/s, minb=48246KB/s, maxb=155241KB/s, mint=900066msec, maxt=900078msec Disk stats (read/write): dm-2: ios=45783724/28, merge=0/0, ticks=14086884/552, in_queue=14099888, util=100.00%, aggrios=15302554/317325, aggrmerge=0/0, aggrticks=5324778/75793, aggrin_queue=5403120, aggrutil=100.00% dm-1: ios=5/833005, merge=0/0, ticks=0/136424, in_queue=136468, util=10.74%, aggrios=45578059/918360, aggrmerge=965/23157, aggrticks=6367068/155120, aggrin_queue=6522812, aggrutil=100.00% xvdc: ios=45578059/918360, merge=965/23157, ticks=6367068/155120, in_queue=6522812, util=100.00% dm-0: ios=45579019/109294, merge=0/0, ticks=6367368/21496, in_queue=6396436, util=100.00% dm-3: ios=328639/9676, merge=0/0, ticks=9606968/69460, in_queue=9676456, util=100.00%, aggrios=328639/8639, aggrmerge=0/1037, aggrticks=9606840/46124, aggrin_queue=9652956, aggrutil=100.00% xvdd: ios=328639/8639, merge=0/1037, ticks=9606840/46124, in_queue=9652956, util=100.00% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc1+243bfd2c052d fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [12052KB/0KB/0KB /s] [3013/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1455: Sat Jul 29 13:50:32 2017 read : io=7490.4MB, bw=8521.4KB/s, iops=2130, runt=900105msec slat (usec): min=0, max=128, avg= 1.71, stdev= 4.39 clat (usec): min=64, max=1526.2K, avg=3752.99, stdev=24771.48 lat (usec): min=73, max=1526.2K, avg=3754.82, stdev=24772.16 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 106], 10.00th=[ 111], 20.00th=[ 115], | 30.00th=[ 117], 40.00th=[ 118], 50.00th=[ 119], 60.00th=[ 125], | 70.00th=[ 131], 80.00th=[ 133], 90.00th=[ 171], 95.00th=[12736], | 99.00th=[101888], 99.50th=[160768], 99.90th=[333824], 99.95th=[428032], | 99.99th=[692224] lat (usec) : 100=0.26%, 250=93.35%, 500=0.50%, 750=0.07%, 1000=0.06% lat (msec) : 2=0.11%, 4=0.01%, 10=0.29%, 20=1.36%, 50=1.87% lat (msec) : 100=1.09%, 250=0.82%, 500=0.18%, 750=0.02%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.19%, sys=0.72%, ctx=1983309, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=1917523/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1456: Sat Jul 29 13:50:32 2017 read : io=2596.8MB, bw=2953.6KB/s, iops=738, runt=900054msec slat (usec): min=0, max=154, avg= 3.92, stdev= 5.64 clat (usec): min=66, max=1527.1K, avg=10829.22, stdev=43048.40 lat (usec): min=72, max=1527.1K, avg=10833.40, stdev=43048.88 clat percentiles (usec): | 1.00th=[ 100], 5.00th=[ 105], 10.00th=[ 106], 20.00th=[ 111], | 30.00th=[ 116], 40.00th=[ 118], 50.00th=[ 120], 60.00th=[ 131], | 70.00th=[ 139], 80.00th=[ 173], 90.00th=[26240], 95.00th=[63232], | 99.00th=[201728], 99.50th=[280576], 99.90th=[501760], 99.95th=[626688], | 99.99th=[937984] lat (usec) : 100=0.47%, 250=83.17%, 500=0.37%, 750=0.09%, 1000=0.07% lat (msec) : 2=0.14%, 4=0.02%, 10=0.60%, 20=3.44%, 50=5.41% lat (msec) : 100=3.19%, 250=2.37%, 500=0.55%, 750=0.08%, 1000=0.02% lat (msec) : 2000=0.01% cpu : usr=0.14%, sys=0.48%, ctx=670686, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=664578/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=10086MB, aggrb=11474KB/s, minb=2953KB/s, maxb=8521KB/s, mint=900054msec, maxt=900105msec Disk stats (read/write): dm-2: ios=2581744/29, merge=0/0, ticks=14383904/4548, in_queue=14390016, util=100.00%, aggrios=896781/322000, aggrmerge=0/0, aggrticks=6180766/42330, aggrin_queue=6223142, aggrutil=100.00% dm-1: ios=5/857732, merge=0/0, ticks=0/101316, in_queue=101320, util=6.42%, aggrios=2363753/933247, aggrmerge=0/32409, aggrticks=294652/118376, aggrin_queue=412980, aggrutil=35.73% xvdc: ios=2363753/933247, merge=0/32409, ticks=294652/118376, in_queue=412980, util=35.73% dm-0: ios=2363748/108251, merge=0/0, ticks=294752/21128, in_queue=315924, util=31.04% dm-3: ios=326590/17, merge=0/0, ticks=18247548/4548, in_queue=18252184, util=100.00%, aggrios=326590/10, aggrmerge=0/7, aggrticks=18247532/2576, aggrin_queue=18250100, aggrutil=100.00% xvdd: ios=326590/10, merge=0/7, ticks=18247532/2576, in_queue=18250100, util=100.00% 4.12.0-rc1+f98e0eb68008 (2017-05-15 15:09:53 -0400)

Over 20k average IOPS on the run against a full cache means this was marked fixed.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc1+f98e0eb68008 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [389.1MB/0KB/0KB /s] [99.9K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1473: Sat Jul 29 15:01:22 2017 read : io=175480MB, bw=199657KB/s, iops=49914, runt=900001msec slat (usec): min=0, max=2865, avg= 7.12, stdev= 3.47 clat (usec): min=32, max=272750, avg=150.49, stdev=61.07 lat (usec): min=82, max=272765, avg=158.13, stdev=61.08 clat percentiles (usec): | 1.00th=[ 105], 5.00th=[ 111], 10.00th=[ 116], 20.00th=[ 123], | 30.00th=[ 129], 40.00th=[ 133], 50.00th=[ 139], 60.00th=[ 145], | 70.00th=[ 155], 80.00th=[ 169], 90.00th=[ 197], 95.00th=[ 235], | 99.00th=[ 326], 99.50th=[ 362], 99.90th=[ 454], 99.95th=[ 502], | 99.99th=[ 1020] lat (usec) : 50=0.01%, 100=0.09%, 250=95.99%, 500=3.87%, 750=0.03% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 500=0.01% cpu : usr=13.53%, sys=47.87%, ctx=18357918, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=44922870/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1474: Sat Jul 29 15:01:22 2017 read : io=166695MB, bw=189662KB/s, iops=47415, runt=900001msec slat (usec): min=0, max=7688, avg= 7.16, stdev= 3.64 clat (usec): min=14, max=600437, avg=158.90, stdev=650.77 lat (usec): min=69, max=600476, avg=166.58, stdev=650.92 clat percentiles (usec): | 1.00th=[ 106], 5.00th=[ 113], 10.00th=[ 118], 20.00th=[ 124], | 30.00th=[ 129], 40.00th=[ 135], 50.00th=[ 141], 60.00th=[ 147], | 70.00th=[ 155], 80.00th=[ 171], 90.00th=[ 197], 95.00th=[ 231], | 99.00th=[ 318], 99.50th=[ 350], 99.90th=[ 454], 99.95th=[ 564], | 99.99th=[24448] lat (usec) : 20=0.01%, 50=0.01%, 100=0.06%, 250=96.33%, 500=3.54% lat (usec) : 750=0.02%, 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01% cpu : usr=13.03%, sys=45.62%, ctx=17895059, majf=0, minf=14 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=42673942/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=342175MB, aggrb=389318KB/s, minb=189661KB/s, maxb=199656KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=87581336/28, merge=0/0, ticks=13288760/1272, in_queue=13321684, util=100.00%, aggrios=29220805/38094, aggrmerge=0/0, aggrticks=4442181/395562, aggrin_queue=4846680, aggrutil=100.00% dm-1: ios=13/48662, merge=0/0, ticks=4/8880, in_queue=8892, util=0.76%, aggrios=87640393/44939, aggrmerge=1/9151, aggrticks=12922660/9064, aggrin_queue=12942364, aggrutil=100.00% xvdc: ios=87640393/44939, merge=1/9151, ticks=12922660/9064, in_queue=12942364, util=100.00% dm-0: ios=87640381/6972, merge=0/0, ticks=12891608/1372, in_queue=12919780, util=100.00% dm-3: ios=22021/58648, merge=0/0, ticks=434932/1176436, in_queue=1611368, util=17.77%, aggrios=22021/58631, aggrmerge=0/17, aggrticks=434920/1170812, aggrin_queue=1605732, aggrutil=17.77% xvdd: ios=22021/58631, merge=0/17, ticks=434920/1170812, in_queue=1605732, util=17.77% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc1+f98e0eb68008 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [378.6MB/0KB/0KB /s] [96.8K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1827: Sat Jul 29 15:19:18 2017 read : io=81213MB, bw=92402KB/s, iops=23100, runt=900001msec slat (usec): min=0, max=653, avg= 6.99, stdev= 3.51 clat (usec): min=25, max=1352.2K, avg=336.83, stdev=5321.69 lat (usec): min=72, max=1352.2K, avg=344.32, stdev=5322.11 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 108], 10.00th=[ 112], 20.00th=[ 119], | 30.00th=[ 125], 40.00th=[ 131], 50.00th=[ 139], 60.00th=[ 149], | 70.00th=[ 159], 80.00th=[ 175], 90.00th=[ 201], 95.00th=[ 229], | 99.00th=[ 314], 99.50th=[ 390], 99.90th=[56576], 99.95th=[101888], | 99.99th=[242688] lat (usec) : 50=0.01%, 100=0.13%, 250=96.71%, 500=2.78%, 750=0.03% lat (usec) : 1000=0.01% lat (msec) : 2=0.02%, 4=0.01%, 10=0.01%, 20=0.07%, 50=0.11% lat (msec) : 100=0.06%, 250=0.04%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.96%, sys=21.63%, ctx=10178854, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=20790563/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1828: Sat Jul 29 15:19:18 2017 read : io=73647MB, bw=83794KB/s, iops=20948, runt=900001msec slat (usec): min=0, max=2496, avg= 7.18, stdev= 3.62 clat (usec): min=23, max=1510.2K, avg=372.08, stdev=6082.40 lat (usec): min=72, max=1510.2K, avg=379.78, stdev=6082.74 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 110], 10.00th=[ 114], 20.00th=[ 122], | 30.00th=[ 129], 40.00th=[ 135], 50.00th=[ 143], 60.00th=[ 151], | 70.00th=[ 163], 80.00th=[ 179], 90.00th=[ 207], 95.00th=[ 237], | 99.00th=[ 330], 99.50th=[ 414], 99.90th=[65280], 99.95th=[113152], | 99.99th=[272384] lat (usec) : 50=0.01%, 100=0.13%, 250=96.03%, 500=3.44%, 750=0.03% lat (usec) : 1000=0.01% lat (msec) : 2=0.02%, 4=0.01%, 10=0.02%, 20=0.08%, 50=0.12% lat (msec) : 100=0.07%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.82%, sys=20.07%, ctx=9121390, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=18853758/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=154861MB, aggrb=176196KB/s, minb=83794KB/s, maxb=92402KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=39629245/29, merge=0/0, ticks=13922184/2932, in_queue=13938180, util=100.00%, aggrios=13236641/192742, aggrmerge=0/0, aggrticks=5404900/39769, aggrin_queue=5447944, aggrutil=71.24% dm-1: ios=35/512628, merge=0/0, ticks=0/63156, in_queue=63168, util=4.26%, aggrios=39512858/560041, aggrmerge=1/16890, aggrticks=5866812/73188, aggrin_queue=5946060, aggrutil=73.11% xvdc: ios=39512858/560041, merge=1/16890, ticks=5866812/73188, in_queue=5946060, util=73.11% dm-0: ios=39512824/64453, merge=0/0, ticks=5858616/12180, in_queue=5880596, util=71.24% dm-3: ios=197065/1146, merge=0/0, ticks=10356084/43972, in_queue=10400068, util=65.55%, aggrios=197065/1143, aggrmerge=0/3, aggrticks=10356072/42172, aggrin_queue=10398236, aggrutil=65.55% xvdd: ios=197065/1143, merge=0/3, ticks=10356072/42172, in_queue=10398236, util=65.55% 4.12.0-rc1+701e03e4e180 (2017-05-14 21:54:33 -0400)

Over 20k average IOPS on the run against a full cache means this was marked fixed.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc1+701e03e4e180 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [383.6MB/0KB/0KB /s] [98.2K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1488: Sun Jul 30 00:38:12 2017 read : io=163416MB, bw=185931KB/s, iops=46482, runt=900001msec slat (usec): min=0, max=6338, avg= 6.90, stdev= 4.29 clat (usec): min=2, max=423004, avg=162.74, stdev=659.06 lat (usec): min=70, max=423007, avg=170.12, stdev=659.18 clat percentiles (usec): | 1.00th=[ 105], 5.00th=[ 112], 10.00th=[ 117], 20.00th=[ 124], | 30.00th=[ 131], 40.00th=[ 137], 50.00th=[ 143], 60.00th=[ 153], | 70.00th=[ 163], 80.00th=[ 179], 90.00th=[ 207], 95.00th=[ 241], | 99.00th=[ 326], 99.50th=[ 362], 99.90th=[ 490], 99.95th=[ 828], | 99.99th=[23680] lat (usec) : 4=0.01%, 20=0.01%, 50=0.01%, 100=0.10%, 250=95.64% lat (usec) : 500=4.17%, 750=0.04%, 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01% cpu : usr=11.29%, sys=43.37%, ctx=18655951, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=41834439/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1489: Sun Jul 30 00:38:12 2017 read : io=163729MB, bw=186287KB/s, iops=46571, runt=900001msec slat (usec): min=0, max=3379, avg= 6.91, stdev= 4.05 clat (usec): min=1, max=647713, avg=162.39, stdev=1109.49 lat (usec): min=71, max=647715, avg=169.78, stdev=1109.58 clat percentiles (usec): | 1.00th=[ 105], 5.00th=[ 112], 10.00th=[ 117], 20.00th=[ 123], | 30.00th=[ 129], 40.00th=[ 135], 50.00th=[ 141], 60.00th=[ 149], | 70.00th=[ 159], 80.00th=[ 173], 90.00th=[ 195], 95.00th=[ 219], | 99.00th=[ 286], 99.50th=[ 322], 99.90th=[ 462], 99.95th=[ 764], | 99.99th=[36096] lat (usec) : 2=0.01%, 20=0.01%, 50=0.01%, 100=0.10%, 250=97.69% lat (usec) : 500=2.13%, 750=0.03%, 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01% cpu : usr=11.62%, sys=43.26%, ctx=18444169, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=41914621/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=327145MB, aggrb=372217KB/s, minb=185930KB/s, maxb=186286KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=83732468/28, merge=0/0, ticks=13437692/4572, in_queue=13474260, util=100.00%, aggrios=27939916/52846, aggrmerge=0/0, aggrticks=4517696/389793, aggrin_queue=4918218, aggrutil=100.00% dm-1: ios=3/87822, merge=0/0, ticks=0/14672, in_queue=14672, util=1.21%, aggrios=83780322/91760, aggrmerge=1107/8163, aggrticks=12562052/16096, aggrin_queue=12583792, aggrutil=100.00% xvdc: ios=83780322/91760, merge=1107/8163, ticks=12562052/16096, in_queue=12583792, util=100.00% dm-0: ios=83781426/12436, merge=0/0, ticks=12554856/2392, in_queue=12589888, util=100.00% dm-3: ios=38320/58280, merge=0/0, ticks=998232/1152316, in_queue=2150096, util=21.10%, aggrios=38320/57106, aggrmerge=0/1182, aggrticks=998232/1128504, aggrin_queue=2126504, aggrutil=21.10% xvdd: ios=38320/57106, merge=0/1182, ticks=998232/1128504, in_queue=2126504, util=21.10% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc1+701e03e4e180 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [396.7MB/0KB/0KB /s] [102K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=4235: Sun Jul 30 00:57:53 2017 read : io=83984MB, bw=95555KB/s, iops=23888, runt=900001msec slat (usec): min=0, max=2757, avg= 6.65, stdev= 4.15 clat (usec): min=0, max=1331.1K, avg=325.84, stdev=5327.61 lat (usec): min=70, max=1331.1K, avg=333.00, stdev=5327.93 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 108], 10.00th=[ 113], 20.00th=[ 118], | 30.00th=[ 124], 40.00th=[ 129], 50.00th=[ 135], 60.00th=[ 141], | 70.00th=[ 149], 80.00th=[ 161], 90.00th=[ 183], 95.00th=[ 205], | 99.00th=[ 278], 99.50th=[ 362], 99.90th=[56064], 99.95th=[100864], | 99.99th=[242688] lat (usec) : 2=0.01%, 20=0.01%, 50=0.01%, 100=0.14%, 250=98.20% lat (usec) : 500=1.30%, 750=0.02%, 1000=0.01% lat (msec) : 2=0.02%, 4=0.01%, 10=0.01%, 20=0.07%, 50=0.10% lat (msec) : 100=0.06%, 250=0.04%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.85%, sys=21.52%, ctx=10254989, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=21499909/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=4236: Sun Jul 30 00:57:53 2017 read : io=75994MB, bw=86464KB/s, iops=21616, runt=900001msec slat (usec): min=0, max=3457, avg= 6.91, stdev= 4.10 clat (usec): min=2, max=1517.1K, avg=360.70, stdev=5924.51 lat (usec): min=72, max=1517.1K, avg=368.13, stdev=5924.81 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 110], 10.00th=[ 114], 20.00th=[ 121], | 30.00th=[ 126], 40.00th=[ 131], 50.00th=[ 137], 60.00th=[ 145], | 70.00th=[ 153], 80.00th=[ 169], 90.00th=[ 195], 95.00th=[ 227], | 99.00th=[ 326], 99.50th=[ 406], 99.90th=[64256], 99.95th=[114176], | 99.99th=[264192] lat (usec) : 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01%, 100=0.11% lat (usec) : 250=96.53%, 500=2.96%, 750=0.03%, 1000=0.01% lat (msec) : 2=0.02%, 4=0.01%, 10=0.02%, 20=0.07%, 50=0.11% lat (msec) : 100=0.07%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.46%, sys=20.26%, ctx=9368609, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=19454516/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=159978MB, aggrb=182019KB/s, minb=86464KB/s, maxb=95555KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=40941858/29, merge=0/0, ticks=13939452/2924, in_queue=13957884, util=100.00%, aggrios=13673355/191614, aggrmerge=0/0, aggrticks=5425713/41914, aggrin_queue=5472632, aggrutil=70.78% dm-1: ios=72/509243, merge=0/0, ticks=12/62896, in_queue=62908, util=4.30%, aggrios=40823706/556590, aggrmerge=1/16815, aggrticks=5811880/73008, aggrin_queue=5889632, aggrutil=72.65% xvdc: ios=40823706/556590, merge=1/16815, ticks=5811880/73008, in_queue=5889632, util=72.65% dm-0: ios=40823635/64315, merge=0/0, ticks=5806500/12056, in_queue=5833472, util=70.78% dm-3: ios=196358/1284, merge=0/0, ticks=10470628/50792, in_queue=10521516, util=65.71%, aggrios=196358/1281, aggrmerge=0/3, aggrticks=10470612/48468, aggrin_queue=10519068, aggrutil=65.71% xvdd: ios=196358/1281, merge=0/3, ticks=10470612/48468, in_queue=10519068, util=65.71% 4.12.0-rc1+78c45607b909 (2017-05-14 21:54:32 -0400)

The first job running against a full cache managed 17,398 average IOPS which isn’t terrible, but the second job only managed 694 average IOPS. Although this is a lot better than the worst performance seen, it is far below the best so this must be considered broken.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc1+78c45607b909 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [127.8MB/0KB/0KB /s] [32.8K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1633: Sun Jul 30 01:32:22 2017 read : io=20052MB, bw=22814KB/s, iops=5703, runt=900027msec slat (usec): min=2, max=79803, avg= 7.19, stdev=36.71 clat (usec): min=13, max=929368, avg=1392.96, stdev=9447.28 lat (usec): min=72, max=929376, avg=1400.61, stdev=9448.72 clat percentiles (usec): | 1.00th=[ 102], 5.00th=[ 105], 10.00th=[ 108], 20.00th=[ 114], | 30.00th=[ 118], 40.00th=[ 121], 50.00th=[ 124], 60.00th=[ 131], | 70.00th=[ 143], 80.00th=[ 175], 90.00th=[ 247], 95.00th=[ 378], | 99.00th=[39168], 99.50th=[61184], 99.90th=[124416], 99.95th=[158720], | 99.99th=[257024] lat (usec) : 20=0.01%, 50=0.01%, 100=0.37%, 250=89.87%, 500=5.63% lat (usec) : 750=0.28%, 1000=0.19% lat (msec) : 2=0.30%, 4=0.01%, 10=0.16%, 20=1.05%, 50=1.44% lat (msec) : 100=0.52%, 250=0.17%, 500=0.01%, 750=0.01%, 1000=0.01% cpu : usr=1.61%, sys=5.89%, ctx=4272767, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=5133359/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1634: Sun Jul 30 01:32:22 2017 read : io=70562MB, bw=80281KB/s, iops=20070, runt=900023msec slat (usec): min=2, max=63808, avg= 6.52, stdev=17.03 clat (usec): min=12, max=1526.4K, avg=389.81, stdev=4546.59 lat (usec): min=69, max=1526.4K, avg=396.77, stdev=4547.95 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 104], 10.00th=[ 106], 20.00th=[ 108], | 30.00th=[ 112], 40.00th=[ 118], 50.00th=[ 121], 60.00th=[ 125], | 70.00th=[ 135], 80.00th=[ 159], 90.00th=[ 227], 95.00th=[ 286], | 99.00th=[ 988], 99.50th=[16064], 99.90th=[59136], 99.95th=[84480], | 99.99th=[168960] lat (usec) : 20=0.01%, 50=0.01%, 100=0.34%, 250=92.06%, 500=6.20% lat (usec) : 750=0.23%, 1000=0.18% lat (msec) : 2=0.34%, 4=0.01%, 10=0.03%, 20=0.19%, 50=0.29% lat (msec) : 100=0.10%, 250=0.03%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=4.86%, sys=17.98%, ctx=12698034, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=18063745/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=90614MB, aggrb=103095KB/s, minb=22814KB/s, maxb=80281KB/s, mint=900023msec, maxt=900027msec Disk stats (read/write): dm-2: ios=23193142/28, merge=0/0, ticks=14243636/2712, in_queue=14251904, util=100.00%, aggrios=7849034/861075, aggrmerge=0/0, aggrticks=5211602/303989, aggrin_queue=5516568, aggrutil=99.10% dm-1: ios=1/2233198, merge=0/0, ticks=0/337548, in_queue=337612, util=25.48%, aggrios=22971489/2441741, aggrmerge=998/75445, aggrticks=3693756/390528, aggrin_queue=4083748, aggrutil=94.88% xvdc: ios=22971489/2441741, merge=998/75445, ticks=3693756/390528, in_queue=4083748, util=94.88% dm-0: ios=22972486/284707, merge=0/0, ticks=3711344/62548, in_queue=3776724, util=94.74% dm-3: ios=574615/65320, merge=0/0, ticks=11923464/511872, in_queue=12435368, util=99.10%, aggrios=574505/64312, aggrmerge=109/1008, aggrticks=11894448/464704, aggrin_queue=12359116, aggrutil=99.10% xvdd: ios=574505/64312, merge=109/1008, ticks=11894448/464704, in_queue=12359116, util=99.10% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc1+78c45607b909 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [71928KB/0KB/0KB /s] [17.1K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=4136: Sun Jul 30 01:48:29 2017 read : io=61171MB, bw=69594KB/s, iops=17398, runt=900076msec slat (usec): min=0, max=3387, avg= 6.22, stdev= 3.45 clat (usec): min=1, max=1174.6K, avg=451.49, stdev=7201.13 lat (usec): min=68, max=1174.6K, avg=458.13, stdev=7201.29 clat percentiles (usec): | 1.00th=[ 102], 5.00th=[ 104], 10.00th=[ 106], 20.00th=[ 108], | 30.00th=[ 110], 40.00th=[ 116], 50.00th=[ 120], 60.00th=[ 121], | 70.00th=[ 123], 80.00th=[ 129], 90.00th=[ 173], 95.00th=[ 239], | 99.00th=[ 548], 99.50th=[ 7008], 99.90th=[91648], 99.95th=[150528], | 99.99th=[313344] lat (usec) : 2=0.01%, 4=0.01%, 20=0.01%, 50=0.01%, 100=0.30% lat (usec) : 250=95.24%, 500=3.42%, 750=0.18%, 1000=0.15% lat (msec) : 2=0.19%, 4=0.01%, 10=0.04%, 20=0.12%, 50=0.17% lat (msec) : 100=0.09%, 250=0.07%, 500=0.02%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=3.99%, sys=15.44%, ctx=12579924, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=15659896/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=4137: Sun Jul 30 01:48:29 2017 read : io=2440.7MB, bw=2776.7KB/s, iops=694, runt=900089msec slat (usec): min=2, max=821, avg= 7.64, stdev= 2.96 clat (usec): min=1, max=1511.6K, avg=11514.55, stdev=48644.41 lat (usec): min=69, max=1511.7K, avg=11522.66, stdev=48644.81 clat percentiles (usec): | 1.00th=[ 99], 5.00th=[ 104], 10.00th=[ 106], 20.00th=[ 112], | 30.00th=[ 117], 40.00th=[ 119], 50.00th=[ 121], 60.00th=[ 123], | 70.00th=[ 126], 80.00th=[ 159], 90.00th=[22656], 95.00th=[70144], | 99.00th=[240640], 99.50th=[329728], 99.90th=[569344], 99.95th=[675840], | 99.99th=[921600] lat (usec) : 2=0.01%, 20=0.01%, 100=1.12%, 250=83.31%, 500=2.64% lat (usec) : 750=0.21%, 1000=0.15% lat (msec) : 2=0.26%, 4=0.04%, 10=0.33%, 20=1.50%, 50=4.00% lat (msec) : 100=2.81%, 250=2.69%, 500=0.77%, 750=0.13%, 1000=0.02% lat (msec) : 2000=0.01% cpu : usr=0.25%, sys=0.76%, ctx=624059, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=624801/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=63612MB, aggrb=72369KB/s, minb=2776KB/s, maxb=69593KB/s, mint=900076msec, maxt=900089msec Disk stats (read/write): dm-2: ios=16284697/27, merge=0/0, ticks=14316884/11932, in_queue=14332000, util=100.00%, aggrios=5485619/348652, aggrmerge=0/0, aggrticks=6603850/63144, aggrin_queue=6667324, aggrutil=100.00% dm-1: ios=2/873767, merge=0/0, ticks=0/151744, in_queue=151760, util=8.83%, aggrios=16104758/960703, aggrmerge=0/80804, aggrticks=2186320/160864, aggrin_queue=2346844, aggrutil=91.57% xvdc: ios=16104758/960703, merge=0/80804, ticks=2186320/160864, in_queue=2346844, util=91.57% dm-0: ios=16104756/172087, merge=0/0, ticks=2187352/31620, in_queue=2219940, util=91.28% dm-3: ios=352100/102, merge=0/0, ticks=17624200/6068, in_queue=17630272, util=100.00%, aggrios=352101/91, aggrmerge=0/11, aggrticks=17624100/3584, aggrin_queue=17627592, aggrutil=100.00% xvdd: ios=352101/91, merge=0/11, ticks=17624100/3584, in_queue=17627592, util=100.00% 4.12.0-rc1+6cf4cc8f8b3b (2017-05-14 21:54:33 -0400)

Over 20k average IOPS on the run against a full cache means this was marked fixed.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc1+6cf4cc8f8b3b fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [399.9MB/0KB/0KB /s] [102K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1314: Sun Jul 30 02:20:12 2017 read : io=132841MB, bw=151143KB/s, iops=37785, runt=900001msec slat (usec): min=2, max=4635, avg= 6.74, stdev= 3.66 clat (usec): min=1, max=1208.1K, avg=202.59, stdev=2398.09 lat (usec): min=68, max=1209.8K, avg=209.79, stdev=2398.38 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 108], 10.00th=[ 111], 20.00th=[ 119], | 30.00th=[ 123], 40.00th=[ 129], 50.00th=[ 135], 60.00th=[ 141], | 70.00th=[ 151], 80.00th=[ 163], 90.00th=[ 189], 95.00th=[ 217], | 99.00th=[ 322], 99.50th=[ 402], 99.90th=[18560], 99.95th=[31360], | 99.99th=[88576] lat (usec) : 2=0.01%, 4=0.01%, 20=0.01%, 50=0.01%, 100=0.17% lat (usec) : 250=97.09%, 500=2.42%, 750=0.06%, 1000=0.03% lat (msec) : 2=0.05%, 4=0.01%, 10=0.02%, 20=0.05%, 50=0.07% lat (msec) : 100=0.02%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=9.52%, sys=36.77%, ctx=17387398, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=34007324/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1315: Sun Jul 30 02:20:12 2017 read : io=134807MB, bw=153381KB/s, iops=38345, runt=900001msec slat (usec): min=2, max=4192, avg= 6.64, stdev= 3.55 clat (usec): min=1, max=1525.4K, avg=199.62, stdev=3082.99 lat (usec): min=66, max=1525.4K, avg=206.71, stdev=3083.18 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 107], 10.00th=[ 111], 20.00th=[ 118], | 30.00th=[ 122], 40.00th=[ 127], 50.00th=[ 133], 60.00th=[ 139], | 70.00th=[ 149], 80.00th=[ 161], 90.00th=[ 187], 95.00th=[ 221], | 99.00th=[ 326], 99.50th=[ 398], 99.90th=[13888], 99.95th=[26496], | 99.99th=[99840] lat (usec) : 2=0.01%, 10=0.01%, 50=0.01%, 100=0.19%, 250=96.79% lat (usec) : 500=2.73%, 750=0.05%, 1000=0.03% lat (msec) : 2=0.06%, 4=0.01%, 10=0.02%, 20=0.05%, 50=0.05% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=9.82%, sys=36.76%, ctx=17659747, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=34510679/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=267648MB, aggrb=304524KB/s, minb=151143KB/s, maxb=153380KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=68501557/28, merge=0/0, ticks=13693912/524, in_queue=13720060, util=100.00%, aggrios=22876855/176918, aggrmerge=0/0, aggrticks=4875376/104950, aggrin_queue=4986021, aggrutil=98.38% dm-1: ios=19/418181, merge=0/0, ticks=4/70080, in_queue=70112, util=5.49%, aggrios=68456360/446285, aggrmerge=2397/24774, aggrticks=9939560/79736, aggrin_queue=10019744, aggrutil=98.40% xvdc: ios=68456360/446285, merge=2397/24774, ticks=9939560/79736, in_queue=10019744, util=98.40% dm-0: ios=68458738/53971, merge=0/0, ticks=9966748/12776, in_queue=9996560, util=98.38% dm-3: ios=171810/58604, merge=0/0, ticks=4659376/231996, in_queue=4891392, util=57.31%, aggrios=171801/56085, aggrmerge=9/2519, aggrticks=4655812/142768, aggrin_queue=4798556, aggrutil=57.31% xvdd: ios=171801/56085, merge=9/2519, ticks=4655812/142768, in_queue=4798556, util=57.31% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc1+6cf4cc8f8b3b fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [397.6MB/0KB/0KB /s] [102K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=3900: Sun Jul 30 02:38:37 2017 read : io=73397MB, bw=83509KB/s, iops=20877, runt=900001msec slat (usec): min=3, max=3087, avg= 6.85, stdev= 3.37 clat (usec): min=9, max=1501.2K, avg=373.90, stdev=6030.69 lat (usec): min=72, max=1501.2K, avg=381.22, stdev=6031.19 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 109], 10.00th=[ 114], 20.00th=[ 121], | 30.00th=[ 125], 40.00th=[ 131], 50.00th=[ 137], 60.00th=[ 143], | 70.00th=[ 153], 80.00th=[ 169], 90.00th=[ 197], 95.00th=[ 235], | 99.00th=[ 338], 99.50th=[ 418], 99.90th=[69120], 99.95th=[118272], | 99.99th=[268288] lat (usec) : 10=0.01%, 50=0.01%, 100=0.10%, 250=96.05%, 500=3.43% lat (usec) : 750=0.03%, 1000=0.02% lat (msec) : 2=0.02%, 4=0.01%, 10=0.02%, 20=0.08%, 50=0.12% lat (msec) : 100=0.07%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.68%, sys=20.45%, ctx=9729438, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=18789539/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=3901: Sun Jul 30 02:38:37 2017 read : io=76248MB, bw=86753KB/s, iops=21688, runt=900001msec slat (usec): min=3, max=3309, avg= 6.82, stdev= 3.51 clat (usec): min=1, max=1436.1K, avg=359.59, stdev=5873.40 lat (usec): min=70, max=1436.1K, avg=366.87, stdev=5873.83 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 107], 10.00th=[ 110], 20.00th=[ 117], | 30.00th=[ 122], 40.00th=[ 127], 50.00th=[ 133], 60.00th=[ 139], | 70.00th=[ 147], 80.00th=[ 159], 90.00th=[ 181], 95.00th=[ 205], | 99.00th=[ 282], 99.50th=[ 370], 99.90th=[67072], 99.95th=[116224], | 99.99th=[268288] lat (usec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (usec) : 100=0.18%, 250=98.06%, 500=1.36%, 750=0.02%, 1000=0.02% lat (msec) : 2=0.02%, 4=0.01%, 10=0.02%, 20=0.07%, 50=0.12% lat (msec) : 100=0.07%, 250=0.05%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=5.57%, sys=21.29%, ctx=9691035, majf=0, minf=14 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=19519474/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=149645MB, aggrb=170262KB/s, minb=83508KB/s, maxb=86753KB/s, mint=900001msec, maxt=900001msec Disk stats (read/write): dm-2: ios=38297973/29, merge=0/0, ticks=13982664/7004, in_queue=14003132, util=100.00%, aggrios=12791550/194298, aggrmerge=0/0, aggrticks=5442113/27136, aggrin_queue=5472816, aggrutil=68.60% dm-1: ios=102/517327, merge=0/0, ticks=12/62076, in_queue=62116, util=4.27%, aggrios=38174313/566608, aggrmerge=0/16117, aggrticks=5431508/72500, aggrin_queue=5506176, aggrutil=70.51% xvdc: ios=38174313/566608, merge=0/16117, ticks=5431508/72500, in_queue=5506176, util=70.51% dm-0: ios=38174211/65537, merge=0/0, ticks=5427632/12328, in_queue=5450628, util=68.60% dm-3: ios=200338/30, merge=0/0, ticks=10898696/7004, in_queue=10905704, util=68.37%, aggrios=200338/12, aggrmerge=0/18, aggrticks=10898668/2728, aggrin_queue=10901396, aggrutil=68.37% xvdd: ios=200338/12, merge=0/18, ticks=10898668/2728, in_queue=10901396, util=68.37% 4.12.0-rc1+4d44ec5ab751 (2017-05-14 21:54:33 -0400)

Although this time against a full cache both jobs manage a fairly consistent 6 or 7 thousand IOPS, this is nowhere near the level of performance we had seen before so this must be considered broken.

At this point git bisect was able to report that 6cf4cc8f8b3b was the first fixed commit.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc1+4d44ec5ab751 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [48908KB/0KB/0KB /s] [12.3K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1745: Sun Jul 30 03:23:34 2017 read : io=24635MB, bw=28027KB/s, iops=7006, runt=900037msec slat (usec): min=1, max=14906, avg= 6.77, stdev= 8.06 clat (usec): min=1, max=1249.6K, avg=1132.59, stdev=9627.84 lat (usec): min=69, max=1249.6K, avg=1139.79, stdev=9628.80 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 104], 10.00th=[ 106], 20.00th=[ 108], | 30.00th=[ 112], 40.00th=[ 118], 50.00th=[ 121], 60.00th=[ 123], | 70.00th=[ 127], 80.00th=[ 147], 90.00th=[ 213], 95.00th=[ 282], | 99.00th=[32128], 99.50th=[57088], 99.90th=[136192], 99.95th=[177152], | 99.99th=[292864] lat (usec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (usec) : 100=0.38%, 250=93.13%, 500=3.55%, 750=0.22%, 1000=0.21% lat (msec) : 2=0.29%, 4=0.01%, 10=0.07%, 20=0.57%, 50=0.98% lat (msec) : 100=0.41%, 250=0.18%, 500=0.02%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=2.13%, sys=6.56%, ctx=5311621, majf=1, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=6306443/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1746: Sun Jul 30 03:23:34 2017 read : io=35476MB, bw=40361KB/s, iops=10090, runt=900046msec slat (usec): min=1, max=23444, avg= 6.51, stdev=11.66 clat (usec): min=1, max=1501.4K, avg=784.01, stdev=8022.09 lat (usec): min=67, max=1501.4K, avg=790.95, stdev=8022.73 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 104], 10.00th=[ 106], 20.00th=[ 108], | 30.00th=[ 111], 40.00th=[ 117], 50.00th=[ 121], 60.00th=[ 123], | 70.00th=[ 127], 80.00th=[ 145], 90.00th=[ 211], 95.00th=[ 278], | 99.00th=[19840], 99.50th=[40192], 99.90th=[113152], 99.95th=[152576], | 99.99th=[280576] lat (usec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (usec) : 100=0.37%, 250=93.16%, 500=4.24%, 750=0.27%, 1000=0.20% lat (msec) : 2=0.29%, 4=0.02%, 10=0.07%, 20=0.37%, 50=0.62% lat (msec) : 100=0.25%, 250=0.11%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=2.58%, sys=9.44%, ctx=6971695, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=9081785/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=60110MB, aggrb=68388KB/s, minb=28027KB/s, maxb=40361KB/s, mint=900037msec, maxt=900046msec Disk stats (read/write): dm-2: ios=15387198/28, merge=0/0, ticks=14306100/6972, in_queue=14316452, util=100.00%, aggrios=5233578/737062, aggrmerge=0/0, aggrticks=5314757/297484, aggrin_queue=5612629, aggrutil=100.00% dm-1: ios=0/1898650, merge=0/0, ticks=0/275700, in_queue=275740, util=20.28%, aggrios=15166622/2060890, aggrmerge=2630/85657, aggrticks=2377016/315952, aggrin_queue=2692844, aggrutil=88.91% xvdc: ios=15166622/2060890, merge=2630/85657, ticks=2377016/315952, in_queue=2692844, util=88.91% dm-0: ios=15169252/249770, merge=0/0, ticks=2424412/51644, in_queue=2477168, util=88.37% dm-3: ios=531482/62766, merge=0/0, ticks=13519860/565108, in_queue=14084980, util=100.00%, aggrios=531464/60098, aggrmerge=18/2668, aggrticks=13515428/447352, aggrin_queue=13962752, aggrutil=100.00% xvdd: ios=531464/60098, merge=18/2668, ticks=13515428/447352, in_queue=13962752, util=100.00% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc1+4d44ec5ab751 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [42352KB/0KB/0KB /s] [10.6K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=4248: Sun Jul 30 03:40:34 2017 read : io=27816MB, bw=31644KB/s, iops=7911, runt=900111msec slat (usec): min=1, max=3143, avg= 6.38, stdev= 3.68 clat (usec): min=1, max=1247.3K, avg=1002.62, stdev=10214.33 lat (usec): min=69, max=1247.4K, avg=1009.42, stdev=10214.63 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 103], 10.00th=[ 105], 20.00th=[ 107], | 30.00th=[ 109], 40.00th=[ 114], 50.00th=[ 119], 60.00th=[ 121], | 70.00th=[ 123], 80.00th=[ 135], 90.00th=[ 207], 95.00th=[ 278], | 99.00th=[25472], 99.50th=[53504], 99.90th=[150528], 99.95th=[195584], | 99.99th=[337920] lat (usec) : 2=0.01%, 4=0.01%, 20=0.01%, 50=0.01%, 100=0.49% lat (usec) : 250=93.10%, 500=4.06%, 750=0.23%, 1000=0.21% lat (msec) : 2=0.29%, 4=0.01%, 10=0.05%, 20=0.36%, 50=0.66% lat (msec) : 100=0.32%, 250=0.20%, 500=0.02%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=2.18%, sys=7.27%, ctx=6208585, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=7120878/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=4249: Sun Jul 30 03:40:34 2017 read : io=21517MB, bw=24481KB/s, iops=6120, runt=900042msec slat (usec): min=1, max=3946, avg= 6.52, stdev= 4.06 clat (usec): min=1, max=1225.4K, avg=1298.37, stdev=13872.59 lat (usec): min=68, max=1225.4K, avg=1305.32, stdev=13872.88 clat percentiles (usec): | 1.00th=[ 100], 5.00th=[ 103], 10.00th=[ 105], 20.00th=[ 107], | 30.00th=[ 108], 40.00th=[ 111], 50.00th=[ 118], 60.00th=[ 121], | 70.00th=[ 122], 80.00th=[ 127], 90.00th=[ 195], 95.00th=[ 270], | 99.00th=[29568], 99.50th=[82432], 99.90th=[207872], 99.95th=[268288], | 99.99th=[452608] lat (usec) : 2=0.01%, 4=0.01%, 20=0.01%, 50=0.01%, 100=0.60% lat (usec) : 250=93.47%, 500=3.81%, 750=0.20%, 1000=0.17% lat (msec) : 2=0.28%, 4=0.01%, 10=0.05%, 20=0.23%, 50=0.44% lat (msec) : 100=0.34%, 250=0.34%, 500=0.05%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=1.71%, sys=5.81%, ctx=5136977, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=5508414/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=49333MB, aggrb=56123KB/s, minb=24480KB/s, maxb=31644KB/s, mint=900042msec, maxt=900111msec Disk stats (read/write): dm-2: ios=12628996/29, merge=0/0, ticks=14325248/9004, in_queue=14337928, util=100.00%, aggrios=4283504/430396, aggrmerge=0/0, aggrticks=6408228/152828, aggrin_queue=6561282, aggrutil=100.00% dm-1: ios=1/1069937, merge=0/0, ticks=0/186088, in_queue=186096, util=11.15%, aggrios=12454547/1169074, aggrmerge=94/92406, aggrticks=1764348/199032, aggrin_queue=1963216, aggrutil=82.20% xvdc: ios=12454547/1169074, merge=94/92406, ticks=1764348/199032, in_queue=1963216, util=82.20% dm-0: ios=12454640/195514, merge=0/0, ticks=1766212/37200, in_queue=1804008, util=82.11% dm-3: ios=395873/25738, merge=0/0, ticks=17458472/235196, in_queue=17693744, util=100.00%, aggrios=395871/25673, aggrmerge=2/65, aggrticks=17458124/214396, aggrin_queue=17672604, aggrutil=100.00% xvdd: ios=395871/25673, merge=2/65, ticks=17458124/214396, in_queue=17672604, util=100.00% Tests against additional kernel builds 4.12.0-rc1 2ea659a9ef48 (2017-05-13 13:19:49 -0700)

A couple of the release candidates with some of Joe’s fixes in had started to produce approaching decent performance so I wanted to go back to the first tagged kernel for this release candidate to get an overview of where Joe had been starting from.

2,069 and 689 average IOPS against a full cache demonstrate the poor performance.

Empty cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.0-rc1+2ea659a9ef48 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [226.2MB/0KB/0KB /s] [57.9K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1410: Sun Jul 30 05:18:46 2017 read : io=119229MB, bw=135648KB/s, iops=33911, runt=900055msec slat (usec): min=3, max=4243, avg= 6.13, stdev= 2.67 clat (usec): min=12, max=777623, avg=227.66, stdev=3034.59 lat (usec): min=69, max=777634, avg=234.21, stdev=3035.00 clat percentiles (usec): | 1.00th=[ 102], 5.00th=[ 105], 10.00th=[ 107], 20.00th=[ 111], | 30.00th=[ 117], 40.00th=[ 121], 50.00th=[ 123], 60.00th=[ 126], | 70.00th=[ 131], 80.00th=[ 143], 90.00th=[ 183], 95.00th=[ 237], | 99.00th=[ 370], 99.50th=[ 466], 99.90th=[29568], 99.95th=[52480], | 99.99th=[140288] lat (usec) : 20=0.01%, 50=0.01%, 100=0.22%, 250=95.59%, 500=3.73% lat (usec) : 750=0.10%, 1000=0.07% lat (msec) : 2=0.10%, 4=0.01%, 10=0.01%, 20=0.03%, 50=0.09% lat (msec) : 100=0.03%, 250=0.02%, 500=0.01%, 750=0.01%, 1000=0.01% cpu : usr=8.43%, sys=31.48%, ctx=19010691, majf=1, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=30522650/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1411: Sun Jul 30 05:18:46 2017 read : io=38320MB, bw=43598KB/s, iops=10899, runt=900034msec slat (usec): min=3, max=2937, avg= 7.12, stdev= 3.52 clat (usec): min=1, max=664000, avg=724.45, stdev=7322.47 lat (usec): min=69, max=664008, avg=732.02, stdev=7323.28 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 104], 10.00th=[ 107], 20.00th=[ 111], | 30.00th=[ 117], 40.00th=[ 120], 50.00th=[ 122], 60.00th=[ 125], | 70.00th=[ 129], 80.00th=[ 137], 90.00th=[ 165], 95.00th=[ 235], | 99.00th=[16512], 99.50th=[38656], 99.90th=[111104], 99.95th=[148480], | 99.99th=[246784] lat (usec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (usec) : 100=0.59%, 250=95.04%, 500=2.69%, 750=0.11%, 1000=0.08% lat (msec) : 2=0.14%, 4=0.02%, 10=0.10%, 20=0.34%, 50=0.52% lat (msec) : 100=0.25%, 250=0.11%, 500=0.01%, 750=0.01% cpu : usr=3.25%, sys=11.57%, ctx=8505263, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=9809817/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=157549MB, aggrb=179244KB/s, minb=43597KB/s, maxb=135647KB/s, mint=900034msec, maxt=900055msec Disk stats (read/write): dm-2: ios=40332438/28, merge=0/0, ticks=14117864/2196, in_queue=14130156, util=100.00%, aggrios=13483192/313952, aggrmerge=0/0, aggrticks=5420121/72913, aggrin_queue=5494664, aggrutil=100.00% dm-1: ios=2/824719, merge=0/0, ticks=0/128108, in_queue=128144, util=10.06%, aggrios=40125306/891909, aggrmerge=1810/39159, aggrticks=5531356/144120, aggrin_queue=5674304, aggrutil=99.96% xvdc: ios=40125306/891909, merge=1810/39159, ticks=5531356/144120, in_queue=5674304, util=99.96% dm-0: ios=40127114/107088, merge=0/0, ticks=5533456/20608, in_queue=5559116, util=99.96% dm-3: ios=322461/10050, merge=0/0, ticks=10726908/70024, in_queue=10796732, util=100.00%, aggrios=322461/8135, aggrmerge=0/1923, aggrticks=10726860/46396, aggrin_queue=10773180, aggrutil=100.00% xvdd: ios=322461/8135, merge=0/1923, ticks=10726860/46396, in_queue=10773180, util=100.00% Full cache $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.0-rc1+2ea659a9ef48 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [11616KB/0KB/0KB /s] [2904/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=3880: Sun Jul 30 05:36:12 2017 read : io=7278.8MB, bw=8279.1KB/s, iops=2069, runt=900169msec slat (usec): min=3, max=749, avg= 7.94, stdev= 6.36 clat (usec): min=2, max=1369.3K, avg=3854.37, stdev=24787.42 lat (usec): min=72, max=1369.3K, avg=3862.76, stdev=24789.32 clat percentiles (usec): | 1.00th=[ 102], 5.00th=[ 104], 10.00th=[ 107], 20.00th=[ 108], | 30.00th=[ 116], 40.00th=[ 118], 50.00th=[ 121], 60.00th=[ 121], | 70.00th=[ 122], 80.00th=[ 125], 90.00th=[ 163], 95.00th=[13632], | 99.00th=[104960], 99.50th=[162816], 99.90th=[337920], 99.95th=[423936], | 99.99th=[659456] lat (usec) : 4=0.01%, 50=0.01%, 100=0.15%, 250=93.36%, 500=0.46% lat (usec) : 750=0.09%, 1000=0.07% lat (msec) : 2=0.11%, 4=0.01%, 10=0.29%, 20=1.36%, 50=1.89% lat (msec) : 100=1.15%, 250=0.85%, 500=0.18%, 750=0.02%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.71%, sys=2.61%, ctx=1927655, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=1863344/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=3881: Sun Jul 30 05:36:12 2017 read : io=2423.8MB, bw=2757.4KB/s, iops=689, runt=900123msec slat (usec): min=3, max=702, avg= 8.61, stdev= 5.54 clat (usec): min=64, max=1534.9K, avg=11593.97, stdev=45491.15 lat (usec): min=70, max=1534.9K, avg=11603.08, stdev=45492.10 clat percentiles (usec): | 1.00th=[ 100], 5.00th=[ 103], 10.00th=[ 104], 20.00th=[ 106], | 30.00th=[ 108], 40.00th=[ 113], 50.00th=[ 119], 60.00th=[ 122], | 70.00th=[ 131], 80.00th=[ 163], 90.00th=[28288], 95.00th=[68096], | 99.00th=[216064], 99.50th=[296960], 99.90th=[544768], 99.95th=[659456], | 99.99th=[929792] lat (usec) : 100=0.68%, 250=82.38%, 500=0.40%, 750=0.11%, 1000=0.08% lat (msec) : 2=0.15%, 4=0.02%, 10=0.62%, 20=3.41%, 50=5.54% lat (msec) : 100=3.34%, 250=2.53%, 500=0.61%, 750=0.10%, 1000=0.02% lat (msec) : 2000=0.01% cpu : usr=0.36%, sys=0.86%, ctx=628886, majf=0, minf=14 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=620477/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=9702.5MB, aggrb=11037KB/s, minb=2757KB/s, maxb=8279KB/s, mint=900123msec, maxt=900169msec Disk stats (read/write): dm-2: ios=2483744/29, merge=0/0, ticks=14379036/2900, in_queue=14383400, util=100.00%, aggrios=863141/314477, aggrmerge=0/0, aggrticks=6157485/39758, aggrin_queue=6197288, aggrutil=100.00% dm-1: ios=33/837830, merge=0/0, ticks=0/96648, in_queue=96660, util=6.07%, aggrios=2269963/907940, aggrmerge=0/35094, aggrticks=285192/112048, aggrin_queue=397148, aggrutil=34.50% xvdc: ios=2269963/907940, merge=0/35094, ticks=285192/112048, in_queue=397148, util=34.50% dm-0: ios=2269930/105586, merge=0/0, ticks=285436/19728, in_queue=305184, util=30.03% dm-3: ios=319462/16, merge=0/0, ticks=18187020/2900, in_queue=18190020, util=100.00%, aggrios=319462/8, aggrmerge=0/8, aggrticks=18187012/1068, aggrin_queue=18188076, aggrutil=100.00% xvdd: ios=319462/8, merge=0/8, ticks=18187012/1068, in_queue=18188076, util=100.00%
Categories: LUG Community Blogs

Steve Kemp: So I'm considering a new project

Fri, 28/07/2017 - 22:00

In the past there used to be a puppet-labs project called puppet-dashboard, which would let you see the state of your managed-nodes. Having even a very basic and simple "report user-interface" is pretty neat when you're pushing out a change, and you want to see it be applied across your fleet of hosts.

There are some other neat features, such as allowing you to identify failures easily, and see nodes that haven't reported-in recently.

This was spun out into a community-supported project which is largely stale:

Having a dashboard is nice, but the current state of the software is less good. It turns out that the implementation is pretty simple though:

  • Puppet runs on a node.
  • The node reports back to the puppet-master what happened.
  • The puppet-master can optionally HTTP-post that report to the reporting node.

The reporting node can thus receive real-time updates, and do what it wants with them. You can even sidestep the extra server if you wish:

  • The puppet-master can archive the reports locally.

For example on my puppet-master I have this:

root@master /var/lib/puppet/reports # ls | tail -n4 smaug.dh.bytemark.co.uk ssh.steve.org.uk www.dns-api.com www.steve.org.uk

Inside each directory is a bunch of YAML files which describe the state of the host, and the recipes that were applied. Parsing those is pretty simple, the hardest part would be making a useful/attractive GUI. But happily we have the existing one to "inspire" us.

I think I just need to write down a list of assumptions and see if they make sense. After all the existing installation(s) won't break, it's just a matter of deciding whether it is useful/worthwhile way to spend some time.

  • Assume you have 100+ hosts running puppet 4.x
  • Assume you want a broad overview:
    • All the nodes you're managing.
    • Whether their last run triggered a change, resulted in an error, or logged anything.
    • If so what changed/failed/was output?
  • For each individual run you want to see:
    • Rough overview.
  • Assume you don't want to keep history indefinitely, just the last 50 runs or so of each host.

Beyond that you might want to export data about the managed-nodes themselves. For example you might want a list of all the hosts which have "bash" installed on them. Or "All nodes with local user "steve"." I've written that stuff already, as it is very useful for auditing & etc.

The hard part about that is that to get the extra data you'll need to include a puppet module to collect it. I suspect a new dashboard would be broadly interesting/useful but unless you have that extra detail it might not be so useful. You can't point to a slightly more modern installation and say "Yes this is worth migrating to". But if you have extra meta-data you can say:

  • Give me a list of all hosts running wheezy.
  • Give me a list of all hosts running exim4 version 4.84.2-2+deb8u4.

And that facility is very useful when you have shellshock, or similar knocking at your door.

Anyway as a hacky start I wrote some code to parse reports, avoiding the magic object-fu that the YAML would usually invoke. The end result is this:

root@master ~# dump-run www.steve.org.uk www.steve.org.uk Puppet Version: 4.8.2 /var/lib/puppet/reports/www.steve.org.uk/201707291813.yaml Runtime: 2.16 Status:changed Time:2017-07-29 18:13:04 +0000 Resources total -> 176 skipped -> 2 failed -> 0 changed -> 3 out_of_sync -> 3 scheduled -> 0 corrective_change -> 3 Changed Resources Ssh_authorized_key[skx@shelob-s-fi] /etc/puppet/code/environments/production/modules/ssh_keys/manifests/init.pp:17 Ssh_authorized_key[skx@deagol-s-fi] /etc/puppet/code/environments/production/modules/ssh_keys/manifests/init.pp:22 Ssh_authorized_key[steve@ssh.steve.org.uk-s-fi] /etc/puppet/code/environments/production/modules/ssh_keys/manifests/init.pp:27 Skipped Resources Exec[clone sysadmin utils] Exec[update sysadmin utils]
Categories: LUG Community Blogs

Andy Smith: lvmcache on Debian experimental kernel 4.12.2-1~exp1

Thu, 27/07/2017 - 15:08

In the previous article I mentioned that it may be worth checking if the kernel version 4.12.2-1~exp1 as (currently) supplied by the linux-image-4.12.0-trunk-amd64 package from Debian experimental was new enough to have the dm-cache performance improvements.

The answer appears to be yes.

Categories: LUG Community Blogs

Andy Smith: lvmcache with a 4.12.3 kernel

Thu, 27/07/2017 - 10:44
Background

In the previous two articles I had discovered that lvmcache had amazing performance on an empty cache but then on every run after that (i.e. when the cache device was full of junk) went scarcely better than baseline HDD.

A few days ago I happened across an email on the linux-lvm list where Mike Snitzer advised:

the [CentOS] 7.4 dm-cache will be much more performant than the 7.3 cache you appear to be using.

…and…

It could be that your workload isn’t accessing the data enough to warrant promotion to the cache. dm-cache is a “hotspot” cache. If you aren’t accessing the data repeatedly then you won’t see much benefit (particularly with the 7.3 and earlier releases).

Just to get a feel, you could try the latest upstream 4.12 kernel to see how effective the 7.4 dm-cache will be for your setup.

I don’t know what kernel version CentOS 7.3 uses, but the VM I’m testing with is Debian testing (buster), so some version of 4.11.x plus backported patches.

That seemed pretty new, but Mike is suggesting 4.12.x so I thought I’d re-test lvmcache with the latest stable upstream kernel, which at the time of writing is version 4.12.3.

Test methodology

This time around I only focused on fio tests, using the same settings as before:

[partial] ioengine=libaio direct=1 iodepth=8 numjobs=2 readwrite=randread random_distribution=zipf:1.2 bs=4k size=2g unlink=1 runtime=20m time_based=1 per_job_logs=1 log_avg_msec=500 write_iops_log=/var/tmp/fio-${FIOTEST}

The only changes were:

  1. to reduce the run time to 20 minutes from 30 minutes, since all the interesting things happened within the first 20 minutes before.
  2. to write an IOPS log entry every 500ms instead of ever 1000ms, as the log files were quite small and some higher resolution might help smooth graphs out.

Last time there was a dramatic difference between the initial run with an empty cache and subsequent runs with a cache volume full of junk, so I did a test for each of those conditions, as well as tests for the baseline SSD and HDD.

The virtual machine had been upgraded from Debian 9 (stretch) to testing (buster), so it still had packaged kernel versions 4.9.30-2 and 4.11.6-1 laying around to test things with. In addition I compiled up version 4.12.3 by copying the .config from 4.11.6-1 then doing make oldconfig accepting all defaults.

Results

Although the fio job spec was essentially the same as in the previous article, I have since worked out how to merge the IOPS logs from both jobs so the graphs will seem to show about double the IOPS as they did before.

All-in-one

Well that’s an interesting set of graphs but rather hard to distinguish. Let’s try that by kernel version.

Baseline SSD by kernel version

A couple of weird things here:

  1. 4.12.3 and 4.11.6-1 are actually fairly consistent, but 4.9.30-2 varies rather a lot.
  2. All kernels show a sharp dip a few minutes in. I don’t know what that is about.

Although these lines do look quite far apart, bear in mind that this graph’s y axis starts at 92k IOPS. The average IOPS didn’t vary that much:

Average IOPS by kernel version 4.9.30-2 4.11.6-1 4.12.3 102,325 102,742 104,352

So there was actually only a 1.9% difference between the worst performer and the best.

Baseline HDD by kernel version

4.9.30-2 and 4.12.3 are close enough here to probably be within the margin of error, but there is something weird going on with 4.11.6-1.

Its average IOPS across the 20 minute test were only 56% of those for 4.12.3 and 53% of those for 4.9.30-2, which is quite a big disparity. I re-ran these tests 5 times to check it wasn’t some anomaly, but no, it’s reproducible.

Maybe something to look into another day.

lvmcache by kernel version

Dragging things back to the point of this article: previously we discovered that lvmcache worked great the first time through, when its cache volume was completely empty, but then subsequent runs all absolutely sucked. They didn’t perform significantly better than HDD baseline.

Let’s graph all the lvmcache results for each kernel version against the SSD baseline for that kernel to see if things changed at all.

lvmcache 4.9.30-2

This is the similar to what we saw before: an empty cache volume produces decent results of around 47k IOPS. Although it’s interesting that the second job is down around 1k IOPS. Again the results on a full cache are poor. In fact the results for the second job of the empty cache are about the same as the results for both jobs on a full cache.

lvmcache 4.11.6-1

Same story again here, although the performance is a little higher. Again the first job on an empty cache is getting the big results of almost 60k IOPS while the second job—and both jobs on a full cache—get only around 1k IOPS.

lvmcache 4.12.3

Wow. Something dramatic has been fixed. The performance on an empty cache is still better, but crucially the performance on a full cache pretty quickly becomes very close to baseline SSD.

Also the runs against both the empty and full cache device result in both jobs getting roughly the same IOPS performance rather than the first job being great and all others very poor.

What’s next?

It’s really encouraging that the performance is so much better with 4.12.3. It’s changed lvmcache from a “hmm, maybe” option to one that I would strongly consider using anywhere I could.

It’s a shame though that such a new kernel is required. The kernel version in Debian testing (buster) is currently 4.11.6-1. Debian experimental’s linux-image-4.12.0-trunk-amd64 package currently has version 4.12.2-1 so I should test if that is new enough I tested to see if that was new enough.

Failing that I think I should git bisect or similar in order to find out exactly which changeset fixed this, so I could have some chance of knowing when it hits a packaged version.

Test output Kernel 4.12.3 Baseline SSD (4.12.3) $ (cd /srv/ssd/fio && FIOTEST=ssd-4.12.3 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [407.9MB/0KB/0KB /s] [104K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=781: Wed Jul 26 19:03:54 2017 read : io=237280MB, bw=202479KB/s, iops=50619, runt=1200001msec slat (usec): min=0, max=2094, avg= 4.55, stdev= 2.02 clat (usec): min=36, max=8139, avg=151.16, stdev=42.74 lat (usec): min=94, max=8140, avg=156.24, stdev=42.71 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 111], 10.00th=[ 117], 20.00th=[ 123], | 30.00th=[ 129], 40.00th=[ 133], 50.00th=[ 139], 60.00th=[ 147], | 70.00th=[ 157], 80.00th=[ 173], 90.00th=[ 201], 95.00th=[ 233], | 99.00th=[ 314], 99.50th=[ 338], 99.90th=[ 406], 99.95th=[ 430], | 99.99th=[ 502] lat (usec) : 50=0.01%, 100=0.10%, 250=96.22%, 500=3.67%, 750=0.01% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01% cpu : usr=12.41%, sys=35.81%, ctx=33753770, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=60743781/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=782: Wed Jul 26 19:03:54 2017 read : io=251874MB, bw=214932KB/s, iops=53733, runt=1200001msec slat (usec): min=0, max=1050, avg= 4.58, stdev= 1.99 clat (usec): min=25, max=8148, avg=142.00, stdev=31.19 lat (usec): min=64, max=8151, avg=147.10, stdev=31.17 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 109], 10.00th=[ 114], 20.00th=[ 120], | 30.00th=[ 125], 40.00th=[ 131], 50.00th=[ 135], 60.00th=[ 141], | 70.00th=[ 149], 80.00th=[ 161], 90.00th=[ 179], 95.00th=[ 199], | 99.00th=[ 243], 99.50th=[ 262], 99.90th=[ 314], 99.95th=[ 334], | 99.99th=[ 414] lat (usec) : 50=0.01%, 100=0.32%, 250=98.86%, 500=0.81%, 750=0.01% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01% cpu : usr=12.91%, sys=38.04%, ctx=34484181, majf=1, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=64479700/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=489154MB, aggrb=417411KB/s, minb=202479KB/s, maxb=214932KB/s, mint=1200001msec, maxt=1200001msec Disk stats (read/write): xvda: ios=125216086/42, merge=14/33, ticks=18252156/0, in_queue=18285152, util=100.00%

Combined average IOPS were 104,352 or 52,176 per job.

Baseline HDD (4.12.3) $ (cd /srv/slow/fio && FIOTEST=hdd-4.12.3 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [3711KB/0KB/0KB /s] [927/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=2125: Wed Jul 26 22:41:52 2017 read : io=2194.8MB, bw=1872.8KB/s, iops=468, runt=1200071msec slat (usec): min=1, max=125, avg=14.18, stdev= 8.85 clat (usec): min=54, max=743383, avg=17067.31, stdev=35286.68 lat (usec): min=65, max=743393, avg=17082.36, stdev=35286.75 clat percentiles (usec): | 1.00th=[ 70], 5.00th=[ 77], 10.00th=[ 87], 20.00th=[ 109], | 30.00th=[ 139], 40.00th=[ 205], 50.00th=[ 1012], 60.00th=[ 1560], | 70.00th=[15424], 80.00th=[26752], 90.00th=[50944], 95.00th=[82432], | 99.00th=[168960], 99.50th=[214016], 99.90th=[313344], 99.95th=[366592], | 99.99th=[481280] lat (usec) : 100=15.90%, 250=25.89%, 500=3.68%, 750=1.82%, 1000=2.48% lat (msec) : 2=10.76%, 4=0.62%, 10=3.27%, 20=10.53%, 50=14.74% lat (msec) : 100=6.77%, 250=3.24%, 500=0.28%, 750=0.01% cpu : usr=0.50%, sys=1.08%, ctx=549488, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=561853/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=2126: Wed Jul 26 22:41:52 2017 read : io=1975.2MB, bw=1685.4KB/s, iops=421, runt=1200063msec slat (usec): min=1, max=169, avg=14.37, stdev= 8.96 clat (usec): min=42, max=894017, avg=18966.69, stdev=38900.32 lat (usec): min=67, max=894053, avg=18981.93, stdev=38900.42 clat percentiles (usec): | 1.00th=[ 70], 5.00th=[ 77], 10.00th=[ 87], 20.00th=[ 110], | 30.00th=[ 143], 40.00th=[ 213], 50.00th=[ 1064], 60.00th=[ 5984], | 70.00th=[16768], 80.00th=[29056], 90.00th=[57088], 95.00th=[91648], | 99.00th=[189440], 99.50th=[234496], 99.90th=[346112], 99.95th=[399360], | 99.99th=[501760] lat (usec) : 50=0.01%, 100=15.65%, 250=25.63%, 500=3.66%, 750=1.72% lat (usec) : 1000=2.30% lat (msec) : 2=10.10%, 4=0.48%, 10=3.19%, 20=10.79%, 50=14.82% lat (msec) : 100=7.33%, 250=3.93%, 500=0.39%, 750=0.01%, 1000=0.01% cpu : usr=0.48%, sys=0.96%, ctx=494683, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=505636/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=4169.9MB, aggrb=3558KB/s, minb=1685KB/s, maxb=1872KB/s, mint=1200063msec, maxt=1200071msec Disk stats (read/write): xvde: ios=1067472/11, merge=0/17, ticks=19179732/2528, in_queue=19182776, util=100.00%

Average combined IOPS were 889.

lvmcache (4.12.3, empty cache device) $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.12.3 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=0): [f(2)] [100.0% done] [395.1MB/0KB/0KB /s] [101K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1861: Wed Jul 26 21:43:37 2017 read : io=233021MB, bw=198845KB/s, iops=49711, runt=1200001msec slat (usec): min=1, max=696, avg= 6.62, stdev= 2.57 clat (usec): min=32, max=536652, avg=151.86, stdev=102.22 lat (usec): min=71, max=536660, avg=158.96, stdev=102.21 clat percentiles (usec): | 1.00th=[ 105], 5.00th=[ 111], 10.00th=[ 117], 20.00th=[ 123], | 30.00th=[ 129], 40.00th=[ 133], 50.00th=[ 139], 60.00th=[ 147], | 70.00th=[ 157], 80.00th=[ 173], 90.00th=[ 203], 95.00th=[ 239], | 99.00th=[ 326], 99.50th=[ 362], 99.90th=[ 454], 99.95th=[ 510], | 99.99th=[ 1128] lat (usec) : 50=0.01%, 100=0.09%, 250=95.76%, 500=4.10%, 750=0.03% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 750=0.01% cpu : usr=12.28%, sys=45.34%, ctx=27534900, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=59653437/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1862: Wed Jul 26 21:43:37 2017 read : io=218979MB, bw=186862KB/s, iops=46715, runt=1200001msec slat (usec): min=1, max=2349, avg= 6.67, stdev= 2.64 clat (usec): min=26, max=512486, avg=162.12, stdev=1013.93 lat (usec): min=64, max=512495, avg=169.27, stdev=1014.06 clat percentiles (usec): | 1.00th=[ 98], 5.00th=[ 109], 10.00th=[ 114], 20.00th=[ 121], | 30.00th=[ 126], 40.00th=[ 131], 50.00th=[ 137], 60.00th=[ 143], | 70.00th=[ 153], 80.00th=[ 165], 90.00th=[ 187], 95.00th=[ 209], | 99.00th=[ 274], 99.50th=[ 306], 99.90th=[ 450], 99.95th=[ 9664], | 99.99th=[44800] lat (usec) : 50=0.01%, 100=1.07%, 250=97.19%, 500=1.66%, 750=0.01% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.02%, 50=0.02% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01% cpu : usr=11.90%, sys=42.52%, ctx=25462344, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=56058591/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=452000MB, aggrb=385706KB/s, minb=186861KB/s, maxb=198844KB/s, mint=1200001msec, maxt=1200001msec Disk stats (read/write): dm-2: ios=115688209/28, merge=0/0, ticks=18037068/4148, in_queue=18097348, util=100.00%, aggrios=38592557/60889, aggrmerge=0/0, aggrticks=6093288/367342, aggrin_queue=6477849, aggrutil=100.00% dm-1: ios=23/117017, merge=0/0, ticks=4/19880, in_queue=19888, util=1.27%, aggrios=115726125/113023, aggrmerge=2/18688, aggrticks=16993500/21068, aggrin_queue=17022812, aggrutil=100.00% xvdc: ios=115726125/113023, merge=2/18688, ticks=16993500/21068, in_queue=17022812, util=100.00% dm-0: ios=115726104/16785, merge=0/0, ticks=16982820/3328, in_queue=17037740, util=100.00% dm-3: ios=51545/48866, merge=0/0, ticks=1297040/1078820, in_queue=2375920, util=20.61%, aggrios=51545/48849, aggrmerge=0/17, aggrticks=1297044/1075220, aggrin_queue=2372260, aggrutil=20.61% xvdd: ios=51545/48849, merge=0/17, ticks=1297044/1075220, in_queue=2372260, util=20.61% lvmcache (4.12.3, full cache device) $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.12.3 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [391.3MB/0KB/0KB /s] [100K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1988: Wed Jul 26 22:13:33 2017 read : io=142920MB, bw=121958KB/s, iops=30489, runt=1200001msec slat (usec): min=0, max=1489, avg= 6.76, stdev= 2.74 clat (usec): min=27, max=1517.9K, avg=253.15, stdev=3966.27 lat (usec): min=66, max=1517.1K, avg=260.40, stdev=3966.63 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 108], 10.00th=[ 113], 20.00th=[ 120], | 30.00th=[ 125], 40.00th=[ 131], 50.00th=[ 137], 60.00th=[ 145], | 70.00th=[ 155], 80.00th=[ 171], 90.00th=[ 203], 95.00th=[ 245], | 99.00th=[ 350], 99.50th=[ 402], 99.90th=[27776], 99.95th=[61696], | 99.99th=[179200] lat (usec) : 50=0.01%, 100=0.86%, 250=94.46%, 500=4.44%, 750=0.04% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.04%, 50=0.06% lat (msec) : 100=0.03%, 250=0.02%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=7.69%, sys=28.25%, ctx=17949976, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=36587548/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1989: Wed Jul 26 22:13:33 2017 read : io=142448MB, bw=121556KB/s, iops=30388, runt=1200001msec slat (usec): min=1, max=2795, avg= 6.79, stdev= 2.76 clat (usec): min=14, max=1271.5K, avg=253.96, stdev=4051.40 lat (usec): min=66, max=1271.5K, avg=261.25, stdev=4051.71 clat percentiles (usec): | 1.00th=[ 90], 5.00th=[ 108], 10.00th=[ 113], 20.00th=[ 120], | 30.00th=[ 125], 40.00th=[ 131], 50.00th=[ 137], 60.00th=[ 143], | 70.00th=[ 153], 80.00th=[ 169], 90.00th=[ 197], 95.00th=[ 231], | 99.00th=[ 330], 99.50th=[ 382], 99.90th=[28288], 99.95th=[63232], | 99.99th=[187392] lat (usec) : 20=0.01%, 50=0.01%, 100=1.41%, 250=94.94%, 500=3.42% lat (usec) : 750=0.03%, 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.04%, 50=0.06% lat (msec) : 100=0.03%, 250=0.02%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=7.79%, sys=28.22%, ctx=17393786, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=36466725/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=285368MB, aggrb=243514KB/s, minb=121555KB/s, maxb=121958KB/s, mint=1200001msec, maxt=1200001msec Disk stats (read/write): dm-2: ios=73045110/29, merge=0/0, ticks=18465212/5508, in_queue=18503620, util=100.00%, aggrios=24373349/191891, aggrmerge=0/0, aggrticks=6909396/40201, aggrin_queue=6960561, aggrutil=80.17% dm-1: ios=78/509945, merge=0/0, ticks=16/63772, in_queue=63792, util=3.22%, aggrios=72922899/556621, aggrmerge=0/17454, aggrticks=10752172/73676, aggrin_queue=10833040, aggrutil=81.60% xvdc: ios=72922899/556621, merge=0/17454, ticks=10752172/73676, in_queue=10833040, util=81.60% dm-0: ios=72922821/64279, merge=0/0, ticks=10747112/12240, in_queue=10792240, util=80.17% dm-3: ios=197149/1449, merge=0/0, ticks=9981060/44592, in_queue=10025652, util=46.75%, aggrios=197149/1433, aggrmerge=0/16, aggrticks=9981044/40808, aggrin_queue=10021852, aggrutil=46.75% xvdd: ios=197149/1433, merge=0/16, ticks=9981044/40808, in_queue=10021852, util=46.75% Kernel 4.11.6-1 Baseline SSD (4.11.6-1) $ (cd /srv/ssd/fio && FIOTEST=ssd-4.11.6-1 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [401.9MB/0KB/0KB /s] [103K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=2028: Wed Jul 26 18:01:52 2017 read : io=240561MB, bw=205278KB/s, iops=51319, runt=1200001msec slat (usec): min=1, max=231, avg= 4.95, stdev= 1.91 clat (usec): min=32, max=7892, avg=148.38, stdev=39.27 lat (usec): min=95, max=7896, avg=153.84, stdev=39.25 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 110], 10.00th=[ 115], 20.00th=[ 122], | 30.00th=[ 127], 40.00th=[ 133], 50.00th=[ 139], 60.00th=[ 145], | 70.00th=[ 155], 80.00th=[ 169], 90.00th=[ 193], 95.00th=[ 223], | 99.00th=[ 294], 99.50th=[ 326], 99.90th=[ 390], 99.95th=[ 410], | 99.99th=[ 474] lat (usec) : 50=0.01%, 100=0.12%, 250=97.04%, 500=2.83%, 750=0.01% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01% cpu : usr=13.47%, sys=38.85%, ctx=32087576, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=61583511/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=2029: Wed Jul 26 18:01:52 2017 read : io=241050MB, bw=205696KB/s, iops=51423, runt=1200001msec slat (usec): min=1, max=1065, avg= 4.98, stdev= 1.91 clat (usec): min=38, max=7985, avg=148.06, stdev=41.67 lat (usec): min=63, max=7992, avg=153.54, stdev=41.65 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 109], 10.00th=[ 113], 20.00th=[ 120], | 30.00th=[ 125], 40.00th=[ 131], 50.00th=[ 137], 60.00th=[ 143], | 70.00th=[ 153], 80.00th=[ 169], 90.00th=[ 197], 95.00th=[ 229], | 99.00th=[ 306], 99.50th=[ 334], 99.90th=[ 398], 99.95th=[ 418], | 99.99th=[ 486] lat (usec) : 50=0.01%, 100=0.14%, 250=96.43%, 500=3.41%, 750=0.01% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01% cpu : usr=13.70%, sys=38.78%, ctx=31782354, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=61708761/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=481610MB, aggrb=410973KB/s, minb=205278KB/s, maxb=205695KB/s, mint=1200001msec, maxt=1200001msec Disk stats (read/write): xvda: ios=123286829/13, merge=0/21, ticks=18148640/0, in_queue=18184316, util=100.00%

Combined average IOPS were 102,742 or 51,371 per job.

Baseline HDD (4.11.6-1) $ (cd /srv/slow/fio && FIOTEST=hdd-4.11.6-1 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [2108KB/0KB/0KB /s] [527/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=904: Thu Jul 27 09:17:21 2017 read : io=1218.3MB, bw=1039.6KB/s, iops=259, runt=1200088msec slat (usec): min=0, max=93, avg=14.37, stdev= 8.02 clat (usec): min=62, max=1126.5K, avg=30763.45, stdev=56969.73 lat (usec): min=68, max=1126.5K, avg=30778.70, stdev=56969.84 clat percentiles (usec): | 1.00th=[ 70], 5.00th=[ 82], 10.00th=[ 96], 20.00th=[ 129], | 30.00th=[ 175], 40.00th=[ 724], 50.00th=[ 4320], 60.00th=[15552], | 70.00th=[29824], 80.00th=[49408], 90.00th=[90624], 95.00th=[140288], | 99.00th=[272384], 99.50th=[333824], 99.90th=[489472], 99.95th=[561152], | 99.99th=[733184] lat (usec) : 100=11.11%, 250=23.85%, 500=3.45%, 750=1.71%, 1000=1.28% lat (msec) : 2=6.65%, 4=1.72%, 10=6.07%, 20=7.46%, 50=16.95% lat (msec) : 100=11.01%, 250=7.46%, 500=1.19%, 750=0.09%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.21%, sys=0.60%, ctx=303130, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=311877/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=905: Thu Jul 27 09:17:21 2017 read : io=1142.7MB, bw=998342B/s, iops=243, runt=1200098msec slat (usec): min=0, max=171, avg=14.44, stdev= 7.98 clat (usec): min=62, max=1288.1K, avg=32802.09, stdev=60366.24 lat (usec): min=65, max=1288.2K, avg=32817.41, stdev=60366.35 clat percentiles (usec): | 1.00th=[ 71], 5.00th=[ 83], 10.00th=[ 98], 20.00th=[ 133], | 30.00th=[ 183], 40.00th=[ 1064], 50.00th=[ 7840], 60.00th=[16768], | 70.00th=[30592], 80.00th=[51456], 90.00th=[95744], 95.00th=[148480], | 99.00th=[292864], 99.50th=[354304], 99.90th=[522240], 99.95th=[593920], | 99.99th=[749568] lat (usec) : 100=10.59%, 250=22.93%, 500=3.11%, 750=1.64%, 1000=1.14% lat (msec) : 2=6.16%, 4=1.30%, 10=5.73%, 20=10.31%, 50=16.53% lat (msec) : 100=11.09%, 250=7.93%, 500=1.43%, 750=0.11%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.20%, sys=0.57%, ctx=283410, majf=0, minf=14 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=292507/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=2360.9MB, aggrb=2014KB/s, minb=974KB/s, maxb=1039KB/s, mint=1200088msec, maxt=1200098msec Disk stats (read/write): xvde: ios=604364/11, merge=0/17, ticks=19188840/3004, in_queue=19192884, util=100.00%

Average combined IOPS were 502.

lvmcache (4.11.6-1, empty cache device) $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.11.6-1 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [229.7MB/0KB/0KB /s] [58.8K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1281: Wed Jul 26 16:36:01 2017 read : io=270469MB, bw=230800KB/s, iops=57700, runt=1200001msec slat (usec): min=1, max=665, avg= 6.06, stdev= 1.39 clat (usec): min=44, max=1129.8K, avg=130.33, stdev=356.51 lat (usec): min=98, max=1129.8K, avg=136.83, stdev=356.51 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 105], 10.00th=[ 107], 20.00th=[ 111], | 30.00th=[ 116], 40.00th=[ 120], 50.00th=[ 123], 60.00th=[ 126], | 70.00th=[ 133], 80.00th=[ 141], 90.00th=[ 161], 95.00th=[ 183], | 99.00th=[ 235], 99.50th=[ 258], 99.90th=[ 310], 99.95th=[ 346], | 99.99th=[ 940] lat (usec) : 50=0.01%, 100=0.32%, 250=99.05%, 500=0.60%, 750=0.01% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=13.58%, sys=47.73%, ctx=30980580, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=69240158/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1282: Wed Jul 26 16:36:01 2017 read : io=5438.4MB, bw=4640.4KB/s, iops=1160, runt=1200094msec slat (usec): min=3, max=501, avg= 8.90, stdev= 3.01 clat (usec): min=1, max=1109.4K, avg=6884.09, stdev=22209.48 lat (usec): min=66, max=1109.5K, avg=6893.50, stdev=22209.59 clat percentiles (usec): | 1.00th=[ 69], 5.00th=[ 76], 10.00th=[ 98], 20.00th=[ 111], | 30.00th=[ 118], 40.00th=[ 122], 50.00th=[ 127], 60.00th=[ 135], | 70.00th=[ 149], 80.00th=[ 564], 90.00th=[23680], 95.00th=[43776], | 99.00th=[99840], 99.50th=[125440], 99.90th=[197632], 99.95th=[276480], | 99.99th=[544768] lat (usec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (usec) : 100=10.37%, 250=68.11%, 500=1.27%, 750=0.77%, 1000=0.64% lat (msec) : 2=1.11%, 4=0.37%, 10=1.68%, 20=4.42%, 50=7.14% lat (msec) : 100=3.13%, 250=0.94%, 500=0.04%, 750=0.02%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.39%, sys=1.53%, ctx=1364109, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=1392218/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=275908MB, aggrb=235422KB/s, minb=4640KB/s, maxb=230800KB/s, mint=1200001msec, maxt=1200094msec Disk stats (read/write): dm-2: ios=70632376/28, merge=0/0, ticks=18853972/3148, in_queue=18878504, util=100.00%, aggrios=23553400/9408, aggrmerge=0/0, aggrticks=6264994/404240, aggrin_queue=6677382, aggrutil=100.00% dm-1: ios=1/372, merge=0/0, ticks=0/120, in_queue=120, util=0.01%, aggrios=70270018/409, aggrmerge=476/2, aggrticks=9314984/128, aggrin_queue=9321056, aggrutil=100.00% xvdc: ios=70270018/409, merge=476/2, ticks=9314984/128, in_queue=9321056, util=100.00% dm-0: ios=70270493/39, merge=0/0, ticks=9339988/4, in_queue=9364408, util=100.00% dm-3: ios=389707/27815, merge=0/0, ticks=9454996/1212596, in_queue=10667620, util=100.00%, aggrios=389707/27324, aggrmerge=0/491, aggrticks=9454912/1121628, aggrin_queue=10576496, aggrutil=100.00% xvdd: ios=389707/27324, merge=0/491, ticks=9454912/1121628, in_queue=10576496, util=100.00% lvmcache (4.11.6-1, full cache device) $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.11.6-1 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [4904KB/0KB/0KB /s] [1226/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1723: Wed Jul 26 17:03:19 2017 read : io=2713.5MB, bw=2315.3KB/s, iops=578, runt=1200126msec slat (usec): min=4, max=103, avg=14.92, stdev= 8.59 clat (usec): min=67, max=943720, avg=13802.04, stdev=37388.05 lat (usec): min=76, max=943732, avg=13817.72, stdev=37388.50 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 112], 10.00th=[ 115], 20.00th=[ 119], | 30.00th=[ 127], 40.00th=[ 131], 50.00th=[ 141], 60.00th=[ 155], | 70.00th=[ 201], 80.00th=[17280], 90.00th=[43776], 95.00th=[79360], | 99.00th=[181248], 99.50th=[234496], 99.90th=[366592], 99.95th=[428032], | 99.99th=[585728] lat (usec) : 100=0.35%, 250=72.25%, 500=0.73%, 750=0.26%, 1000=0.22% lat (msec) : 2=0.53%, 4=0.06%, 10=1.18%, 20=6.18%, 50=9.45% lat (msec) : 100=5.33%, 250=3.06%, 500=0.39%, 750=0.02%, 1000=0.01% cpu : usr=0.37%, sys=1.29%, ctx=688778, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=694643/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1724: Wed Jul 26 17:03:19 2017 read : io=2493.3MB, bw=2127.5KB/s, iops=531, runt=1200115msec slat (usec): min=3, max=113, avg=15.20, stdev= 8.84 clat (usec): min=29, max=929665, avg=15021.94, stdev=39251.61 lat (usec): min=76, max=929675, avg=15037.91, stdev=39252.04 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 112], 10.00th=[ 116], 20.00th=[ 124], | 30.00th=[ 129], 40.00th=[ 133], 50.00th=[ 143], 60.00th=[ 159], | 70.00th=[ 217], 80.00th=[19328], 90.00th=[48896], 95.00th=[85504], | 99.00th=[191488], 99.50th=[242688], 99.90th=[374784], 99.95th=[440320], | 99.99th=[602112] lat (usec) : 50=0.01%, 100=0.40%, 250=70.92%, 500=0.78%, 750=0.31% lat (usec) : 1000=0.23% lat (msec) : 2=0.60%, 4=0.08%, 10=1.17%, 20=5.98%, 50=9.80% lat (msec) : 100=5.83%, 250=3.46%, 500=0.43%, 750=0.03%, 1000=0.01% cpu : usr=0.35%, sys=1.20%, ctx=632752, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=638283/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=5206.8MB, aggrb=4442KB/s, minb=2127KB/s, maxb=2315KB/s, mint=1200115msec, maxt=1200126msec Disk stats (read/write): dm-2: ios=1332926/29, merge=0/0, ticks=19182084/4352, in_queue=19187080, util=100.00%, aggrios=445250/12578, aggrmerge=0/0, aggrticks=6224384/4612, aggrin_queue=6229029, aggrutil=100.00% dm-1: ios=60/34939, merge=0/0, ticks=16/6800, in_queue=6820, util=0.28%, aggrios=954236/35982, aggrmerge=0/1722, aggrticks=141920/7080, aggrin_queue=148992, aggrutil=11.70% xvdc: ios=954236/35982, merge=0/1722, ticks=141920/7080, in_queue=148992, util=11.70% dm-0: ios=954176/2767, merge=0/0, ticks=142036/720, in_queue=142780, util=11.53% dm-3: ios=381514/28, merge=0/0, ticks=18531100/6316, in_queue=18537488, util=100.00%, aggrios=381514/13, aggrmerge=0/15, aggrticks=18530984/2632, aggrin_queue=18533840, aggrutil=100.00% xvdd: ios=381514/13, merge=0/15, ticks=18530984/2632, in_queue=18533840, util=100.00% Kernel 4.9.30-2 Baseline SSD (4.9.30-2) $ (cd /srv/ssd/fio && FIOTEST=ssd-4.9.30-2 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=0): [f(2)] [100.0% done] [399.8MB/0KB/0KB /s] [102K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=770: Thu Jul 27 02:27:43 2017 read : io=241655MB, bw=206212KB/s, iops=51553, runt=1200001msec slat (usec): min=1, max=1046, avg= 4.42, stdev= 1.59 clat (usec): min=28, max=4828, avg=148.39, stdev=42.92 lat (usec): min=95, max=4830, avg=153.26, stdev=42.90 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 109], 10.00th=[ 113], 20.00th=[ 120], | 30.00th=[ 125], 40.00th=[ 131], 50.00th=[ 137], 60.00th=[ 143], | 70.00th=[ 153], 80.00th=[ 169], 90.00th=[ 197], 95.00th=[ 235], | 99.00th=[ 322], 99.50th=[ 350], 99.90th=[ 422], 99.95th=[ 454], | 99.99th=[ 532] lat (usec) : 50=0.01%, 100=0.15%, 250=96.00%, 500=3.84%, 750=0.01% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01% cpu : usr=13.51%, sys=36.36%, ctx=33974378, majf=1, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=61863654/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=771: Thu Jul 27 02:27:43 2017 read : io=238123MB, bw=203198KB/s, iops=50799, runt=1200001msec slat (usec): min=1, max=997, avg= 4.39, stdev= 1.60 clat (usec): min=42, max=4827, avg=150.72, stdev=47.68 lat (usec): min=65, max=4832, avg=155.56, stdev=47.66 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 109], 10.00th=[ 113], 20.00th=[ 120], | 30.00th=[ 125], 40.00th=[ 131], 50.00th=[ 137], 60.00th=[ 143], | 70.00th=[ 155], 80.00th=[ 171], 90.00th=[ 203], 95.00th=[ 249], | 99.00th=[ 342], 99.50th=[ 386], 99.90th=[ 458], 99.95th=[ 486], | 99.99th=[ 564] lat (usec) : 50=0.01%, 100=0.14%, 250=94.88%, 500=4.95%, 750=0.03% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01% cpu : usr=13.44%, sys=35.68%, ctx=34074578, majf=0, minf=18 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=60959403/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=479778MB, aggrb=409409KB/s, minb=203197KB/s, maxb=206212KB/s, mint=1200001msec, maxt=1200001msec Disk stats (read/write): xvda: ios=122797792/38, merge=0/33, ticks=18273356/0, in_queue=18319740, util=100.00%

Combined average IOPS were 102,352 or 51,176 per job.

Baseline HDD (4.9.30-2) $ (cd /srv/slow/fio && FIOTEST=hdd-4.9.30-2 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [3847KB/0KB/0KB /s] [961/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=1290: Thu Jul 27 02:57:41 2017 read : io=2348.2MB, bw=2003.7KB/s, iops=500, runt=1200052msec slat (usec): min=2, max=88, avg=12.51, stdev= 7.84 clat (usec): min=28, max=877693, avg=15952.77, stdev=32920.15 lat (usec): min=65, max=877697, avg=15966.13, stdev=32920.23 clat percentiles (usec): | 1.00th=[ 69], 5.00th=[ 77], 10.00th=[ 86], 20.00th=[ 108], | 30.00th=[ 139], 40.00th=[ 207], 50.00th=[ 996], 60.00th=[ 1464], | 70.00th=[14912], 80.00th=[25472], 90.00th=[47872], 95.00th=[76288], | 99.00th=[156672], 99.50th=[197632], 99.90th=[296960], 99.95th=[346112], | 99.99th=[473088] lat (usec) : 50=0.01%, 100=15.94%, 250=25.74%, 500=3.72%, 750=1.98% lat (usec) : 1000=2.70% lat (msec) : 2=11.26%, 4=0.46%, 10=3.02%, 20=11.04%, 50=14.68% lat (msec) : 100=6.41%, 250=2.83%, 500=0.20%, 750=0.01%, 1000=0.01% cpu : usr=0.47%, sys=1.14%, ctx=589357, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=601124/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=1291: Thu Jul 27 02:57:41 2017 read : io=2094.6MB, bw=1787.3KB/s, iops=446, runt=1200065msec slat (usec): min=2, max=136, avg=12.86, stdev= 8.11 clat (usec): min=57, max=1113.9K, avg=17886.33, stdev=37285.63 lat (usec): min=62, max=1113.1K, avg=17900.05, stdev=37285.72 clat percentiles (usec): | 1.00th=[ 69], 5.00th=[ 77], 10.00th=[ 87], 20.00th=[ 109], | 30.00th=[ 141], 40.00th=[ 209], 50.00th=[ 1032], 60.00th=[ 1832], | 70.00th=[16064], 80.00th=[27520], 90.00th=[52992], 95.00th=[85504], | 99.00th=[179200], 99.50th=[226304], 99.90th=[346112], 99.95th=[391168], | 99.99th=[536576] lat (usec) : 100=15.88%, 250=25.66%, 500=3.65%, 750=1.82%, 1000=2.45% lat (msec) : 2=10.62%, 4=0.32%, 10=2.87%, 20=11.21%, 50=14.75% lat (msec) : 100=6.98%, 250=3.46%, 500=0.34%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.39%, sys=1.08%, ctx=526250, majf=0, minf=14 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=536196/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=4442.7MB, aggrb=3790KB/s, minb=1787KB/s, maxb=2003KB/s, mint=1200052msec, maxt=1200065msec Disk stats (read/write): xvde: ios=1137197/11, merge=0/17, ticks=19176720/2496, in_queue=19179612, util=100.00%

Average combined IOPS were 946.

lvmcache (4.9.30-2, empty cache device) $ (cd /srv/cache/fio && FIOTEST=lvmcache-empty-4.9.30-2 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [93173KB/0KB/0KB /s] [23.3K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=8863: Wed Jul 26 15:33:04 2017 read : io=222011MB, bw=189442KB/s, iops=47360, runt=1200047msec slat (usec): min=1, max=536, avg= 5.50, stdev= 1.29 clat (usec): min=45, max=1066.2K, avg=161.37, stdev=2258.81 lat (usec): min=70, max=1066.2K, avg=167.26, stdev=2258.82 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 104], 10.00th=[ 106], 20.00th=[ 109], | 30.00th=[ 114], 40.00th=[ 118], 50.00th=[ 121], 60.00th=[ 124], | 70.00th=[ 129], 80.00th=[ 137], 90.00th=[ 165], 95.00th=[ 197], | 99.00th=[ 282], 99.50th=[ 322], 99.90th=[ 422], 99.95th=[ 1176], | 99.99th=[90624] lat (usec) : 50=0.01%, 100=0.30%, 250=97.69%, 500=1.93%, 750=0.01% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.02% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=10.37%, sys=37.16%, ctx=31754180, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=56834728/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=8864: Wed Jul 26 15:33:04 2017 read : io=6439.3MB, bw=5494.3KB/s, iops=1373, runt=1200133msec slat (usec): min=2, max=363, avg= 7.53, stdev= 2.49 clat (usec): min=1, max=1082.6K, avg=5814.26, stdev=21389.31 lat (usec): min=67, max=1082.6K, avg=5822.26, stdev=21389.42 clat percentiles (usec): | 1.00th=[ 72], 5.00th=[ 97], 10.00th=[ 104], 20.00th=[ 109], | 30.00th=[ 116], 40.00th=[ 119], 50.00th=[ 123], 60.00th=[ 127], | 70.00th=[ 139], 80.00th=[ 318], 90.00th=[18304], 95.00th=[35072], | 99.00th=[89600], 99.50th=[120320], 99.90th=[218112], 99.95th=[337920], | 99.99th=[593920] lat (usec) : 2=0.01%, 4=0.01%, 20=0.01%, 50=0.01%, 100=5.70% lat (usec) : 250=73.84%, 500=1.37%, 750=0.81%, 1000=0.81% lat (msec) : 2=1.32%, 4=0.26%, 10=1.52%, 20=5.19%, 50=6.12% lat (msec) : 100=2.26%, 250=0.72%, 500=0.05%, 750=0.03%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.43%, sys=1.55%, ctx=1616527, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=1648391/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=228450MB, aggrb=194922KB/s, minb=5494KB/s, maxb=189441KB/s, mint=1200047msec, maxt=1200133msec Disk stats (read/write): dm-2: ios=58483119/28, merge=0/0, ticks=18951620/1096, in_queue=18967248, util=100.00%, aggrios=19503443/9385, aggrmerge=0/0, aggrticks=6300254/611164, aggrin_queue=6914877, aggrutil=100.00% dm-1: ios=1/917, merge=0/0, ticks=0/316, in_queue=316, util=0.01%, aggrios=58032708/925, aggrmerge=600/80, aggrticks=7695844/316, aggrin_queue=7697340, aggrutil=99.47% xvdc: ios=58032708/925, merge=600/80, ticks=7695844/316, in_queue=7697340, util=99.47% dm-0: ios=58033307/88, merge=0/0, ticks=7713444/8, in_queue=7724700, util=99.46% dm-3: ios=477022/27152, merge=0/0, ticks=11187320/1833168, in_queue=13019616, util=100.00%, aggrios=477022/26569, aggrmerge=0/598, aggrticks=11187220/1810144, aggrin_queue=12996476, aggrutil=100.00% xvdd: ios=477022/26569, merge=0/598, ticks=11187220/1810144, in_queue=12996476, util=100.00% lvmcache (4.9.30-2, full cache device) $ (cd /srv/cache/fio && FIOTEST=lvmcache-full-4.9.30-2 fio ~/partial.fio) partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [5953KB/0KB/0KB /s] [1488/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=9014: Wed Jul 26 16:09:32 2017 read : io=3180.6MB, bw=2713.1KB/s, iops=678, runt=1200039msec slat (usec): min=3, max=172, avg=13.34, stdev= 7.73 clat (usec): min=65, max=1177.6K, avg=11773.54, stdev=32057.40 lat (usec): min=72, max=1177.7K, avg=11787.57, stdev=32057.79 clat percentiles (usec): | 1.00th=[ 105], 5.00th=[ 113], 10.00th=[ 118], 20.00th=[ 125], | 30.00th=[ 131], 40.00th=[ 133], 50.00th=[ 145], 60.00th=[ 159], | 70.00th=[ 213], 80.00th=[16512], 90.00th=[36096], 95.00th=[63232], | 99.00th=[148480], 99.50th=[197632], 99.90th=[337920], 99.95th=[415744], | 99.99th=[610304] lat (usec) : 100=0.45%, 250=70.90%, 500=0.60%, 750=0.28%, 1000=0.30% lat (msec) : 2=0.67%, 4=0.04%, 10=1.48%, 20=8.16%, 50=10.28% lat (msec) : 100=4.53%, 250=2.07%, 500=0.23%, 750=0.02%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.34%, sys=1.45%, ctx=806505, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=814209/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=9015: Wed Jul 26 16:09:32 2017 read : io=3083.6MB, bw=2631.2KB/s, iops=657, runt=1200059msec slat (usec): min=3, max=251, avg=13.31, stdev= 7.88 clat (usec): min=65, max=1195.8K, avg=12144.40, stdev=32642.85 lat (usec): min=72, max=1195.8K, avg=12158.39, stdev=32643.28 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 111], 10.00th=[ 115], 20.00th=[ 122], | 30.00th=[ 127], 40.00th=[ 133], 50.00th=[ 143], 60.00th=[ 159], | 70.00th=[ 221], 80.00th=[17024], 90.00th=[37632], 95.00th=[65280], | 99.00th=[152576], 99.50th=[199680], 99.90th=[342016], 99.95th=[419840], | 99.99th=[610304] lat (usec) : 100=0.49%, 250=70.26%, 500=0.58%, 750=0.27%, 1000=0.30% lat (msec) : 2=0.77%, 4=0.04%, 10=1.51%, 20=8.18%, 50=10.43% lat (msec) : 100=4.74%, 250=2.17%, 500=0.24%, 750=0.02%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.43%, sys=1.30%, ctx=782175, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=789392/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=6264.7MB, aggrb=5345KB/s, minb=2631KB/s, maxb=2713KB/s, mint=1200039msec, maxt=1200059msec Disk stats (read/write): dm-2: ios=1603515/29, merge=0/0, ticks=19178856/4820, in_queue=19184536, util=100.00%, aggrios=535452/12461, aggrmerge=0/0, aggrticks=6232844/4946, aggrin_queue=6237806, aggrutil=100.00% dm-1: ios=59/34655, merge=0/0, ticks=16/7544, in_queue=7560, util=0.29%, aggrios=1124948/36669, aggrmerge=0/683, aggrticks=166868/8120, aggrin_queue=174932, aggrutil=13.61% xvdc: ios=1124948/36669, merge=0/683, ticks=166868/8120, in_queue=174932, util=13.61% dm-0: ios=1124889/2697, merge=0/0, ticks=166996/736, in_queue=167772, util=13.44% dm-3: ios=481409/31, merge=0/0, ticks=18531520/6560, in_queue=18538088, util=100.00%, aggrios=481409/14, aggrmerge=0/17, aggrticks=18531404/2968, aggrin_queue=18534356, aggrutil=100.00% xvdd: ios=481409/14, merge=0/17, ticks=18531404/2968, in_queue=18534356, util=100.00%
Categories: LUG Community Blogs

Debian Bits: DebConf17 Schedule Published!

Tue, 25/07/2017 - 00:15

The DebConf17 orga team is proud to announce that over 120 activities have been scheduled so far, including 45- and 20-minute talks, team meetings, and workshops, as well as a variety of other events.

Most of the talks and BoFs will be streamed and recorded, thanks to our amazing video team!

We'd like to remind you that Saturday August 5th is also our much anticipated Open Day! This means a program for a wider audience, including special activities for newcomers, such as an AMA session about Debian, a beginners workshop on packaging, a thoughtful talk about freedom with regard to today's popular gadgets and more.

In addition to the published schedule, we'll provide rooms for ad-hoc sessions where attendees will be able to schedule activities at any time during the whole conference.

The current schedule is available at https://debconf17.debconf.org/schedule/

This is also available through an XML feed. You can use ConfClerk in Debian to consume this, or Giggity on Android devices: https://debconf17.debconf.org/schedule/mobile/

We look forward to seeing you in Montreal!

Categories: LUG Community Blogs

Andy Smith: 12 hours of lvmcache

Fri, 21/07/2017 - 20:59

In the previous post I noted that the performance of lvmcache was still increasing and it might be worth testing it for longer than 3 hours.

Here’a a 12 hour test $ cd /srv/cache/fio && FIOTEST=lvmcache-12h fio ~/lvmcache-12h.fio partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [6272KB/0KB/0KB /s] [1568/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=11130: Fri Jul 21 16:37:30 2017 read : io=136145MB, bw=3227.2KB/s, iops=806, runt=43200062msec slat (usec): min=3, max=586402, avg=14.27, stdev=619.54 clat (usec): min=2, max=1517.9K, avg=9897.80, stdev=29334.14 lat (usec): min=71, max=1517.9K, avg=9912.72, stdev=29344.74 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 110], 10.00th=[ 113], 20.00th=[ 119], | 30.00th=[ 124], 40.00th=[ 129], 50.00th=[ 133], 60.00th=[ 143], | 70.00th=[ 157], 80.00th=[11840], 90.00th=[30848], 95.00th=[56576], | 99.00th=[136192], 99.50th=[179200], 99.90th=[309248], 99.95th=[382976], | 99.99th=[577536] lat (usec) : 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01%, 100=0.03% lat (usec) : 250=76.84%, 500=0.26%, 750=0.13%, 1000=0.13% lat (msec) : 2=0.25%, 4=0.02%, 10=1.19%, 20=6.67%, 50=8.59% lat (msec) : 100=3.93%, 250=1.77%, 500=0.18%, 750=0.02%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.44%, sys=1.63%, ctx=34524570, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=34853153/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=11131: Fri Jul 21 16:37:30 2017 read : io=134521MB, bw=3188.7KB/s, iops=797, runt=43200050msec slat (usec): min=3, max=588479, avg=14.35, stdev=613.38 clat (usec): min=2, max=1530.3K, avg=10017.42, stdev=29196.28 lat (usec): min=70, max=1530.3K, avg=10032.43, stdev=29207.06 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 109], 10.00th=[ 112], 20.00th=[ 118], | 30.00th=[ 124], 40.00th=[ 127], 50.00th=[ 133], 60.00th=[ 143], | 70.00th=[ 157], 80.00th=[12352], 90.00th=[31360], 95.00th=[57600], | 99.00th=[138240], 99.50th=[179200], 99.90th=[301056], 99.95th=[370688], | 99.99th=[561152] lat (usec) : 4=0.01%, 20=0.01%, 50=0.01%, 100=0.04%, 250=76.56% lat (usec) : 500=0.26%, 750=0.12%, 1000=0.13% lat (msec) : 2=0.26%, 4=0.02%, 10=1.20%, 20=6.75%, 50=8.65% lat (msec) : 100=4.01%, 250=1.82%, 500=0.17%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.45%, sys=1.60%, ctx=34118324, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=34437257/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=270666MB, aggrb=6415KB/s, minb=3188KB/s, maxb=3227KB/s, mint=43200050msec, maxt=43200062msec Disk stats (read/write): dm-2: ios=69290239/1883, merge=0/0, ticks=690078104/246240, in_queue=690329868, util=100.00%, aggrios=23098543/27863, aggrmerge=0/0, aggrticks=229728200/637782, aggrin_queue=230366965, aggrutil=100.00% dm-1: ios=247/64985, merge=0/0, ticks=36/15464, in_queue=15504, util=0.02%, aggrios=53025553/63449, aggrmerge=0/7939, aggrticks=7413340/14760, aggrin_queue=7427028, aggrutil=16.42% xvdc: ios=53025553/63449, merge=0/7939, ticks=7413340/14760, in_queue=7427028, util=16.42% dm-0: ios=53025306/6403, merge=0/0, ticks=7417028/1852, in_queue=7419784, util=16.42% dm-3: ios=16270078/12201, merge=0/0, ticks=681767536/1896032, in_queue=683665608, util=100.00%, aggrios=16270077/12200, aggrmerge=1/1, aggrticks=681637744/1813744, aggrin_queue=683453224, aggrutil=100.00% xvdd: ios=16270077/12200, merge=1/1, ticks=681637744/1813744, in_queue=683453224, util=100.00%

It’s still going up, slowly. The cache hit rate was 76.53%. In the 30 minute test the hit rate was 73.64%.

Over 30 minutes the average IOPS was 1,484.

Over 12 hours the average IOPS was 1,603.

I was kind of hoping to reach the point where the hit rate is so high that it just takes off like bcache does and we start to see tens of thousands of IOPS, but it wasn’t to be.

24 hours of lvmcache

…so I went ahead and ran the same thing for 24 hours.

I’ve skipped the first 2 hours of results since we know what they look like. It appears to still be going up, although the results past 20 hours leave some doubt there.

Here’s the full fio output.

$ cd /srv/cache/fio && FIOTEST=lvmcache-24h fio ~/lvmcache-24h.fio partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [7152KB/0KB/0KB /s] [1788/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=14676: Sat Jul 22 21:34:12 2017 read : io=278655MB, bw=3302.6KB/s, iops=825, runt=86400091msec slat (usec): min=3, max=326, avg=12.43, stdev= 6.97 clat (usec): min=1, max=1524.1K, avg=9673.02, stdev=28748.45 lat (usec): min=71, max=1525.7K, avg=9686.11, stdev=28748.87 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 106], 10.00th=[ 111], 20.00th=[ 116], | 30.00th=[ 119], 40.00th=[ 125], 50.00th=[ 131], 60.00th=[ 139], | 70.00th=[ 155], 80.00th=[11456], 90.00th=[30336], 95.00th=[55552], | 99.00th=[134144], 99.50th=[177152], 99.90th=[305152], 99.95th=[374784], | 99.99th=[569344] lat (usec) : 2=0.01%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (usec) : 100=0.03%, 250=77.16%, 500=0.22%, 750=0.12%, 1000=0.12% lat (msec) : 2=0.23%, 4=0.02%, 10=1.18%, 20=6.65%, 50=8.54% lat (msec) : 100=3.84%, 250=1.70%, 500=0.17%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.47%, sys=1.64%, ctx=70653446, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=71335660/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=14677: Sat Jul 22 21:34:12 2017 read : io=283280MB, bw=3357.4KB/s, iops=839, runt=86400074msec slat (usec): min=3, max=330, avg=12.44, stdev= 6.98 clat (usec): min=2, max=1515.9K, avg=9514.83, stdev=28128.86 lat (usec): min=71, max=1515.2K, avg=9527.92, stdev=28129.29 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 109], 10.00th=[ 112], 20.00th=[ 118], | 30.00th=[ 123], 40.00th=[ 126], 50.00th=[ 133], 60.00th=[ 141], | 70.00th=[ 157], 80.00th=[11328], 90.00th=[29824], 95.00th=[55040], | 99.00th=[132096], 99.50th=[173056], 99.90th=[292864], 99.95th=[362496], | 99.99th=[544768] lat (usec) : 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01%, 100=0.03% lat (usec) : 250=77.29%, 500=0.23%, 750=0.11%, 1000=0.12% lat (msec) : 2=0.23%, 4=0.02%, 10=1.18%, 20=6.65%, 50=8.49% lat (msec) : 100=3.81%, 250=1.66%, 500=0.15%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.47%, sys=1.67%, ctx=71794214, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=72519640/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=561935MB, aggrb=6659KB/s, minb=3302KB/s, maxb=3357KB/s, mint=86400074msec, maxt=86400091msec Disk stats (read/write): dm-2: ios=143855123/29, merge=0/0, ticks=1380761492/1976, in_queue=1380772508, util=100.00%, aggrios=47953627/25157, aggrmerge=0/0, aggrticks=459927326/7329, aggrin_queue=459937080, aggrutil=100.00% dm-1: ios=314/70172, merge=0/0, ticks=40/15968, in_queue=16008, util=0.01%, aggrios=110839338/72760, aggrmerge=0/2691, aggrticks=15300392/17100, aggrin_queue=15315432, aggrutil=16.92% xvdc: ios=110839338/72760, merge=0/2691, ticks=15300392/17100, in_queue=15315432, util=16.92% dm-0: ios=110839024/5279, merge=0/0, ticks=15308540/1768, in_queue=15312588, util=16.93% dm-3: ios=33021544/20, merge=0/0, ticks=1364473400/4252, in_queue=1364482644, util=100.00%, aggrios=33021544/19, aggrmerge=0/1, aggrticks=1364468920/4076, aggrin_queue=1364476064, aggrutil=100.00% xvdd: ios=33021544/19, merge=0/1, ticks=1364468920/4076, in_queue=1364476064, util=100.00%

So, 1,664 average IOPS (825 + 839), 77.05% (110,839,338 / (71,335,660 + 72,519,640)*100) cache hit rate.

Not sure I can be bothered to run a multi-day test on this now!

Categories: LUG Community Blogs

Andy Smith: bcache and lvmcache

Wed, 19/07/2017 - 07:06
Background

Over at BitFolk we offer both SSD-backed storage and HDD-backed archive storage. The SSDs we use are really nice and basically have removed all IO performance problems we have ever encountered in the past. I can’t begin to describe how pleasant it is to just never have to think about that whole class of problems.

The main downside of course is that SSD capacity is still really expensive. That’s why we introduced the HDD-backed archive storage: for bulk storage of things that didn’t need to have high performance.

You’d really think though that by now there would be some kind of commodity tiered storage that would allow a relatively small amount of fast SSD storage to accelerate a much larger pool of HDDs. Of course there are various enterprise solutions, and there is also ZFS where SSDs could be used for the ZIL and L2ARC while HDDs are used for the pool.

ZFS is quite appealing but I’m not ready to switch everything to that yet, and I’m certainly not going to buy some enterprise storage solution. I also don’t necessarily want to commit to putting all storage behind such a system.

I decided to explore Linux’s own block device caching solutions.

Scope

I’ve restricted the scope to the two solutions which are part of the mainline Linux kernel as of July 2017, these being bcache and lvmcache.

lvmcache is based upon dm-cache which has been included with the mainline kernel since April 2013. It’s quite conservative, and having been around for quite a while is considered stable. It has the advantage that it can work with any LVM logical volume no matter what the contents. That brings the disadvantage that you do need to run LVM.

bcache has been around for a little longer but is a much more ambitious project. Being completely dedicated to accelerating slow block devices with fast ones it is claimed to be able to achieve higher performance than other caching solutions, but as it’s much more complicated than dm-cache there are still bugs being found. Also it requires you format your block devices as bcache before you use them for anything.

Test environment

I’m testing this on a Debian testing (buster) Xen virtual machine with a 20GiB xvda virtual disk containing the main operating system. That disk is backed by a software (md) RAID-10 composed of two Samsung sm863 SSDs. It was also used for testing the baseline SSD performance from the directory /srv/ssd.

The virtual machine had 1GiB of memory but the pagecache was cleared between each test run in an attempt to prevent anything being cached in memory.

A 5GiB xvdc virtual disk was provided, backed again on the SSD RAID. This was used for the cache role both in bcache and lvmcache.

A 50GiB xvdd virtual disk was provided, backed by a pair of Seagate ST4000LM016-1N2170 HDDs in software RAID-1. This was used for the HDD backing store in each of the caching implementations. The resulting cache device was mounted at /srv/cache.

Finally a 50GiB xvde virtual disk also backed on HDD was used to test baseline HDD performance, mounted at /srv/slow.

The filesystem in use in all cases was ext4 with default options. In dom0, deadline scheduler was used in all cases.

TL;DR, I just want graphs

In case you can’t be bothered to read the rest of this article, here’s just the graphs with some attempt at interpreting them. Down at the tests section you’ll find details of the actual testing process and more commentary on why certain graphs were produced.

git test graphs

Times to git clone and git grep.

fio IOPS graphs

These are graphs of IOPS across the 30 minutes of testing. There’s two important things to note about these graphs:

  1. They’re a Bezier curve fitted to the data points which are one per second. The actual data points are all over the place, because achieved IOPS depends on how many cache hits/misses there were, which is statistical.
  2. Only the IOPS for the first job is graphed. Even when using the per_job_logs=0 setting my copy of fio writes a set of results for each job. I couldn’t work out how to easily combine these so I’ve shown only the results for the first job.

    For all tests except bcache (sequential_cutoff=0) you just have to bear in mind that there is a second job working in parallel doing pretty much the same amount of IOPS. Strangely for that second bcache test the second job only managed a fraction of the IOPS (though still more than 10k IOPS) and I don’t know why.

IOPS over time for all tests

Well, those results are so extreme that it kind of makes it hard to distinguish between the low-end results.

A couple of observations:

  • SSD is incredibly and consistently fast.
  • For everything else there is a short steep section at the beginning which is likely to be the effect of HDD drive cache.
  • With sequential_cutoff set to 0, bcache very quickly reaches near-SSD performance for this workload (4k reads, 90% hitting 10% of data that fits entirely in the bcache). This is probably because the initial write put data straight into cache as it’s set to writeback.
  • When starting with a completely empty cache, lvmcache is no slouch either. It’s not quite as performant as bcache but that is still up near the 48k IOPS per process region, and very predictable.
  • When sequential_cutoff is left at its default of 4M, bcache performs much worse though still blazing compared to an HDD on its own. At the end of this 30 minute test performance was still increasing so it might be worth performing a longer test
  • The performance of lvmcache when starting with a cache already full of junk data seems to be not that much better than HDD baseline.
IOPS over time for low-end results

Leaving the high-performers out to see if there is anything interesting going on near the bottom of the previous graph.

Apart from the initial spike, HDD results are flat as expected.

Although the lvmcache (full cache) results in the previous graph seemed flat too, looking closer we can see that performance is still increasing, just very slowly. It may be interesting to test for longer to see if performance does continue to increase.

Both HDD and lvmcache have a very similar spike at the start of the test so let’s look closer at that.

IOPS for first 30 seconds

For all the lower-end performers the first 19 seconds are steeper and I can only think this is the effect of HDD drive cache. Once that is filled, HDD remains basically flat, lvmcache (full cache) increases performance more slowly and bcache with the default sequential_cutoff starts to take off.

SSDs don’t have the same sort of cache and bcache with no sequential_cutoff spikes up too quickly to really be noticeable at this scale.

3-hour lvmcache test

Since it seemed like lvmcache with a full cache device was still slowly increasing in performance I did a 3-hour testing on that one.

Skipping the first 20 minutes which show stronger growth, even after 3 hours there is still some performance increase happening. It seems like even a full cache would eventually promote read hot spots, but it could take a very very long time.

The tests

This being caching, we would only really expect it to be effective for reads of already-known data. Sequential reads on an HDD aren’t too bad but what really makes things feel slow are random reads all over the place: the time taken to seek to a new location on an HDD is really a killer for performance. So, we’d like a test that involves multiple reads of the same data but non-sequential.

git tests

I decided to do a git clone of the Linux kernel and then do a git grep of this, which would be guaranteed to be random access. The git grep step was done repeatedly to see if we could spot the cache being warmed up, also with the kernel’s pagecache being cleared after each step to avoid a situation where the entirety of the kernel source ended up in RAM. The clone process was also timed just to see if there would be any effect on write time.

fio tests

Despite git being a real application, the workload itself (searching for a string over and over) seemed rather too synthetic and simplistic. I decided to also try using the excellent fio to gather some detailed statistics. This is also synthetic of course, relating to no real workload, but it’s quite sophisticated and hopefully of interest.

The fio job spec used in all cases was as follows:

[partial] ioengine=libaio direct=1 iodepth=8 numjobs=2 readwrite=randread random_distribution=zipf:1.2 bs=4k size=2g unlink=1 runtime=1800 time_based=1 per_job_logs=1 log_avg_msec=1000 write_iops_log=/var/tmp/fio-${FIOTEST}

This causes two 2GiB files of random data to be written out and then two processes read from them for 30 minutes with up to 8 IOs in flight at once for each process. Between the use of direct=1 and the fact that there is 4GiB of data and only 1GiB of RAM, that should be enough to avoid testing any memory cache.

log_avg_msec=1000 is specified in order to restrict the IOPS log to one entry per second, otherwise you get one entry per IO. As that is kept in memory until fio completes, unless the number of entries is reduced then due to the sheer number of IOs the SSDs can service the fio process will cause the VM to run out of memory. This directive causes just one averaged entry per 1000ms, which is adequate for graphing.

unlink is used to make sure that data files are deleted after each test run. Leaving the files in place would cause additional promotions into cache, which would favour tests that are run multiple times.

By default when told to do a random read, fio would ensure it reads the entire file once only (before starting again, if the test is time-based). That would not be any use for testing cache warming as there would only ever be one read from each part of the file. To avoid this, a Zipf distribution is used so that read locations are chosen according to a somewhat repetitive distribution, simulating hotspots. I have no real idea about this part so if anyone has any input on that I’d be interested to hear, but my understanding is that a Zipf distribution with exponent 1.2 should cause about 91% of the reads to be from about 10% of the blocks.

In fact I would be very interested to hear of other fio job configurations that people think would be useful here.

If you’re unfamiliar with the tool, fio output can appear quite daunting. There’s a wealth of information in there though, so it’s worth looking into if you ever need to test storage performance. Too many times people just reach for dd or bonnie as performance testing tools but apart from the most trivial of tests they are just too simple. I found this article by Albert Tobey useful in understanding fio‘s output.

Each test run was performed against:

  1. A purely SSD-backed filesystem.
  2. A purely HDD-backed filesystem.
  3. A filesystem on top of a bcache device, sequential_cutoff=4M (default).
  4. A filesystem on top of a bcache device, sequential_cutoff=0.
  5. A filesystem on top of an lvmcache device, empty cache.
  6. A filesystem on top of an lvmcache device, cache full of junk.
Baseline SSD git grep test $ cd /srv/ssd $ /usr/bin/time -f '%U user %S system %E elapsed %P CPU' \ > git clone git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git Cloning into 'linux-stable'... remote: Counting objects: 5948690, done. remote: Compressing objects: 100% (124799/124799), done. remote: Total 5948690 (delta 427756), reused 549517 (delta 425675) Receiving objects: 100% (5948690/5948690), 1.21 GiB | 39.61 MiB/s, done. Resolving deltas: 100% (4961965/4961965), done. Checking out files: 100% (59844/59844), done. 195.45 user 22.10 system 2:37.89 elapsed 137% CPU $ cd /srv/ssd/linux-stable $ for i in {1..20}; do > echo 3 | sudo tee /proc/sys/vm/drop_caches >/dev/null ; \ > /usr/bin/time -f '%U user %S system %E elapsed %P CPU' \ > git --no-pager grep -q blah ; \ > sleep 1; \ > done 0.98 user 1.28 system 0:02.28 elapsed 99% CPU 0.95 user 1.15 system 0:02.18 elapsed 96% CPU 0.89 user 1.30 system 0:02.23 elapsed 98% CPU 1.04 user 1.13 system 0:02.25 elapsed 96% CPU 0.95 user 1.22 system 0:02.24 elapsed 96% CPU 0.95 user 1.23 system 0:02.23 elapsed 97% CPU 1.00 user 1.16 system 0:02.24 elapsed 96% CPU 1.00 user 1.16 system 0:02.23 elapsed 96% CPU 0.89 user 1.28 system 0:02.22 elapsed 97% CPU 0.97 user 1.20 system 0:02.24 elapsed 97% CPU 1.00 user 1.18 system 0:02.23 elapsed 97% CPU 0.96 user 1.21 system 0:02.23 elapsed 97% CPU 0.96 user 1.22 system 0:02.24 elapsed 97% CPU 1.01 user 1.17 system 0:02.24 elapsed 97% CPU 0.94 user 1.22 system 0:02.24 elapsed 96% CPU 1.00 user 1.16 system 0:02.24 elapsed 96% CPU 0.98 user 1.18 system 0:02.24 elapsed 96% CPU 0.92 user 1.24 system 0:02.24 elapsed 97% CPU 0.92 user 1.28 system 0:02.25 elapsed 98% CPU 1.09 user 1.07 system 0:02.23 elapsed 97% CPU

The initial git clone took 157.89 seconds and each git grep reliably took about 2.24 seconds.

fio test $ cd /srv/ssd/fio && FIOTEST=ssd fio ~/partial.fio partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [409.8MB/0KB/0KB /s] [105K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=29306: Wed Jul 19 05:17:39 2017 read : io=379033MB, bw=215628KB/s, iops=53906, runt=1800001msec slat (usec): min=2, max=1547, avg= 4.38, stdev= 1.68 clat (usec): min=26, max=4350, avg=141.54, stdev=29.67 lat (usec): min=94, max=4353, avg=146.39, stdev=29.65 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 108], 10.00th=[ 112], 20.00th=[ 119], | 30.00th=[ 124], 40.00th=[ 129], 50.00th=[ 135], 60.00th=[ 141], | 70.00th=[ 149], 80.00th=[ 161], 90.00th=[ 179], 95.00th=[ 197], | 99.00th=[ 243], 99.50th=[ 262], 99.90th=[ 314], 99.95th=[ 338], | 99.99th=[ 406] lat (usec) : 50=0.01%, 100=0.14%, 250=99.08%, 500=0.77%, 750=0.01% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01% cpu : usr=14.30%, sys=39.70%, ctx=50601996, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=97032448/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=29307: Wed Jul 19 05:17:39 2017 read : io=358188MB, bw=203769KB/s, iops=50942, runt=1800001msec slat (usec): min=2, max=225, avg= 4.46, stdev= 1.71 clat (usec): min=21, max=4385, avg=150.08, stdev=43.00 lat (usec): min=69, max=4391, avg=155.01, stdev=42.98 clat percentiles (usec): | 1.00th=[ 104], 5.00th=[ 110], 10.00th=[ 115], 20.00th=[ 121], | 30.00th=[ 127], 40.00th=[ 133], 50.00th=[ 137], 60.00th=[ 145], | 70.00th=[ 155], 80.00th=[ 171], 90.00th=[ 199], 95.00th=[ 237], | 99.00th=[ 322], 99.50th=[ 350], 99.90th=[ 414], 99.95th=[ 438], | 99.99th=[ 498] lat (usec) : 50=0.01%, 100=0.10%, 250=95.82%, 500=4.06%, 750=0.01% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.01% cpu : usr=13.80%, sys=37.97%, ctx=48695297, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=91696255/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=737221MB, aggrb=419396KB/s, minb=203769KB/s, maxb=215627KB/s, mint=1800001msec, maxt=1800001msec Disk stats (read/write): xvda: ios=188724848/22, merge=0/28, ticks=27221784/0, in_queue=27267336, util=100.00%

This provides a nice illustration of just how fast the SSDs are: each process managed a very consistent 52 thousand IOPS and between the two of them they managed to read ~720GiB of data in 30 minutes.

Baseline HDD git grep test $ cd /srv/slow $ /usr/bin/time -f '%U user %S system %E elapsed %P CPU' \ > git clone git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git Cloning into 'linux-stable'... remote: Counting objects: 5948690, done. remote: Compressing objects: 100% (124799/124799), done. remote: Total 5948690 (delta 427756), reused 549517 (delta 425675) Receiving objects: 100% (5948690/5948690), 1.21 GiB | 38.36 MiB/s, done. Resolving deltas: 100% (4961965/4961965), done. Checking out files: 100% (59844/59844), done. 194.61 user 21.48 system 3:42.23 elapsed 97% CPU $ cd /srv/slow/linux-stable $ for i in {1..20}; do > echo 3 | sudo tee /proc/sys/vm/drop_caches >/dev/null ; \ > /usr/bin/time -f '%U user %S system %E elapsed %P CPU' \ > git --no-pager grep -q blah ; \ > sleep 1; \ > done 1.22 user 1.60 system 0:23.60 elapsed 11% CPU 1.35 user 1.32 system 0:23.08 elapsed 11% CPU 1.24 user 1.43 system 0:22.49 elapsed 11% CPU 1.15 user 1.57 system 0:22.37 elapsed 12% CPU 1.25 user 1.44 system 0:22.09 elapsed 12% CPU 1.26 user 1.48 system 0:22.60 elapsed 12% CPU 1.20 user 1.52 system 0:22.47 elapsed 12% CPU 1.21 user 1.48 system 0:22.82 elapsed 11% CPU 1.14 user 1.55 system 0:22.45 elapsed 12% CPU 1.28 user 1.49 system 0:22.32 elapsed 12% CPU 1.21 user 1.48 system 0:22.43 elapsed 12% CPU 1.26 user 1.51 system 0:22.95 elapsed 12% CPU 1.21 user 1.49 system 0:22.33 elapsed 12% CPU 1.23 user 1.45 system 0:22.54 elapsed 11% CPU 1.19 user 1.51 system 0:22.08 elapsed 12% CPU 1.26 user 1.42 system 0:22.42 elapsed 12% CPU 1.20 user 1.51 system 0:22.47 elapsed 12% CPU 1.19 user 1.53 system 0:22.42 elapsed 12% CPU 1.15 user 1.56 system 0:22.48 elapsed 12% CPU 1.24 user 1.47 system 0:22.10 elapsed 12% CPU

The initial git clone took 222.23 seconds and each git grep reliably took about 22.53 seconds.

Of course, a significant part of the clone time was spent downloading the data.

fio test $ cd /srv/slow/fio && FIOTEST=hdd fio ~/partial.fio partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [3795KB/0KB/0KB /s] [948/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=29158: Wed Jul 19 04:33:19 2017 read : io=3276.3MB, bw=1863.8KB/s, iops=465, runt=1800078msec slat (usec): min=2, max=132, avg=13.00, stdev= 8.10 clat (usec): min=59, max=830656, avg=17151.07, stdev=36630.33 lat (usec): min=64, max=830668, avg=17164.94, stdev=36630.39 clat percentiles (usec): | 1.00th=[ 70], 5.00th=[ 78], 10.00th=[ 88], 20.00th=[ 111], | 30.00th=[ 143], 40.00th=[ 205], 50.00th=[ 988], 60.00th=[ 1496], | 70.00th=[15424], 80.00th=[26240], 90.00th=[50432], 95.00th=[82432], | 99.00th=[177152], 99.50th=[224256], 99.90th=[346112], 99.95th=[403456], | 99.99th=[544768] lat (usec) : 100=15.10%, 250=26.82%, 500=3.79%, 750=1.99%, 1000=2.47% lat (msec) : 2=10.84%, 4=0.39%, 10=2.91%, 20=10.92%, 50=14.68% lat (msec) : 100=6.53%, 250=3.21%, 500=0.34%, 750=0.02%, 1000=0.01% cpu : usr=0.39%, sys=1.17%, ctx=822685, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=838728/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=29159: Wed Jul 19 04:33:19 2017 read : io=3328.4MB, bw=1893.4KB/s, iops=473, runt=1800076msec slat (usec): min=2, max=157, avg=13.20, stdev= 8.22 clat (usec): min=4, max=1046.5K, avg=16882.45, stdev=35737.35 lat (usec): min=64, max=1046.6K, avg=16896.52, stdev=35737.42 clat percentiles (usec): | 1.00th=[ 70], 5.00th=[ 78], 10.00th=[ 88], 20.00th=[ 111], | 30.00th=[ 143], 40.00th=[ 211], 50.00th=[ 996], 60.00th=[ 1528], | 70.00th=[15296], 80.00th=[25728], 90.00th=[49920], 95.00th=[81408], | 99.00th=[173056], 99.50th=[218112], 99.90th=[329728], 99.95th=[391168], | 99.99th=[522240] lat (usec) : 10=0.01%, 50=0.01%, 100=15.03%, 250=26.59%, 500=3.86% lat (usec) : 750=2.07%, 1000=2.49% lat (msec) : 2=10.70%, 4=0.36%, 10=3.03%, 20=11.45%, 50=14.46% lat (msec) : 100=6.49%, 250=3.17%, 500=0.30%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.47%, sys=1.15%, ctx=834258, majf=0, minf=14 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=852044/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=6604.6MB, aggrb=3757KB/s, minb=1863KB/s, maxb=1893KB/s, mint=1800076msec, maxt=1800078msec Disk stats (read/write): xvde: ios=1690665/12, merge=0/15, ticks=28764332/2848, in_queue=28767992, util=100.00%

Both jobs together managed an average of 938 IOPS here, which is pretty weak but still considerably more than a pair of HDDs should be able to do (100–150 IOPS each, so 300 if you were amazingly lucky). This indicates that the drive cache must have been involved, and indeed the IOPS rate does very rapidly increase within the first few seconds of the test run but reaches a stable maximum within 20 seconds. That marks the time when the disk cache is full and cache hit rate can’t be maintained.

A total of 6,604.6MiB was read during the 30 minute test.

bcache Setup

By default bcache uses writethrough cache mode, where writes go to both the backing store and the cache and are only acknowledged once they’ve completed at the backing store. In this mode bcache will never accelerate writes, but it is safe to lose the cache device at any time.

I felt comfortable switching the bcache device to writeback mode as the underlying storage used for the cache is already reliable. In writeback mode the writes are acknowledged as soon as they are completed on the cache device and are later sequentially written to backing device.

This should result in higher write performance as well as the higher read performance we’d expect from using a cache.

$ sudo make-bcache -B /dev/xvdd UUID: cf0353eb-ac08-45bd-a3d5-f8b92fbc0fb8 Set UUID: 8c06d317-e735-41f5-8dec-52da26a37883 version: 1 block_size: 1 data_offset: 16 $ sudo mkfs.ext4 -L cached /dev/bcache0 mke2fs 1.43.4 (31-Jan-2017) Discarding device blocks: done Creating filesystem with 13107198 4k blocks and 3276800 inodes Filesystem UUID: 7db8568e-282b-4fa2-bf19-a42143708a0d Superblock backups stored on blocks: 32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208, 4096000, 7962624, 11239424 Allocating group tables: done Writing inode tables: done Creating journal (65536 blocks): done Writing superblocks and filesystem accounting information: done $ sudo make-bcache -C /dev/xvdc UUID: ae581052-4449-460d-8cf5-b11d6467403d Set UUID: 2fcc7937-9f5f-4089-9400-552675929ea5 version: 0 nbuckets: 10240 block_size: 1 bucket_size: 1024 nr_in_set: 1 nr_this_dev: 0 first_bucket: 1 $ sudo bcache-super-show /dev/xvdc sb.magic ok sb.first_sector 8 [match] sb.csum 206ADF979437131E [match] sb.version 3 [cache device] dev.label (empty) dev.uuid ae581052-4449-460d-8cf5-b11d6467403d dev.sectors_per_block 1 dev.sectors_per_bucket 1024 dev.cache.first_sector 1024 dev.cache.cache_sectors 10484736 dev.cache.total_sectors 10485760 dev.cache.ordered yes dev.cache.discard no dev.cache.pos 0 dev.cache.replacement 0 [lru] cset.uuid 2fcc7937-9f5f-4089-9400-552675929ea5 $ echo 2fcc7937-9f5f-4089-9400-552675929ea5 | sudo tee /sys/block/bcache0/bcache/attach >/dev/null $ echo writeback | sudo tee /sys/block/bcache0/bcache/cache_mode >/dev/null $ sudo mount /dev/bcache0 /srv/cache $ sudo chown andy: /srv/cache git grep test $ cd /srv/cache $ /usr/bin/time -f '%U user %S system %E elapsed %P CPU' \ > git clone git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git Cloning into 'linux-stable'... remote: Counting objects: 5948690, done. remote: Compressing objects: 100% (124799/124799), done. remote: Total 5948690 (delta 427755), reused 549518 (delta 425675) Receiving objects: 100% (5948690/5948690), 1.21 GiB | 40.30 MiB/s, done. Resolving deltas: 100% (4961964/4961964), done. Checking out files: 100% (59844/59844), done. 194.88 user 21.30 system 3:34.10 elapsed 100% CPU $ cd /srv/cache/linux-stable $ for i in {1..20}; do > echo 3 | sudo tee /proc/sys/vm/drop_caches >/dev/null ; \ > /usr/bin/time -f '%U user %S system %E elapsed %P CPU' \ > git --no-pager grep -q blah ; \ > sleep 1; \ > done 1.03 user 1.40 system 0:03.02 elapsed 80% CPU 1.01 user 1.25 system 0:02.55 elapsed 88% CPU 1.05 user 1.21 system 0:02.63 elapsed 85% CPU 0.90 user 1.34 system 0:02.56 elapsed 87% CPU 0.96 user 1.28 system 0:02.59 elapsed 86% CPU 0.90 user 1.35 system 0:02.54 elapsed 88% CPU 1.04 user 1.22 system 0:02.51 elapsed 90% CPU 1.00 user 1.25 system 0:02.50 elapsed 89% CPU 0.93 user 1.34 system 0:02.54 elapsed 89% CPU 1.02 user 1.26 system 0:02.53 elapsed 90% CPU 0.91 user 1.43 system 0:02.67 elapsed 88% CPU 0.94 user 1.39 system 0:02.57 elapsed 90% CPU 1.04 user 1.28 system 0:02.57 elapsed 90% CPU 1.00 user 1.37 system 0:02.58 elapsed 91% CPU 1.02 user 1.33 system 0:02.84 elapsed 82% CPU 1.11 user 1.21 system 0:02.58 elapsed 90% CPU 1.03 user 1.30 system 0:02.57 elapsed 90% CPU 0.96 user 1.37 system 0:02.58 elapsed 90% CPU 1.02 user 1.36 system 0:02.62 elapsed 90% CPU 0.92 user 1.41 system 0:02.62 elapsed 89% CPU

The initial git clone took 214.10 seconds and each git grep on average took about 2.61 seconds. It can be seen that the first run is slightly slower but very quickly things settle down.

The git grep results are very encouraging being only 11.6% of the time taken for the baseline HDD tests, or almost 9x faster. Each run was only 0.37 seconds slower than when run directly against SSD.

The initial clone results are disappointing though, being only 8.13 seconds faster than when run against baseline HDD. It looks like most data must have bypassed the cache and waited for HDD. bcache does bypass cache for sequential writes which is what the git clone would have done, and we can prove this:

$ cat /sys/block/bcache0/bcache/stats_total/bypassed 270M

So yes, there we go, all 270MiB bypassed the cache and waited on HDD completion.

We can give bcache hints about when to consider writes sequential, and at the extreme end tell it to never consider writes sequential, so let’s try that.

$ echo 0 | sudo tee /sys/block/bcache0/bcache/sequential_cutoff >/dev/null $ cd /srv/cache $ rm -r linux-stable $ /usr/bin/time -f '%U user %S system %E elapsed %P CPU' \ > git clone git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git Cloning into 'linux-stable'... remote: Counting objects: 5948690, done. remote: Compressing objects: 100% (124799/124799), done. remote: Total 5948690 (delta 427756), reused 549517 (delta 425675) Receiving objects: 100% (5948690/5948690), 1.21 GiB | 39.51 MiB/s, done. Resolving deltas: 100% (4961965/4961965), done. Checking out files: 100% (59844/59844), done. 195.10 user 23.53 system 2:39.87 elapsed 136% CPU $ cat /sys/block/bcache0/bcache/stats_total/bypassed 299M

This time only 29MiB of data bypassed cache, and the time for clone was 159.87 seconds. That’s 62.36 seconds faster than against HDD, and only 1.98 seconds slower than against SSD!

fio test – sequential_cutoff=4M $ cd /srv/cache/fio && FIOTEST=bcache-sc4m fio ~/partial.fio partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [119.9MB/0KB/0KB /s] [30.7K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=8640: Thu Jul 20 18:13:08 2017 read : io=45237MB, bw=25734KB/s, iops=6433, runt=1800041msec slat (usec): min=3, max=5707, avg= 6.14, stdev=10.97 clat (usec): min=2, max=1238.4K, avg=1235.14, stdev=10675.04 lat (usec): min=70, max=1238.5K, avg=1241.70, stdev=10675.36 clat percentiles (usec): | 1.00th=[ 101], 5.00th=[ 104], 10.00th=[ 106], 20.00th=[ 108], | 30.00th=[ 111], 40.00th=[ 117], 50.00th=[ 119], 60.00th=[ 121], | 70.00th=[ 122], 80.00th=[ 124], 90.00th=[ 133], 95.00th=[ 151], | 99.00th=[35584], 99.50th=[67072], 99.90th=[148480], 99.95th=[185344], | 99.99th=[305152] lat (usec) : 4=0.01%, 50=0.01%, 100=0.37%, 250=97.11%, 500=0.08% lat (usec) : 750=0.02%, 1000=0.02% lat (msec) : 2=0.03%, 4=0.01%, 10=0.13%, 20=0.65%, 50=0.87% lat (msec) : 100=0.46%, 250=0.24%, 500=0.01%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=1.81%, sys=6.79%, ctx=10722687, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=11580733/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=8641: Thu Jul 20 18:13:08 2017 read : io=43743MB, bw=24885KB/s, iops=6221, runt=1800037msec slat (usec): min=3, max=5703, avg= 6.27, stdev=10.53 clat (usec): min=2, max=1081.7K, avg=1277.44, stdev=10884.69 lat (usec): min=70, max=1081.8K, avg=1284.13, stdev=10884.99 clat percentiles (usec): | 1.00th=[ 100], 5.00th=[ 103], 10.00th=[ 104], 20.00th=[ 106], | 30.00th=[ 107], 40.00th=[ 109], 50.00th=[ 111], 60.00th=[ 117], | 70.00th=[ 120], 80.00th=[ 122], 90.00th=[ 129], 95.00th=[ 145], | 99.00th=[37632], 99.50th=[69120], 99.90th=[150528], 99.95th=[189440], | 99.99th=[317440] lat (usec) : 4=0.01%, 20=0.01%, 50=0.01%, 100=0.68%, 250=96.73% lat (usec) : 500=0.07%, 750=0.02%, 1000=0.02% lat (msec) : 2=0.03%, 4=0.01%, 10=0.13%, 20=0.68%, 50=0.90% lat (msec) : 100=0.48%, 250=0.25%, 500=0.02%, 750=0.01%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=1.74%, sys=6.57%, ctx=10425775, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=11198290/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=88981MB, aggrb=50618KB/s, minb=24884KB/s, maxb=25734KB/s, mint=1800037msec, maxt=1800041msec Disk stats (read/write): bcache0: ios=22775854/880, merge=0/0, ticks=28668080/36768, in_queue=197531580, util=100.00%, aggrios=11393419/289009, aggrmerge=0/182, aggrticks=14329508/435284, aggrin_queue=14764868, aggrutil=100.00% xvdc: ios=22230796/570128, merge=0/348, ticks=2656456/37312, in_queue=2693812, util=67.39% xvdd: ios=556042/7890, merge=0/17, ticks=26002560/833256, in_queue=26835924, util=100.00%

The ~15k IOPS per process achieved towards the end of this test are quite impressive and suggest that there is more scope for performance increase in a longer test. The combined average IOPS across 30 minutes were “only” 12.7k though. Vastly superior to a HDD and we would be glad to take it.

22,779,023 IOs were issued (11,580,733+11,198,290) of which 22,230,796 (97.6%) hit the cache. That’s a really great result, but we learned earlier that setting sequential_cutoff to 0 allows us to get data we write straight into the cache, so let’s see how that changes things.

fio test – sequential_cutoff=0 $ cd /srv/cache/fio && FIOTEST=bcache-sc0 fio ~/partial.fio partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [276.1MB/0KB/0KB /s] [70.1K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=9190: Thu Jul 20 20:34:49 2017 read : io=367029MB, bw=208799KB/s, iops=52199, runt=1800001msec slat (usec): min=3, max=350673, avg= 5.49, stdev=54.90 clat (usec): min=22, max=558335, avg=145.49, stdev=593.01 lat (usec): min=67, max=558342, avg=151.41, stdev=597.19 clat percentiles (usec): | 1.00th=[ 102], 5.00th=[ 106], 10.00th=[ 109], 20.00th=[ 116], | 30.00th=[ 120], 40.00th=[ 123], 50.00th=[ 127], 60.00th=[ 133], | 70.00th=[ 139], 80.00th=[ 153], 90.00th=[ 185], 95.00th=[ 223], | 99.00th=[ 310], 99.50th=[ 338], 99.90th=[ 442], 99.95th=[ 868], | 99.99th=[20352] lat (usec) : 50=0.01%, 100=0.27%, 250=96.55%, 500=3.11%, 750=0.02% lat (usec) : 1000=0.01% lat (msec) : 2=0.02%, 4=0.01%, 10=0.01%, 20=0.01%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01%, 500=0.01%, 750=0.01% cpu : usr=13.23%, sys=41.79%, ctx=46793234, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=93959531/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=9191: Thu Jul 20 20:34:49 2017 read : io=116645MB, bw=66357KB/s, iops=16589, runt=1800020msec slat (usec): min=3, max=86150, avg= 6.31, stdev=22.07 clat (usec): min=16, max=581877, avg=473.49, stdev=3466.14 lat (usec): min=67, max=581884, avg=480.26, stdev=3466.62 clat percentiles (usec): | 1.00th=[ 102], 5.00th=[ 107], 10.00th=[ 110], 20.00th=[ 116], | 30.00th=[ 120], 40.00th=[ 123], 50.00th=[ 126], 60.00th=[ 131], | 70.00th=[ 135], 80.00th=[ 143], 90.00th=[ 161], 95.00th=[ 183], | 99.00th=[14656], 99.50th=[23680], 99.90th=[47360], 99.95th=[59648], | 99.99th=[94720] lat (usec) : 20=0.01%, 50=0.01%, 100=0.23%, 250=97.81%, 500=0.37% lat (usec) : 750=0.03%, 1000=0.02% lat (msec) : 2=0.02%, 4=0.02%, 10=0.18%, 20=0.66%, 50=0.58% lat (msec) : 100=0.08%, 250=0.01%, 500=0.01%, 750=0.01% cpu : usr=4.86%, sys=16.81%, ctx=24164179, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=29861123/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=483674MB, aggrb=275153KB/s, minb=66357KB/s, maxb=208798KB/s, mint=1800001msec, maxt=1800020msec Disk stats (read/write): bcache0: ios=123818417/28, merge=0/0, ticks=27945560/5768, in_queue=429639892, util=100.00%, aggrios=61919282/259095, aggrmerge=0/25, aggrticks=13927904/125270, aggrin_queue=14059068, aggrutil=100.00% xvdc: ios=123345393/500268, merge=0/35, ticks=17064408/30904, in_queue=17107112, util=99.99% xvdd: ios=493172/17923, merge=0/15, ticks=10791400/219636, in_queue=11011024, util=100.00%

123,820,654 (93,959,531+29,861,123) IOs issued, 123,345,393 (99.62%) hit cache. It’s just ~2% higher than the previous test, but what a difference it makes. The IOPS log indicates that performance very quickly increases to near-SSD levels, averaging 68.8k IOPS across 30 minutes.

Interestingly unlike all other tests (even unlike bcache with the default sequential_cutoff), the second job doesn’t manage anywhere near the same level of performance as the first job. It’s 52.2k IOPS vs. 16.6k IOPS. Perhaps at this high level of cache utilisation there is some sort of locking going on. Might be worth following up with upstream.

lvmcache Setup $ sudo pvcreate /dev/xvdc Physical volume "/dev/xvdc" successfully created. $ sudo pvcreate /dev/xvdd Physical volume "/dev/xvdd" successfully created. $ sudo vgcreate cachedvg /dev/xvd{d,c} Volume group "cachedvg" successfully created $ sudo lvcreate -L50m -n lv_cache_meta cachedvg /dev/xvdc Rounding up size to full physical extent 52.00 MiB Logical volume "lv_cache_meta" created. $ sudo lvcreate -L4g -n lv_cache cachedvg /dev/xvdc Logical volume "lv_cache" created. $ sudo lvcreate -l100%PVS -n lv cachedvg /dev/xvdd Logical volume "lv" created. $ sudo lvs LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert lv cachedvg -wi-a----- 50.00g lv_cache cachedvg -wi-a----- 4.00g lv_cache_meta cachedvg -wi-a----- 52.00m $ sudo lvconvert --type cache-pool --cachemode writeback --poolmetadata cachedvg/lv_cache_meta cachedvg/lv_cache Converted cachedvg/lv_cache to cache pool. $ sudo lvconvert --type cache --cachepool cachedvg/lv_cache cachedvg/lv Logical volume cachedvg/lv is now cached. $ sudo lvs -a -o +devices LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert Devices lv cachedvg Cwi-a-C--- 50.00g [lv_cache] [lv_corig] 0.00 1.01 0.00 lv_corig(0) [lv_cache] cachedvg Cwi---C--- 4.00g 0.00 1.01 0.00 lv_cache_cdata(0) [lv_cache_cdata] cachedvg Cwi-ao---- 4.00g /dev/xvdc(13) [lv_cache_cmeta] cachedvg ewi-ao---- 52.00m /dev/xvdc(0) [lv_corig] cachedvg owi-aoC--- 50.00g /dev/xvdd(0) [lvol0_pmspare] cachedvg ewi------- 52.00m /dev/xvdc(1037) $ sudo mkfs.ext4 -L cached /dev/cachedvg/lv mke2fs 1.43.4 (31-Jan-2017) Creating filesystem with 13106176 4k blocks and 3276800 inodes Filesystem UUID: bff28530-339a-403d-889d-dd42f4465ab5 Superblock backups stored on blocks: 32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208, 4096000, 7962624, 11239424 Allocating group tables: done Writing inode tables: done Creating journal (65536 blocks): done Writing superblocks and filesystem accounting information: done $ sudo mount /dev/cachedvg/lv /srv/cache $ sudo chown andy: /srv/cache git grep test $ cd /srv/cache $ /usr/bin/time -f '%U user %S system %E elapsed %P CPU' \ > git clone git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git Cloning into 'linux-stable'... remote: Counting objects: 5948690, done. remote: Compressing objects: 100% (124799/124799), done. remote: Total 5948690 (delta 427755), reused 549518 (delta 425675) Receiving objects: 100% (5948690/5948690), 1.21 GiB | 35.05 MiB/s, done. Resolving deltas: 100% (4961964/4961964), done. Checking out files: 100% (59844/59844), done. 196.00 user 23.40 system 2:43.13 elapsed 134% CPU $ cd /srv/cache/linux-stable $ for i in {1..20}; do > echo 3 | sudo tee /proc/sys/vm/drop_caches >/dev/null ; \ > /usr/bin/time -f '%U user %S system %E elapsed %P CPU' \ > git --no-pager grep -q blah ; \ > sleep 1; \ > done 1.00 user 1.58 system 0:06.41 elapsed 40% CPU 0.93 user 1.46 system 0:04.22 elapsed 56% CPU 1.09 user 1.30 system 0:04.09 elapsed 58% CPU 1.02 user 1.39 system 0:03.89 elapsed 62% CPU 0.97 user 1.41 system 0:03.89 elapsed 61% CPU 1.03 user 1.34 system 0:03.52 elapsed 67% CPU 1.06 user 1.30 system 0:03.60 elapsed 65% CPU 0.94 user 1.46 system 0:03.79 elapsed 63% CPU 1.05 user 1.33 system 0:03.28 elapsed 72% CPU 1.00 user 1.39 system 0:03.81 elapsed 62% CPU 1.03 user 1.37 system 0:03.25 elapsed 74% CPU 1.05 user 1.37 system 0:03.30 elapsed 73% CPU 1.07 user 1.30 system 0:03.18 elapsed 74% CPU 1.07 user 1.32 system 0:03.28 elapsed 72% CPU 0.96 user 1.46 system 0:02.95 elapsed 82% CPU 1.07 user 1.31 system 0:03.14 elapsed 76% CPU 0.98 user 1.40 system 0:03.04 elapsed 78% CPU 0.96 user 1.42 system 0:03.19 elapsed 74% CPU 1.00 user 1.40 system 0:03.12 elapsed 77% CPU 0.89 user 1.48 system 0:02.89 elapsed 82% CPU

The initial git clone took 163.13 seconds, not far off native SSD timings (157.68 seconds). Each git grep on average took about 3.35 seconds. These started off slower (6.41 seconds) but settled down with the last 15 runs averaging at 3.29 seconds.

fio test – empty cache $ cd /srv/cache/fio && FIOTEST=lvmcache-empty fio ~/partial.fio partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [373.5MB/0KB/0KB /s] [95.7K/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=6652: Thu Jul 20 12:14:06 2017 read : io=333831MB, bw=189912KB/s, iops=47478, runt=1800001msec slat (usec): min=2, max=245, avg= 6.41, stdev= 2.17 clat (usec): min=29, max=234093, avg=159.69, stdev=401.74 lat (usec): min=69, max=234098, avg=166.57, stdev=401.75 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 108], 10.00th=[ 113], 20.00th=[ 119], | 30.00th=[ 124], 40.00th=[ 129], 50.00th=[ 135], 60.00th=[ 143], | 70.00th=[ 153], 80.00th=[ 175], 90.00th=[ 221], 95.00th=[ 274], | 99.00th=[ 370], 99.50th=[ 410], 99.90th=[ 502], 99.95th=[ 996], | 99.99th=[16768] lat (usec) : 50=0.01%, 100=0.16%, 250=92.76%, 500=6.97%, 750=0.05% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.01%, 10=0.02%, 20=0.02%, 50=0.01% lat (msec) : 100=0.01%, 250=0.01% cpu : usr=12.04%, sys=44.39%, ctx=42073879, majf=0, minf=16 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=85460611/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=6653: Thu Jul 20 12:14:06 2017 read : io=313896MB, bw=178572KB/s, iops=44642, runt=1800001msec slat (usec): min=2, max=335, avg= 6.55, stdev= 2.25 clat (usec): min=22, max=212646, avg=170.22, stdev=721.29 lat (usec): min=63, max=212656, avg=177.25, stdev=721.32 clat percentiles (usec): | 1.00th=[ 103], 5.00th=[ 108], 10.00th=[ 112], 20.00th=[ 118], | 30.00th=[ 122], 40.00th=[ 127], 50.00th=[ 131], 60.00th=[ 137], | 70.00th=[ 145], 80.00th=[ 157], 90.00th=[ 177], 95.00th=[ 197], | 99.00th=[ 258], 99.50th=[ 314], 99.90th=[12096], 99.95th=[15424], | 99.99th=[28032] lat (usec) : 50=0.01%, 100=0.22%, 250=98.57%, 500=0.91%, 750=0.01% lat (usec) : 1000=0.01% lat (msec) : 2=0.01%, 4=0.02%, 10=0.11%, 20=0.11%, 50=0.03% lat (msec) : 100=0.01%, 250=0.01% cpu : usr=11.36%, sys=42.77%, ctx=41146564, majf=0, minf=14 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=80357365/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=647726MB, aggrb=368484KB/s, minb=178571KB/s, maxb=189912KB/s, mint=1800001msec, maxt=1800001msec Disk stats (read/write): dm-2: ios=165815070/5, merge=0/0, ticks=27172896/564, in_queue=27297688, util=100.00%, aggrios=55273566/7259, aggrmerge=0/0, aggrticks=8986252/1498, aggrin_queue=9007888, aggrutil=100.00% dm-1: ios=2/19047, merge=0/0, ticks=0/3072, in_queue=3072, util=0.14%, aggrios=165379217/17690, aggrmerge=0/4083, aggrticks=23890596/3616, aggrin_queue=23926592, aggrutil=100.00% xvdc: ios=165379217/17690, merge=0/4083, ticks=23890596/3616, in_queue=23926592, util=100.00% dm-0: ios=165379215/2726, merge=0/0, ticks=23917816/860, in_queue=23979056, util=100.00% dm-3: ios=441483/6, merge=0/0, ticks=3040940/564, in_queue=3041536, util=75.31%, aggrios=441483/5, aggrmerge=0/1, aggrticks=3040896/464, aggrin_queue=3041312, aggrutil=75.30% xvdd: ios=441483/5, merge=0/1, ticks=3040896/464, in_queue=3041312, util=75.30%

We can see that the first job issued 85,460,611 IOs and the second job issued 80,357,365 IOs, for a total of 165817976 IOs. Of these, 165,379,217 (99.74%) came from xvdc, the cache (SSD) device. Only 441,483 (0.27%) came from xvdd, the backing (HDD) device. Again this is very aggressive cache population, with a combined average of 92k IOPS being achieved over 30 minutes.

80k IOPS are achieved within 20 seconds of the start of testing, which is probably helped by the disk cache, but the steady maximum of ~90k IOPS is still achieved by 40 seconds in. Average IOPS ended up at 88% that of the SSD.

This is incredible performance, but I could only get it to happen the first time, against an empty cache device. Since the next test run will unlink the previous run’s files, the cache device will at that point be full of data that is of no use as it’s been deleted. The next test will write out 4GiB of new, uncached data.

fio test – full cache $ cd /srv/cache/fio && FIOTEST=lvmcache-full fio ~/partial.fio partial: (g=0): rw=randread, bs=4K-4K/4K-4K/4K-4K, ioengine=libaio, iodepth=8 ... fio-2.16 Starting 2 processes partial: Laying out IO file(s) (1 file(s) / 2048MB) partial: Laying out IO file(s) (1 file(s) / 2048MB) Jobs: 2 (f=2): [r(2)] [100.0% done] [6072KB/0KB/0KB /s] [1518/0/0 iops] [eta 00m:00s] partial: (groupid=0, jobs=1): err= 0: pid=6857: Thu Jul 20 12:58:00 2017 read : io=5266.1MB, bw=2996.2KB/s, iops=749, runt=1800111msec slat (usec): min=3, max=117, avg=12.93, stdev= 7.50 clat (usec): min=58, max=1060.8K, avg=10663.38, stdev=31025.49 lat (usec): min=73, max=1060.8K, avg=10676.99, stdev=31025.92 clat percentiles (usec): | 1.00th=[ 105], 5.00th=[ 112], 10.00th=[ 117], 20.00th=[ 125], | 30.00th=[ 127], 40.00th=[ 133], 50.00th=[ 137], 60.00th=[ 153], | 70.00th=[ 179], 80.00th=[14144], 90.00th=[33024], 95.00th=[59136], | 99.00th=[144384], 99.50th=[191488], 99.90th=[337920], 99.95th=[411648], | 99.99th=[618496] lat (usec) : 100=0.29%, 250=74.34%, 500=0.46%, 750=0.20%, 1000=0.21% lat (msec) : 2=0.47%, 4=0.03%, 10=1.30%, 20=7.26%, 50=9.23% lat (msec) : 100=4.09%, 250=1.88%, 500=0.22%, 750=0.02%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.48%, sys=1.52%, ctx=1335232, majf=0, minf=17 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=1348336/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 partial: (groupid=0, jobs=1): err= 0: pid=6858: Thu Jul 20 12:58:00 2017 read : io=5171.2MB, bw=2941.7KB/s, iops=735, runt=1800133msec slat (usec): min=3, max=320, avg=13.05, stdev= 7.56 clat (usec): min=63, max=1145.3K, avg=10861.20, stdev=30827.62 lat (usec): min=72, max=1145.3K, avg=10874.93, stdev=30828.07 clat percentiles (usec): | 1.00th=[ 109], 5.00th=[ 116], 10.00th=[ 120], 20.00th=[ 126], | 30.00th=[ 131], 40.00th=[ 133], 50.00th=[ 141], 60.00th=[ 155], | 70.00th=[ 185], 80.00th=[14912], 90.00th=[33536], 95.00th=[59136], | 99.00th=[142336], 99.50th=[189440], 99.90th=[333824], 99.95th=[411648], | 99.99th=[602112] lat (usec) : 100=0.33%, 250=73.29%, 500=0.50%, 750=0.22%, 1000=0.22% lat (msec) : 2=0.52%, 4=0.03%, 10=1.38%, 20=7.58%, 50=9.63% lat (msec) : 100=4.18%, 250=1.88%, 500=0.21%, 750=0.02%, 1000=0.01% lat (msec) : 2000=0.01% cpu : usr=0.43%, sys=1.55%, ctx=1310390, majf=0, minf=15 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=100.0%, 16=0.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.1%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% issued : total=r=1323812/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0 latency : target=0, window=0, percentile=100.00%, depth=8 Run status group 0 (all jobs): READ: io=10438MB, aggrb=5937KB/s, minb=2941KB/s, maxb=2996KB/s, mint=1800111msec, maxt=1800133msec Disk stats (read/write): dm-2: ios=2671993/27, merge=0/0, ticks=28764200/8380, in_queue=28773988, util=100.00%, aggrios=891886/15335, aggrmerge=0/0, aggrticks=9387100/6312, aggrin_queue=9393456, aggrutil=100.00% dm-1: ios=173/42639, merge=0/0, ticks=16/9316, in_queue=9336, util=0.24%, aggrios=1967684/42191, aggrmerge=0/3788, aggrticks=291076/9440, aggrin_queue=300476, aggrutil=15.68% xvdc: ios=1967684/42191, merge=0/3788, ticks=291076/9440, in_queue=300476, util=15.68% dm-0: ios=1967511/3340, merge=0/0, ticks=291184/1240, in_queue=292496, util=15.55% dm-3: ios=707976/28, merge=0/0, ticks=27870100/8380, in_queue=27878536, util=100.00%, aggrios=707976/12, aggrmerge=0/16, aggrticks=27870028/2908, aggrin_queue=27872968, aggrutil=100.00% xvdd: ios=707976/12, merge=0/16, ticks=27870028/2908, in_queue=27872968, util=100.00%

Testing with an already-full cache produces dramatically different results. We can see that a total of 2,672,148 (1,348,336+1,323,812) IOs were issued, with 1,967,684 (73.64%) coming from the cache. 707,976 (26.49%) came from the backing device. Although on the face of it that sounds like a good hit rate, it was poor enough to severely limit the average IOPS to 1484 across both jobs. We saw how much the difference was between 99.6% and 97.6% cache hit rate, and now we see what the difference is between ~99% and ~73%.

1,484 IOPS represents a 158% increase on the baseline HDD IOPS of 938 so it’s nothing to be sniffed at, and as I say it may be the case that it only gets better over much longer time span.

Even after this 30 minute test the performance of the lvmcache was still very slowly going up, so I decided to run another test across 3 hours instead of 30 seconds. The performance did continue to increase very very slowly peaking at about 1,640 aggregate IOPS after 3 hours. It seems like it is quite difficult to evict data from the cache in favour of new data that is being read, whereas when the cache was empty it promoted everything that was read immediately. Possibly if this was left for long enough there would eventually be enough cache hits to achieve similar performance to the first run.

bcache did not suffer from this problem: every test run against bcache was the same, except for the effect of preloading the cache device by messing with sequential_cutoff. I suspect that this is because bcache is a write cache; when set to writethrough or writeback mode, writes are going through the cache and immediately being cached. lvmcache is not a write cache; it will only promote things when they are read.

It seems like lvmcache needs a very large number of reads before it will evict data from the cache and bring new data in.

Further tests?

If you have any further tests you’d like to suggest please leave a comment or a tweet. I think it would be good to come up with more fio tests for this, but bear in mind that it really needs to test random read IO with a repetitive element so that a cache can actually be effective. If you read every part of 100GiB equally as often then there will be no hot spots and a comparatively small cache will not be effective, no matter what technology is used.

Also real-world benchmarks would be good, so perhaps a really complex SQL SELECT on a data set that won’t fit in (1GiB) memory but will fit in (5GiB) SSD cache.

It might be interesting to use a cache device much smaller than the working set. These tests laid out two 2GiB files then used a Zipf distribution to randomly read 10% of each (so two sets of ~205MiB data) while the cache devices were 4 or 5GiB in size. It is difficult to guess what the hot spots would be on a production file system, but a 90/10 or 80/20 distribution does seem reasonable.

What may be less reasonable is starting off with such a small data set. What would it be like to try to cache 1TiB with 10GiB of SSD? I’d rather not write out 1TiB of data to check, but I could simulate by writing out 51GiB (5%) and having a cache device that is only 512MiB. On a 90/10 distribution it would still want to cache ~5.2GiB and only have 512MiB to put it in.

Conclusions

SSDs are lovely but they’re still really expensive. Although they are fun for turning out impressive benchmarks, in the real world you only need things to be fast enough. If you want consistently great performance and you have the money, buy some decent SSDs. If you don’t have the money then there is a whole world of trade-offs.

bcache from what I can measure provides really impressive performance but I am concerned about relying upon it. The existing implementation is no longer the focus of the main developer’s attention as the developer has now moved on to bcachefs, a complete filesystem in its own right that builds upon the bcache concept.

I’m really conservative when it comes to filesystems and I know it can take years, even decades, to get to feature parity with all the nasty bugs hammered out. I also worry about continued maintenance of a separate filesystem.

lvmcache on the other hand is providing only a very modest performance boost. In many cases it won’t be worth it. I do like its simplicity though and I’m already very familiar with LVM.

The way that any logical volume can have a cache attached without needing to start over with the filesystem makes lvmcache very flexible; it also means that some volumes can be cached and others not. Furthermore with a slow PV and a fast PV both in the volume group, if necessary an individual logical volume can be allocated entirely from the fast PV.

I may experiment with bcache some more in situations where I can quickly restore the data in event of calamity, or fairly easily move to a different solution if bcache becomes abandonware. I may experiment with lvmcache in places where a modest speedup would be welcome.

Categories: LUG Community Blogs

Steve Kemp: bind9 considered harmful

Mon, 10/07/2017 - 22:00

Recently there was another bind9 security update released by the Debian Security Team. I thought that was odd, so I've scanned my mailbox:

  • 11 January 2017
    • DSA-3758 - bind9
  • 26 February 2017
    • DSA-3795-1 - bind9
  • 14 May 2017
    • DSA-3854-1 - bind9
  • 8 July 2017
    • DSA-3904-1 - bind9

So in the year to date there have been 7 months, in 3 of them nothing happened, but in 4 of them we had bind9 updates. If these trends continue we'll have another 2.5 updates before the end of the year.

I don't run a nameserver. The only reason I have bind-packages on my system is for the dig utility.

Rewriting a compatible version of dig in Perl should be trivial, thanks to the Net::DNS::Resolver module:

These are about the only commands I ever run:

dig -t a steve.fi +short dig -t aaaa steve.fi +short dig -t a steve.fi @8.8.8.8

I should do that then. Yes.

Categories: LUG Community Blogs