Linux-BTRFS Archive on lore.kernel.org
 help / color / Atom feed
From: Dennis Zhou <dennis@kernel.org>
To: dsterba@suse.cz, Dennis Zhou <dennis@kernel.org>,
	David Sterba <dsterba@suse.com>, Chris Mason <clm@fb.com>,
	Josef Bacik <josef@toxicpanda.com>,
	Omar Sandoval <osandov@osandov.com>,
	kernel-team@fb.com, linux-btrfs@vger.kernel.org
Subject: Re: [PATCH v6 00/22] btrfs: async discard support
Date: Thu, 19 Dec 2019 15:17:55 -0600
Message-ID: <20191219211755.GA38803@dennisz-mbp.dhcp.thefacebook.com> (raw)
In-Reply-To: <20191219203438.GS3929@twin.jikos.cz>

On Thu, Dec 19, 2019 at 09:34:38PM +0100, David Sterba wrote:
> On Tue, Dec 17, 2019 at 07:06:00PM -0500, Dennis Zhou wrote:
> > > Regarding the slow io submission, I tried to increase the iops value,
> > > default was 10, but 100 and 1000 made no change. Increasing the maximum
> > > discard request size to 128M worked (when there was such long extent
> > > ready). I was expecting a burst of like 4 consecutive IOs after a 600MB
> > > file is deleted.  I did not try to tweak bps_limit because there was
> > > nothing to limit.
> > 
> > Ah so there's actually a max time between discards set to 10 seconds as
> > the maximum timeout is calculated over 6 hours. So if we only have 6
> > extents, we'd discard 1 per hour(ish given it decays), but this is
> > clamped to 10 seconds.
> > 
> > At least on our servers, we seem to discard at a reasonable rate to
> > prevent performance penalties during a large number of reads and writes
> > while maintaining reasonable write amplification performance. Also,
> > metadata blocks aren't tracked, so on freeing of a whole metadata block
> > group (minus relocation), we'll trickle discards slightly slower than
> > expected.
> 
> So after watching the sysfs numbers, my observation is that the overall
> strategy of the async discard is to wait for larger ranges and discard
> one range every 10 seconds. This is a slow process, but this makes sense
> when there are reads or writes going on so the discard IO penalty is
> marginal.
> 

Yeah, (un)fortunately on our systems we're running chef fairly
frequently which results in a lot of IO in addition to package
deployment. This actually drives the system to have a fairly high steady
state number of untrimmed extents and results in a bit faster paced
discarding rate.

> Running full fstrim will flush all the discardable extents so there's a
> way to reset the discardable queue. What I still don't see as optimal is
> the single discard request sent per one period. Namely because there's
> the iops_limit knob.
> 

Yeah, it's not really ideal at the moment for much slower paced systems
such as our own laptops. Adding persistence would also make a big
difference here.

> My idea is that each timeout, 'iops_limit' times 'max_discard_size' is
> called, so the discard batches are large in total. However, this has
> impact on reads and writes and also on the device itself, I'm not sure
> if the too frequent discards are not making things worse (as this is a
> known problem).
> 

I spent a bit of time looking at the impact of discard on some drives
and my conclusion was that the iops rate is more impactful than the size
of the discards (within reason, which is why there's the
max_discard_size). On a particular drive, I noticed if I went over 10
iops of discards on a sustained simple read write workload, the
latencies would double. That's kind of where the 10 iops limit comes
from. Given the latency impact, that's why this more or less trickles it
down in pieces rather than as a larger batch.

> I'm interested in more strategies that you could have tested in your
> setups, either bps based or rate limited etc. The current one seems ok
> for first implementation but we might want to tune it once we get
> feedback from more users.

Definitely, one of the things I want to do is experiment with different
limits and see how this all correlates with write amplification. I'm
sure there's some happy medium that we can identify that's a lot less
arbitrary than what's current set forth. I imagine it should result in
some graph that we can correlate delay and rate of discarding to a
particular write amp given a fixed workload.

Thanks,
Dennis

  reply index

Thread overview: 49+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-12-14  0:22 Dennis Zhou
2019-12-14  0:22 ` [PATCH 01/22] bitmap: genericize percpu bitmap region iterators Dennis Zhou
2019-12-14  0:22 ` [PATCH 02/22] btrfs: rename DISCARD opt to DISCARD_SYNC Dennis Zhou
2019-12-14  0:22 ` [PATCH 03/22] btrfs: keep track of which extents have been discarded Dennis Zhou
2019-12-14  0:22 ` [PATCH 04/22] btrfs: keep track of cleanliness of the bitmap Dennis Zhou
2019-12-14  0:22 ` [PATCH 05/22] btrfs: add the beginning of async discard, discard workqueue Dennis Zhou
2019-12-14  0:22 ` [PATCH 06/22] btrfs: handle empty block_group removal Dennis Zhou
2019-12-14  0:22 ` [PATCH 07/22] btrfs: discard one region at a time in async discard Dennis Zhou
2019-12-14  0:22 ` [PATCH 08/22] btrfs: add removal calls for sysfs debug/ Dennis Zhou
2019-12-18 11:45   ` Anand Jain
2019-12-14  0:22 ` [PATCH 09/22] btrfs: make UUID/debug have its own kobject Dennis Zhou
2019-12-18 11:45   ` Anand Jain
2019-12-14  0:22 ` [PATCH 10/22] btrfs: add discard sysfs directory Dennis Zhou
2019-12-18 11:45   ` Anand Jain
2019-12-14  0:22 ` [PATCH 11/22] btrfs: track discardable extents for async discard Dennis Zhou
2019-12-14  0:22 ` [PATCH 12/22] btrfs: keep track of discardable_bytes Dennis Zhou
2019-12-14  0:22 ` [PATCH 13/22] btrfs: calculate discard delay based on number of extents Dennis Zhou
2019-12-30 16:50   ` David Sterba
2020-01-02 16:45     ` Dennis Zhou
2019-12-14  0:22 ` [PATCH 14/22] btrfs: add bps discard rate limit Dennis Zhou
2019-12-30 17:58   ` David Sterba
2020-01-02 16:46     ` Dennis Zhou
2019-12-14  0:22 ` [PATCH 15/22] btrfs: limit max discard size for async discard Dennis Zhou
2019-12-30 18:00   ` David Sterba
2019-12-30 18:08   ` David Sterba
2020-01-02 16:48     ` Dennis Zhou
2019-12-14  0:22 ` [PATCH 16/22] btrfs: make max async discard size tunable Dennis Zhou
2019-12-30 18:05   ` David Sterba
2020-01-02 16:50     ` Dennis Zhou
2019-12-14  0:22 ` [PATCH 17/22] btrfs: have multiple discard lists Dennis Zhou
2019-12-14  0:22 ` [PATCH 18/22] btrfs: only keep track of data extents for async discard Dennis Zhou
2019-12-30 17:39   ` David Sterba
2020-01-02 16:55     ` Dennis Zhou
2019-12-14  0:22 ` [PATCH 19/22] btrfs: keep track of discard reuse stats Dennis Zhou
2019-12-30 17:33   ` David Sterba
2020-01-02 16:57     ` Dennis Zhou
2019-12-14  0:22 ` [PATCH 20/22] btrfs: add async discard header Dennis Zhou
2019-12-14  0:22 ` [PATCH 21/22] btrfs: increase the metadata allowance for the free_space_cache Dennis Zhou
2019-12-14  0:22 ` [PATCH 22/22] btrfs: make smaller extents more likely to go into bitmaps Dennis Zhou
2019-12-17 14:55 ` [PATCH v6 00/22] btrfs: async discard support David Sterba
2019-12-18  0:06   ` Dennis Zhou
2019-12-19  2:03     ` Dennis Zhou
2019-12-19 20:06       ` David Sterba
2019-12-19 21:22         ` Dennis Zhou
2019-12-19 20:34     ` David Sterba
2019-12-19 21:17       ` Dennis Zhou [this message]
2019-12-30 18:13 ` David Sterba
2019-12-30 18:49   ` Dennis Zhou
2020-01-02 13:22     ` David Sterba

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20191219211755.GA38803@dennisz-mbp.dhcp.thefacebook.com \
    --to=dennis@kernel.org \
    --cc=clm@fb.com \
    --cc=dsterba@suse.com \
    --cc=dsterba@suse.cz \
    --cc=josef@toxicpanda.com \
    --cc=kernel-team@fb.com \
    --cc=linux-btrfs@vger.kernel.org \
    --cc=osandov@osandov.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-BTRFS Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-btrfs/0 linux-btrfs/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-btrfs linux-btrfs/ https://lore.kernel.org/linux-btrfs \
		linux-btrfs@vger.kernel.org
	public-inbox-index linux-btrfs

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-btrfs


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git