linux-btrfs.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Stephane Chazelas <stephane_chazelas@yahoo.fr>
To: linux-btrfs <linux-btrfs@vger.kernel.org>
Subject: Re: write(2) taking 4s
Date: Tue, 19 Jul 2011 10:32:52 +0100	[thread overview]
Message-ID: <20110719093252.GA24559@yahoo.fr> (raw)
In-Reply-To: <20110718193725.GA6298@yahoo.fr>

2011-07-18 20:37:25 +0100, Stephane Chazelas:
> 2011-07-18 11:39:12 +0100, Stephane Chazelas:
> > 2011-07-17 10:17:37 +0100, Stephane Chazelas:
> > > 2011-07-16 13:12:10 +0100, Stephane Chazelas:
> > > > Still on my btrfs-based backup system. I still see one BUG()
> > > > reached in btrfs-fixup per boot time, no memory exhaustion
> > > > anymore. There is now however something new: write performance
> > > > is down to a few bytes per second.
> > > [...]
> > > 
> > > The condition that was causing that seems to have cleared by
> > > itself this morning before 4am.
> > > 
> > > flush-btrfs-1 and sync are still in D state.
> > > 
> > > Can't really tell what cleared it. Could be when the first of
> > > the rsyncs ended as all the other ones (and ntfsclones from nbd
> > > devices) ended soon after
> > [...]
> > 
> > New nightly backup, and it's happening again. Started about 40
> > minutes after the start of the backup.
> [...]
> > Actively  running at the moment are 1 rsync and 3 ntfsclone.
> [...]
> 
> And then again today.
> 
> Interestingly, I "killall -STOP"ed all the ntfsclone and rsync
> processes and:
[...]
> Now 95% of the write(2)s take 4 seconds (while it was about 15%
> before I stopped the processes).
[...]

And this morning, after killing everything so that nothing was
writing to the FS anymore, 95% of write(2)s were delayed as well
(according to strace -Te write yes > file-on-btrfs).

Then I rebooted (sysrq-b) and am trying btrfsck (from
integration-20110705) on it, but btrfsck is using 8G of memory
on a system that has only 5G so it's swapping in and out
constantly and getting nowhere (and renders the system hardly usable)

I found
http://thread.gmane.org/gmane.comp.file-systems.btrfs/5716/focus=5728
from last year. Is that still the case?

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
 1950 root      20   0 7684m 4.4g  232 R    4 91.1   4:22.87 btrfsck
(and still growing)

 vmstat 1
procs -----------memory---------- ---swap-- -----io---- -system-- ----cpu----
 r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa
 2  2 3232016 115232   4524   3520  698  708  3305   716  991  570  3  1 56 39
 0  2 3231816 111536   5976   3428 2964  532  4912   532 1569  683  1  0 46 53
 0  2 3231144 105832   8144   3536 3140   24  5324    24 1612  392  1  1 38 60
 0  2 3231532 104964   8180   3684 2672  900  2708   900 1017  324  1  1 34 64

-- 
Stephane

      reply	other threads:[~2011-07-19  9:32 UTC|newest]

Thread overview: 37+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-07-03 19:09 Memory leak? Stephane Chazelas
2011-07-03 19:38 ` cwillu
2011-07-06  8:11   ` Stephane Chazelas
2011-07-07  8:09     ` Stephane Chazelas
2011-07-07  8:20       ` Li Zefan
2011-07-07  8:37         ` Stephane Chazelas
2011-07-08 12:44     ` Stephane Chazelas
2011-07-08 15:06       ` Chris Mason
2011-07-08 15:41         ` Stephane Chazelas
2011-07-08 16:11           ` Stephane Chazelas
2011-07-08 16:17             ` Chris Mason
2011-07-08 16:57               ` Stephane Chazelas
2011-07-09 17:11               ` Stephane Chazelas
2011-07-08 16:15           ` Chris Mason
2011-07-08 17:06             ` Stephane Chazelas
2011-07-08 20:04             ` Stephane Chazelas
2011-07-08 20:12               ` Chris Mason
2011-07-09  7:09                 ` Stephane Chazelas
2011-07-09  7:42                   ` A lot of writing to FS only read (Was: Memory leak?) Stephane Chazelas
2011-07-10  5:58                   ` Memory leak? Stephane Chazelas
2011-07-09 17:09         ` Stephane Chazelas
2011-07-09 19:25           ` cwillu
2011-07-09 20:36             ` Stephane Chazelas
2011-07-10 12:44               ` Chris Mason
2011-07-10 18:37                 ` Stephane Chazelas
2011-07-11  9:01                   ` Stephane Chazelas
2011-07-11 15:00                     ` Chris Mason
2011-07-11 15:35                       ` Stephane Chazelas
2011-07-11 16:25                         ` Chris Mason
2011-07-11 16:34                           ` Stephane Chazelas
2011-07-12 11:40                     ` Stephane Chazelas
2011-07-16 12:12                     ` write(2) taking 4s. (Was: Memory leak?) Stephane Chazelas
2011-07-16 16:22                       ` Stephane Chazelas
2011-07-17  9:17                       ` Stephane Chazelas
2011-07-18 10:39                         ` write(2) taking 4s Stephane Chazelas
2011-07-18 19:37                           ` Stephane Chazelas
2011-07-19  9:32                             ` Stephane Chazelas [this message]

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=20110719093252.GA24559@yahoo.fr \
    --to=stephane_chazelas@yahoo.fr \
    --cc=linux-btrfs@vger.kernel.org \
    /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
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).