From: Christian Reis <kiko@async.com.br>
To: NFS@lists.sourceforge.net
Cc: linux-kernel@vger.kernel.org
Subject: Re: NFS client locking hangs for period
Date: Fri, 25 Apr 2003 01:57:30 -0300 [thread overview]
Message-ID: <20030425015730.A29574@blackjesus.async.com.br> (raw)
Well, since I've more or less moved on from my original problems, I
should probably post a summary of what was going on, and what I did to
work around it.
Details can be read out from [1]: after a certain amount of time a
number diskless clients, which were mounting everything from the same
NFS server, started getting hung lock requests from the server. The
server ran 2.4.20, reiserfs over RAID-1 mounted with 2 SCSI disks on an
Adaptec 29160. The clients were debian woodys running 2.4.20.
Our diskless setup is a bit unusual: all the clients mount the same root
partition. I tried to be very careful to make sure no files were written
to on /, but I never got to the point where the clients could mount the
directory read-only. I used devfs to make sure that the /dev directories
were `localized' and syslog/console ownership and permissions kept sane.
The locking problem, however, was not related to the root filesystem --
it seems to have happened with files on the /var/log mount, which is
separate for each box (but still coming from a shared filesystem
/export/root on the server, which contains all the client directories).
If I mounted /var/log with the nolock option, they ran fine. This took
me a very long time to figure out, and I'd advise anyone with locking
problems to give it a go.
I should point out that this *does* seem to be a bug in the NFS server
code. I think it is associated with reiserfs, being that I haven't seen
it happen on other partition types. Rebooting the server cleared up the
problem. Erasing or changing files in /var/lib/nfs did not. While I was
initially using a volatile /var/lib/nfs directory on the *clients*, I
changed this on Trond's suggestion [2]. It did not fix the problem.
However, since I know little about the code itself, and it's not very
clear how one should debug, I was unable to pinpoint the exact source of
the problem, which very much saddens me. The workaround, however, was
quite effective.
[1] http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&th=9db70994c3458f46&rnum=1
[2] http://groups.google.com/groups?q=christian+reis+nfs+locking&hl=en&lr=&ie=UTF-8&scoring=d&selm=20030126231006%246e11%40gated-at.bofh.it&rnum=3
Take care,
--
Christian Reis, Senior Engineer, Async Open Source, Brazil.
http://async.com.br/~kiko/ | [+55 16] 261 2331 | NMFL
next reply other threads:[~2003-04-25 4:46 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2003-04-25 4:57 Christian Reis [this message]
-- strict thread matches above, loose matches on Subject: below --
2003-01-24 20:49 Christian Reis
2003-01-25 3:54 ` Neil Brown
2003-01-26 16:02 ` Christian Reis
2003-01-28 8:00 ` Denis Vlasenko
2003-01-28 16:44 ` Christian Reis
2003-01-29 21:53 ` Daniel Egger
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=20030425015730.A29574@blackjesus.async.com.br \
--to=kiko@async.com.br \
--cc=NFS@lists.sourceforge.net \
--cc=linux-kernel@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
all inboxes | Powered by JetHome®