From: Trond Myklebust <trond.myklebust@fys.uio.no>
To: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kenny Simpson <theonetruekenny@yahoo.com>,
Andrew Morton <akpm@osdl.org>,
linux-kernel@vger.kernel.org
Subject: Re: nfs unhappiness with memory pressure
Date: Mon, 05 Dec 2005 19:48:19 -0500 [thread overview]
Message-ID: <1133830100.8007.6.camel@lade.trondhjem.org> (raw)
In-Reply-To: <4394CFFF.3080009@yahoo.com.au>
On Tue, 2005-12-06 at 10:40 +1100, Nick Piggin wrote:
> > ...and most important of all: 'unstable' does _not_ mean that I/O is
> > active on those pages (unlike the apparent assumption in
> > vm_throttle_write.
> > That is why the choice is either to kick pdflush there, or to remove
> > nr_unstable from the accounting in that loop.
> >
>
> Doesn't matter if IO is actually active or not, if you've allocated
> memory for these unstable pages, then page reclaim can scan itself
> to death, which is what seems to have happened here. And which is
> what vm_throttle_write is supposed to help.
Unless someone somehow triggers an NFS commit, then nr_unstable is not
ever going to decrease, and your process will end up looping forever. In
fact, those nr_writeback that refer to NFS pages, will end up being
added to nr_unstable (because they have been written to the server, but
not committed to disk).
Cheers,
Trond
next prev parent reply other threads:[~2005-12-06 0:48 UTC|newest]
Thread overview: 89+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <20051115224645.27832.qmail@web34103.mail.mud.yahoo.com>
2005-11-15 23:47 ` mmap over nfs leads to excessive system load Kenny Simpson
2005-11-16 4:31 ` William Lee Irwin III
2005-11-16 6:05 ` Kenny Simpson
2005-11-16 7:45 ` Andrew Morton
2005-11-16 14:03 ` Trond Myklebust
2005-11-16 15:01 ` Kenny Simpson
2005-11-16 17:44 ` Trond Myklebust
2005-11-16 18:00 ` Andrew Morton
2005-11-16 18:34 ` Trond Myklebust
2005-11-16 18:38 ` Christoph Hellwig
2005-11-17 13:08 ` Nikita Danilov
2005-11-16 19:09 ` Andrew Morton
2005-11-16 20:05 ` Trond Myklebust
2005-11-16 20:56 ` Kenny Simpson
2005-11-16 21:02 ` Kenny Simpson
2005-11-16 21:09 ` Trond Myklebust
2005-11-16 21:17 ` Kenny Simpson
2005-11-16 21:41 ` Kenny Simpson
2005-11-16 21:57 ` Trond Myklebust
2005-11-16 22:04 ` Kenny Simpson
2005-11-16 22:39 ` Kenny Simpson
2005-11-16 23:06 ` Trond Myklebust
2005-11-17 15:40 ` Chuck Lever
2005-11-17 16:56 ` Kenny Simpson
2005-11-17 16:01 ` Kenny Simpson
2005-11-17 21:04 ` Andrew Morton
2005-11-17 21:15 ` Kenny Simpson
2005-11-18 16:55 ` Kenny Simpson
2005-11-18 17:26 ` Kenny Simpson
2005-11-18 21:57 ` Kenny Simpson
2005-11-21 17:13 ` infinite loop? with mmap, nfs, pwrite, O_DIRECT Kenny Simpson
2005-11-21 17:49 ` Chuck Lever
2005-11-21 18:40 ` Kenny Simpson
2005-11-21 20:39 ` Andrew Morton
2005-11-21 21:39 ` Kenny Simpson
2005-11-21 22:42 ` Trond Myklebust
2005-11-21 23:14 ` Kenny Simpson
2005-11-21 23:34 ` Andrew Morton
2005-11-21 23:58 ` Trond Myklebust
2005-11-22 0:09 ` Andrew Morton
2005-11-22 0:18 ` Trond Myklebust
2005-11-22 0:25 ` Trond Myklebust
2005-11-22 0:28 ` Andrew Morton
2005-11-22 0:49 ` Trond Myklebust
2005-11-30 20:04 ` nfs unhappiness with memory pressure Kenny Simpson
2005-11-30 21:42 ` Keith Mannthey
2005-11-30 22:18 ` Kenny Simpson
2005-12-01 14:49 ` Kenny Simpson
2005-12-05 18:01 ` Kenny Simpson
2005-12-05 19:44 ` Kenny Simpson
2005-12-05 20:14 ` Kenny Simpson
2005-12-05 20:13 ` Trond Myklebust
2005-12-05 20:33 ` Trond Myklebust
2005-12-05 20:52 ` Nick Piggin
2005-12-05 21:18 ` Trond Myklebust
2005-12-05 21:23 ` Trond Myklebust
2005-12-05 23:40 ` Nick Piggin
2005-12-06 0:48 ` Trond Myklebust [this message]
2005-12-05 21:51 ` Kenny Simpson
2005-12-05 21:04 ` Kenny Simpson
2005-12-05 22:39 ` Trond Myklebust
2005-12-06 3:36 ` Andrew Morton
2005-12-06 3:36 ` Andrew Morton
2005-12-06 4:40 ` Trond Myklebust
2005-12-06 5:42 ` Nick Piggin
2005-12-06 12:18 ` Trond Myklebust
2005-12-06 19:31 ` Kenny Simpson
2005-12-06 15:51 ` Kenny Simpson
2005-11-21 21:54 ` infinite loop? with mmap, nfs, pwrite, O_DIRECT Kenny Simpson
2005-11-21 20:12 ` Kenny Simpson
2005-11-17 17:02 ` mmap over nfs leads to excessive system load Chuck Lever
2005-11-17 17:07 ` Trond Myklebust
2005-11-18 19:59 ` Kenny Simpson
2005-11-16 21:31 ` Andrew Morton
2005-11-16 21:49 ` Trond Myklebust
2005-11-16 22:10 ` Andrew Morton
2005-11-16 22:23 ` Trond Myklebust
2005-11-16 22:38 ` Trond Myklebust
2005-11-16 22:50 ` Andrew Morton
2005-11-16 22:44 ` Andrew Morton
2005-11-16 23:10 ` Trond Myklebust
2005-11-17 0:06 ` Trond Myklebust
2005-11-17 0:25 ` Andrew Morton
2005-11-17 0:28 ` Trond Myklebust
2005-11-17 0:38 ` Andrew Morton
2005-11-17 0:47 ` Trond Myklebust
2005-11-16 18:48 ` Kenny Simpson
2005-11-16 19:06 ` Kenny Simpson
2005-12-06 13:40 nfs unhappiness with memory pressure Albert Herranz
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=1133830100.8007.6.camel@lade.trondhjem.org \
--to=trond.myklebust@fys.uio.no \
--cc=akpm@osdl.org \
--cc=linux-kernel@vger.kernel.org \
--cc=nickpiggin@yahoo.com.au \
--cc=theonetruekenny@yahoo.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
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®