[280] in The Cryptographic File System users list
Re: cfs and memory usage . . .
daemon@ATHENA.MIT.EDU (Brian T. Schellenberger)
Tue Dec 3 11:18:12 2002
From owner-cfs-users@crypto.com Tue Dec 03 16:18:12 2002
Return-Path: <owner-cfs-users@crypto.com>
Delivered-To: cfs-mtg@CHARON.mit.edu
Received: (qmail 20040 invoked from network); 3 Dec 2002 16:18:12 -0000
Received: from softdnserror (HELO mx.crypto.com) (207.140.168.138)
by charon.mit.edu with SMTP; 3 Dec 2002 16:18:12 -0000
Received: (from majordomo@localhost)
by MultiHostMXServer (8.9.3/8.9.x4) id LAA04508
for cfs-users-list; Tue, 3 Dec 2002 11:00:08 -0500 (EST)
Received: from dockmaster.research.att.com (H-135-207-24-155.research.att.com [135.207.24.155])
by MultiHostMXServer (8.9.3/8.9.x4) with ESMTP id LAA10837
for <cfs-users@crypto.com>; Tue, 3 Dec 2002 11:00:06 -0500 (EST)
Received: from mail-green.research.att.com (mail-green.research.att.com [135.207.30.103])
by dockmaster.research.att.com (8.12.1/8.12.1) with ESMTP id gB3F91xi020050
for <cfs-users@nsa.research.att.com>; Tue, 3 Dec 2002 10:09:01 -0500 (EST)
Received: from janus (fpfw.research.att.com [135.207.1.2])
by mail-green.research.att.com (Postfix) with SMTP id 5213F1E163
for <cfs-users@nsa.research.att.com>; Tue, 3 Dec 2002 10:59:36 -0500 (EST)
Received: from mail-red.research.att.com ([192.20.225.110]) by janus; Tue, 03 Dec 2002 10:59:35 -0500 (EST)
Received: (from postfixfilter@localhost)
by mail-red.research.att.com (8.11.6/8.11.6) id gB3Fxi022561
for cfs-users@nsa.research.att.com; Tue, 3 Dec 2002 10:59:44 -0500
X-Authentication-Warning: mail-red.research.att.com: postfixfilter set sender to bts@fake.com using -f
Received: from ncsmtp03.ogw.rr.com (ncsmtp03.ogw.rr.com [24.93.67.84])
by mail-red.research.att.com (Postfix) with ESMTP id BD7961AB4BD
for <cfs-users@nsa.research.att.com>; Tue, 3 Dec 2002 10:59:41 -0500 (EST)
Received: from mail6.nc.rr.com (fe6 [24.93.67.53])
by ncsmtp03.ogw.rr.com (8.12.5/8.12.2) with ESMTP id gB3FwWix011008;
Tue, 3 Dec 2002 10:58:45 -0500 (EST)
Received: from this.is.fake.com ([24.162.238.30]) by mail6.nc.rr.com with Microsoft SMTPSVC(5.5.1877.757.75);
Tue, 3 Dec 2002 10:59:19 -0500
Received: by this.is.fake.com (Postfix, from userid 111)
id 35004BB39; Tue, 3 Dec 2002 10:56:48 -0500 (EST)
Content-Type: text/plain;
charset="iso-8859-1"
From: "Brian T. Schellenberger" <bts@babbleon.org>
To: Mark Knight <markk@knigma.org>
Subject: Re: cfs and memory usage . . .
Date: Tue, 3 Dec 2002 10:56:47 -0500
User-Agent: KMail/1.4.2
References: <200210310412.48667.bts@babbleon.org> <QLpsybKY5J79EwnK@lap.knigma.org>
In-Reply-To: <QLpsybKY5J79EwnK@lap.knigma.org>
Cc: cfs-users@nsa.research.att.com, freebsd-ports@freebsd.org
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Message-Id: <200212031056.47931.bts@babbleon.org>
X-Spam-Status: No, hits=-2.6 required=5.0
tests=DRASTIC_REDUCED,IN_REP_TO,REFERENCES,SIGNATURE_SHORT_DENSE,
SPAM_PHRASE_01_02,USER_AGENT,USER_AGENT_KMAIL
version=2.43-cvs
X-Spam-Level:
Sender: owner-cfs-users@crypto.com
Precedence: bulk
You have made my day by merely responding.
I have posted to the various lists and never gotten *any* response
before; I was really wondering if these were still used.
Note that I've tried drastically reducing "NINSTANCES" and I still run
into the problem.
I find that I can reproduce the problem quite easily by just using the
"find" command to do something that actually opens the files in one of
my big directories.
I would be more than happy to collaborate with you on doing this.
The one quasi-useful discovery that I *have* made is that it's perfectly
possible to just kill cfsd and start it up again without bothering to
detach and unmount and mount and re-attach; it yields "stale NFS
handles" in all the shells and/or other programs that have a file open
in CFS but is otherwise harmless and for the shells once merely has to
cd back to where one already was in order to get everything stable
again.
This at least lets me avoid crashing or rebooting all the time.
On Tuesday 03 December 2002 07:06 am, you wrote:
| In message <200210310412.48667.bts@babbleon.org>, Brian T.
| Schellenberger <bts@babbleon.org> writes
|
| >I use the FreeBSD "port" of cfs, and I have noticed that it is very
| >greedy in using memory.
| >
| >Now, I'm a little unfriendly to it now in that I have a number of
| > cfs file systems set up but even when I only had one or two it
| > would behave the same way:
| >
| >Memory usage is fine for "normal" activities, but if I do something
| > that scans lots of files, such as running a "find" command that
| > greps over the files in the CFS area or something, memory usage
| > will shoot up to astronomical levels (eg, over 300M or more).
| >
| >And that memory will never be freed.
| >
| >This will so exhaust memory that the system will frequently crash
| > not long afterwards.
| >
| >Now, it looks like cfs is *meant* to respond to a SIGALRM and clean
| > up memory that's no longer needed, and additionally to go through
| > this cleanup automatically every 60 seconds:
| >
| > signal(SIGALRM,grimreap);
| > alarm(60); /* every 60 secs */
| >
| >(at the end of main).
| >
| >But it looks like something is preventing this from happening, or
| > the grimreap routine just never agrees to clean up anything.
| >
| >I know that there are parameters I can tune to use less total memory
| > but I'm wondering if other people have seen this behavior and what,
| > if anything you did to deal with it.
|
| Just a note to say that I also have this problem, particularly when
| using directories with very large numbers of files. I've also
| reported it to the mailing list in the past, without joy.
|
| You've renewed my interest in finding a fix, I might have a look...
|
| Cheers,
--
Brian, the man from Babble-On . . . . bts@babbleon.org (personal)