NetBSD Problem Report #9212

Received: (qmail 17867 invoked from network); 17 Jan 2000 02:39:07 -0000
Message-Id: <200001170237.SAA20320@trebuchet.rc.vix.com>
Date: Sun, 16 Jan 2000 18:37:15 -0800 (PST)
From: gson@nominum.com (Andreas Gustafsson)
Reply-To: gson@nominum.com (Andreas Gustafsson)
To: gnats-bugs@gnats.netbsd.org
Subject: File system access slows down after weeks of uptime
X-Send-Pr-Version: 3.95

>Number:         9212
>Category:       kern
>Synopsis:       File system access slows down after weeks of uptime
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    kern-bug-people
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sun Jan 16 18:42:00 +0000 2000
>Closed-Date:    Fri May 31 03:43:41 +0000 2002
>Last-Modified:  Fri May 31 03:43:41 +0000 2002
>Originator:     Andreas Gustafsson
>Release:        19991223
>Organization:
Nominum, Inc.
>Environment:

System: NetBSD trebuchet.rc.vix.com 1.4P NetBSD 1.4P (TREBUCHET) #0: Thu Dec 23 22:42:19 PST 1999 gson@trebuchet.rc.vix.com:/usr/src/sys/arch/i386/compile/TREBUCHET i386

>Description:

My machine works flawlessly for a week or two after being rebooted.
Then compilations and other disk intensive activities suddenly start
taking many times longer than usual, and are accompanied by the rattle
of constant disk accesses.  I'm sure the machine is not swapping:
pstat -s says it is using only 8 out of 3336480 swap blocks.  Rather,
it seems as if the buffer cache somehow loses its effectiveness.

On the machine in case (550 MHz AMD Athlon, 256 MB RAM, IDE disk),
executing the command

  find /usr/share/man -print >/dev/null

for the first time after a reboot takes 2.2 seconds of real time.
Subsequent invocations usually take only 0.04 seconds (!), no doubt
thanks to the buffer cache.  However, when the machine is experiencing
the slowdown, each invocation takes 1.2 seconds, or 30 times the
usual.

Rebooting the machine fixes the problem, but it returns after 1-3
weeks of uptime.  It has now happened three times.

I have not enabled SOFTDEP, in case that makes a difference. 

>How-To-Repeat:

See above.

>Fix:

Unknown.
>Release-Note:
>Audit-Trail:

From: gson@nominum.com (Andreas Gustafsson)
To: gnats-bugs@gnats.netbsd.org, kern-bug-people@netbsd.org
Cc: Andreas.Gustafsson@nominum.com (Andreas Gustafsson)
Subject: Re: kern/9212: File system access slows down after weeks of uptime
Date: Fri, 11 Feb 2000 15:55:09 -0800 (PST)

 Michael Graff <explorer@flame.org> suggested that I try increasing the
 value of kern.maxvnodes.  The next time the slowdown appeared, after
 some 8 days of uptime, I did just that:

   # sysctl -w kern.maxvnodes=1500
   kern.maxvnodes: 778 -> 1500

 This made the problem go away, but only for another week.  It looks as
 if the kernel is losing vnodes at a fairly constant rate of about 100
 a day.  However, running "pstat -T" or "pstat -v" shows only some 174
 vnodes being in use.
 -- 
 Andreas Gustafsson, gson@nominum.com
State-Changed-From-To: open->closed 
State-Changed-By: gson 
State-Changed-When: Thu May 30 20:42:18 PDT 2002 
State-Changed-Why:  
I have not seen this problem in the last year or so.  Presumably 
whatever caused it has been fixed. 
>Unformatted:

NetBSD Home
NetBSD PR Database Search

(Contact us) $NetBSD: query-full-pr,v 1.51 2026/08/10 02:28:17 riastradh Exp $
$NetBSD: gnats_config.sh,v 1.10 2026/05/13 22:00:09 riastradh Exp $
Copyright © 1994-2026 The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.