Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: 8BIT
Subject: RE: 2.6.3 userspace freeze
Date: Thu, 18 Mar 2004 16:47:45 +0100
Message-ID: <222BE5975A4813449559163F8F8CF503458463@cohsrv1.cohaesio.com>
Thread-Topic: 2.6.3 userspace freeze
Thread-Index: AcQIGgZDYBA+pesuToqxuSQSyRYWQAAAfjogATjgjzA=
From: "Anders K. Pedersen" <akp@cohaesio.com>
To: "Con Kolivas" <kernel@kolivas.org>
Cc: "Jan Kara" <jack@suse.cz>, "Andrew Morton" <akpm@osdl.org>,
       <linux-kernel@vger.kernel.org>
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2589
Lines: 65

Hello,

> > -----Original Message-----
> > From: Con Kolivas [mailto:kernel@kolivas.org] 
> > Sent: Friday, March 12, 2004 11:09 AM
> > To: Anders K. Pedersen
> > Cc: Jan Kara; Andrew Morton; linux-kernel@vger.kernel.org
> > Subject: Re: 2.6.3 userspace freeze

> > Each log you've shown so far shows you getting updatedb stuck 
> > in D which 
> > appears to be the common link. It could be your updatedb is 
> > busy scanning 
> > directories it probably shouldn't.
> > Check your updatedb.conf (usually in /etc) and see that you 
> > have at least 
> > these entries in PRUNEFS
> > 
> > PRUNEFS="nfs,smbfs,ncpfs,proc,devpts,supermount,vfat,iso9660,u
> > df,usbdevfs,devfs,usbfs,sysfs"
> 
> Thank you for the suggestion - I had the following in my 
> updatedb.conf:
> 
> PRUNEFS="devpts NFS nfs afs proc smbfs autofs auto iso9660"
> 
> so I have just added:
> 
> PRUNEFS="$PRUNEFS ncpfs supermount vfat udf usbdevfs devfs 
> usbfs sysfs"
> 
> Of these, the only active one is sysfs. I will give it 
> another try with these settings to night. However, while 
> you're correct that top shows updatedb in D state in the 
> latest test (in the mail dated 2004-mar-12 09:47 - crashed at 
> 04:02:21) as well as the first one I submitted (mail dated 
> 2004-mar-10 10:13 - crashed at 04:02:33), the one in between 
> (mail dated 2004-mar-11 01:46) doesn't as it crashed at 
> 01:26:32, and updatedb doesn't start until 04:02 as part of 
> cron.daily. Also, when I originally upgraded this server to 
> 2.6.3 it rebooted four or five time during the night, before 
> we downgraded it to 2.4 again, and updatedb couldn't have 
> been the cause for more than one of these. One thing, I will 
> try though, is to run updatedb manually with and without the 
> additions above to see if it triggers the deadlock immediately.

I've tried running updatedb manually a few times under 2.6.4 now - both
with and without the PRUNEFS addition above. All of these completed
without any problems.

Also, I left the server running 2.6.4 for a night, where it rebooted
five times with different intervals (ranging from ~20 minutes to ~3
hours), before I switched it back to 2.4. In other words, it doesn't
seem to be any of the cron.daily'es (i.e. updatedb), that specifically
triggers it.

Any other suggestions would be most welcome.

Regards,
Anders K. Pedersen
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/