Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 22 May 2009 07:42:41 -0400
From:      Doug Lee <dgl@dlee.org>
To:        freebsd-questions@freebsd.org
Subject:   More info (was Re: 4.11 panic every 23 hours 55 minutes or so)
Message-ID:  <20090522114241.GA69281@mini.local>
In-Reply-To: <20090517110657.GC2706@mini.local>
References:  <20090517110657.GC2706@mini.local>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, May 17, 2009 at 07:06:57AM -0400, Doug Lee wrote:
> One of the weirder things I've seen in a while here...
> 
> OS: FreeBSD 4.11 (yeah I know, old, but generally stable)
> CPU: Intel(R) Pentium(R) 4 CPU 2.00GHz
> real memory  = 536608768 (524032K bytes)
> Hds: IDE
> 
> Problem:  Ever since a suspitious power outage (I say suspitious
> because we think a surge was also involved), this box has been
> exhibiting kernel panics about every 23 hours 55 minutes, give or
> take about 4 minutes either way.  Obviously hardware is suspect,
> and hopefully in line for upgrade; but as FreeBSD has always proven
> so stable for me, I'm curious what on earth could cause this sort
> of regular panic?
> 
> It's not time of day; if I reboot at 2:00 AM, 3:55 PM, or any other
> time, it's 23:55 or so later I get a panic, whenever that may be.
> I think this rules out cron jobs, external attacks, and load-based
> issues.

Update: I killed mysqld, four nfsiods, Apache2, mpd, and maybe a
couple more no-longer-needed processes two mornings ago.  I also
disabled them at that time in rc.conf.  the next morning, the system
restarted with a panic as usual, BUT...

This morning, on the first boot that never ran all those processes, I
have not seen a restart yet, and we're at 1 day 1 hour as I speak.

I looked in /var/at earlier in the week and never found any scheduled
jobs.  It shouldn't be Cron, since it's sensitive to boot time, not
clock time.

Is there some way one of those processes, like mysqld, could be
scheduling an event to occur 24 hours after launch, without using
`at', and without having to be running 24 hours later?  Example:
Could mysqld schedule something without `at' that will run 24 hours
after mysqld starts even if mysqld is no longer running?

Also, is it even possible that any process could cause a  kernel-mode
page fault without there being damaged hardware?  Example:  Could some
mysql file be so corrupt that it would panic a perfectly fine machine?
I should hope not, but I wonder.


-- 
Doug Lee                 dgl@dlee.org                http://www.dlee.org
SSB BART Group           doug.lee@ssbbartgroup.com   http://www.ssbbartgroup.com
"When there is no enemy within, the enemies outside cannot hurt you."
--African Proverb



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20090522114241.GA69281>