At 02:00 PM 9.7.2003 -0400, Chuck Swiger wrote:
>Jack L. Stone wrote:
>> A while back, on a couple of occasions, I posted a query about some bad
>> behavior on my mail server. For the past several months, it has been either
>> crashing/reboot or just rebooting. It's ALWAYS triggered by a SSH login,
>> but at random and ONLY at the "su" to root -- usually the most time before
>> reboot is about 2+ weeks and then contrasted by 2 in a row right after the
>> reboot -- actually no pattern. It has never happened directly at the
>[ ... ]
>> There are no indications of anything in the logs, and no core dumps. It
>> just stops and reboots, and any random time it pick. Only a couple of times
>> it has crashed without the remote login.
>These two paragraphs contradict each other, at least in part.  :-)

Except, I doubt if those 2 nighttime reboots had the same problem....that's
why I said always triggered by login to root.... forget the 2 unrelated ones.

>You're seeing frequent crashes, which seem to be strongly correlated with 
>logging in as root, but you've also noticed crashes "without the remote
>too?  You should build a debug kernel, and enable dumping the system to swap 
>upon a panic ("man crash"), so that you have more information about the
>> One tip was that I might have stale NFS mountabs -- cleared them out, but
>> problem persisted.
>> The above tip was suggested when I mentioned that on a couple or more of
>> the occurrences, I managed to get to the console quickly enough to see (in
>> bright bold) "lockmgr locking against myself" -- or close to that. My
>> google of that error does mention stale mounts, but mostly about esoteric
>> code stuff. No fix found anywhere.
>Hmm.  Are you performing local mail delivery to NFS volumes?
No, just running backups to backup server over NFS... and share the:
/usr/ports ... /usr/obj ... and /usr/src from the "build" machines.

>Normally (or historically, anyway), NFS locking problems cause rpc.lockd to 
>crash or wedge, thus resulting in NFS locking not working and possibly grim 
>results to file consistency for anything being changed by two or more
>at the same time.
>However, NFS locking problems generally do not result in a system panic.
>[ ... ]
>> http://sageweb/tmp/1-lsof.txt
>> http://sageweb/tmp/2-lsof.txt
>These URLs aren't fully-qualified hostnames.  Please try again.  :-)
Yeah, drats! Already sent these:


Best regards,
Jack L. Stone,

SageOne Net
[EMAIL PROTECTED] mailing list
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Reply via email to