Re: 2.1.115: BAD problem with quotas

Carlos Carvalho (carlos@fisica.ufpr.br)
Tue, 11 Aug 1998 11:48:56 -0300 (EST)


Bill Hawes (whawes@transmeta.com) wrote on 10 August 1998 22:17:
>> Our server is crashing every 15min with a strange problem: anything
>> that accesses disk gets stalled. Consequently, the load average starts
>> to climb until the machine gets completely stuck. The kernel still
>> responds, but root cannot log in nor give any command. I'm obliged to
>> hard-rest it :-(
>>
>> This is apparently due to quotas. Disabling them makes the problem
>> disappear. The last crash happened as soon as I gave the command
>> edquota -p <prototype> user1 user2.
>>
>> I've seen a few fixes for dquot/free, and they're applied in .115, but
>> it looks like there's more to it :-(
>
>Any oopsen or other clues in your log files?

Nothing :-( (otherwise I'd have given the info). This is
understandable, since the disk becomes "locked", that is,
inaccessible. There are also no messages on the console. The machine
still tries to answer. If you press enter at the console you get a
login prompt, you type the user name and <enter>, the password prompt
appears but then no more answers. You can switch vt's, but the same
happens.

I've been using quotas for some weeks already, and the crash never
happened. Several users have filled their quota on /home, without
problems. The news SEEM to be that this time a user has filled his
quota on the /var partition, where the mail arrives. Could this
lockout be caused by a root process (the mail delivery agent) trying
to write beyond the user's quota on a user's file? Root has no
quotas...

The machine runs with raid5 on all partitions, and two aic7xxx
adapters, but this has no influence on the problem. It's enough to
turn quotas off, as they're now, and the machine is rock solid.

Carlos

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html