> Good morning to everyone on the list. I hope that this note finds
> whatever part of the day going well for you.
>
> As the title implies we subjected 2.1.125 to a production load test
> this morning and had some difficulties.
Hi Greg,
Thanks for your test report on quota operations. I've made various comments below.
> At about 08:45 we experienced our first serious problem when users
> began reporting that they were unable to establish IMAP connections to
> the server. At this time the following string of debug messages
> appeared:
>
> ---------------------------------------------------------------------------
> Oct 14 08:49:06 pobox1 kernel: get_empty_dquot: pruning 128
> Oct 14 08:50:02 pobox1 kernel: get_empty_dquot: pruning 128
These are debugging messages and can be ignored, but they indicate that a shortage
of dquot structures is causing dentries to be pruned. (Pruning dentries frees
inodes, and a freed inode will release any dquot structures it's using.)
> During this time period the following syslog messages were generated.
> These would seem to be of more concern:
>
> Oct 14 08:54:41 pobox1 kernel: VFS: No free dquots, contact mvw@planets.elm.net
> Oct 14 08:56:43 pobox1 kernel: VFS: No free dquots, contact mvw@planets.elm.net
These show that even after freeing dentries, there still weren't any available
dquot structures.
> I am not sure that this was a true load average of 75 since the system
> was still reasonably responsive from the console. There were a large
> number of 'stuck' IMAP process in a state of 'dpget'. When we tried
> to run a quota command on a user the quota command hung as well.
The dquot allocation isn't allowed to fail, so the stuck processes were waiting
for dquots to be freed.
You didn't mention what value dquot-max was set to, but my guess is that it's way
to low for what is probably a very busy server. Try upping the available dquots
and see if that helps.
Regards,
Bill
-
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.tux.org/lkml/