[Dovecot] Gah, pop3 falling over

[Timo Sirainen]

On Mon, 2008-10-27 at 09:51 -0700, Brandon Lamb wrote:
> > Two servers were working fine and then all of a sudden this morning
> > bam 500 load and theres a million dovecots and tons of these lines in
> > the logs.

I guess the load is about I/O load, not CPU load?

> > It is like everyone gets up at exactly 8:30 and hits the pop3 server at once.
> >
> > Anyway, dovecot 1.1.5, nfs mount, debian lenny/sid amd64,
> > 2.6.26-1-amd64 i had the performance mode options for login
> > configured, i just set back to the single process per login though, I
> > think that hasnt fallen over that way but I really dont remember
> > anymore. I included the dovecot -n using the suggest performance login
> > settings? Do I have something totally out of whack?

If the virtual -> physical size change didn't help, I guess that wasn't
the problem then. Another possibility could be Maildir++ quota working
non-optimally. You could try temporarily disabling it for POP3 to see if
that helps.

> dotlock_use_excl = yes
> fsync_disable = no
> lock_method = fcntl
> 
> Could I be running into some nfs locking issue? Wouldnt a reboot clear
> any nfs um, handles or whatnot that the mail server(s) might have? Do
> these look like the right values for using NFS? Should I try using
> lock_method = dotlock instead?

I guess you could try lock_method=dotlock too.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
Url : http://dovecot.org/pipermail/dovecot/attachments/20081027/3a5fa36b/attachment.bin