From: Craig Ringer (no email)
Date: Thu Jan 08 2004 - 11:17:11 EST
Ramprasad A Padmanabhan wrote:
> But what happens if you get corrupt mailboxes because in the dead of the
> night the partition run out of disk space
Tricky. I'm just not used to the idea of the server being /able/ to run
out of disk space ; my Cyrus mail spools are now on a separate volume
that always has several gigs of free space. It can be dynamically grown
while mounted read/write (I love reiserfs + lvm) and I get email alerts
if it's filling up. Combine this with a 512/128 ADSL link that puts a
very finite limit on the amount of mail we can recieve from outside the
business, and it's rather unlikely to fill up.
I can see how this could be different with a larger server, less
predictable loads, and a faster 'net connection. I take it that
significantly over-provisioning storage for the mail spools is not an
option?
> or because there is some one accessing using pop over a very slow link
> which breaks off every now and then
Interesting. I don't have remote POP3 users, so I have little experience
there. Many of my remote IMAPs users are using 56k however, and I
haven't seen any problems with that.
I would think that a mailbox left in a corrupt state after a failed POP3
session would indicate a problem with Cyrus's pop3d, but perhaps there's
something I'm not understanding going on.
> I am thinking of a hack in the ipurge source with another commandline
> option
> something like
> ipurge -f -d 30 --skipuntil 'user/someuser'
>
> Is that feasible
I don't know; sorry. Might it not be simpler overall, however, to give
ipurge the ability to skip corrupt mailboxes and move on to the next one?
Craig Ringer
|
|
|