Re: Courier or Cyrus
On Thu, 17 Mar 2005, Michael Loftis wrote:
> >On Tue, 15 Mar 2005, Michael Loftis wrote:
> >>million emails/day. The only issue we've run into is an occasional
> >>deadlock with POP3 that requires rebooting. And that happens once every
> >>few months, if that. This is with 2.1.17.
> >
> >The Debian packages should fix that POP3 deadlock possibility, I think.
> >Something else that can cause POP3 deadlocks is apop (enabled by default),
> >if your box runs out of /dev/random entropy. Just disable apop if you
> >don't need it.
>
> Just had it do that this AM on our production server. 85 days 20 hrs 40
> minutes.
Weird, very weird. What does that production server run?
> >But a deadlock that requires a reboot? Don't you mean a Cyrus restart? If
> >it requires a _reboot_, something is very wrong indeed and I had never
> >heard of anythink like this.
>
> No, a reboot. Not cyrus restart. Cyrus restart doesn't' fix the POP3
> deadlock (and yes I make sure no pop3/pop3proxy processes are lying
Any stuck locks in the filesystem?
> around). Complete, full reboot. It did this to us in RedHat 7.3 systems
> too, so it's nothing specific to Debian woody+your fulldirhash package.
[...]
> As to what/why/how, I don't know. We've totally changed hardware since
> RH7.3 but we're still in the 2.1 series (2.1.17 Debian version, from you)
Ah. You might want to get the 2.1.18 packages from sid, they have some
security fixes.
This looks like some very weird bug :( next time, please strace -p the hung
processes, that might give us a clue. Send them some signals like sighup
and sigterm, to see how they behave (after attaching strace to them).
> except for one, I have to disable the groupchecks stuff since it ends up
> downloading nearly our entire LDAP database for every login. The other
Yeah, that thing is hideous on nsswitch setups where the databases are on
LDAP.
> Debian Cyrus 2.1 seems to handle lookout just fine, or atleast we don't get
> any complaints. :) (Honest, as much as I want to I'm no BOFH :))
Good. I had to track down and add some very weird patches to get it to
behave with outcrook... and it *was* Cyrus' fault, for once. It did not
syncronize seen states between two different concurrent sessions as well as
outlook wants.
> Yeah but what if the mailbox doesn't exist in the current mailboxes db?
I just add them back using a script.
> pain. If reconstruct could fix/readd mailboxes it'd probably work.
It should be trivial to readd that functionality in 2.2 and 2.3...
> >Hmm, come to think about it, maildir is probably superior in that it
> >stores the message flags in-band which is nicer for restores -- cyrus
> >will mark all restored messages as unread and lose all other flags on
> >that message as well). OTOH, how does one manage to have annotations and
> >per-user flags (such as seen state) using maildir?
>
> I think they get stored in an X-IMAP-Flags: or like header.
One for every per-user state? Creepy.
> >AS LONG AS you do not need cluster-wide shared mailboxes, at which point a
> >Cyrus murder cluster is the only thing that will work, AFAIK (and I would
> >love to hear otherwise!).
>
> Yup and the MURDER needs very little care and feeding once setup. It just
> works :)
Yep. The only thing I don't like is the mupdate master, but one can do HA
there rather easily anyway.
And murder makes it completely trivial to move mailboxes around the cluser,
allows for weird setups where some nodes are on the far side of WAN links
(and thus much closer to the customer's lan), etc. And it gives as many
caching proxies for lmtp, imap and pop3 as one could want :-)
--
"One disk to rule them all, One disk to find them. One disk to bring
them all and in the darkness grind them. In the Land of Redmond
where the shadows lie." -- The Silicon Valley Tarot
Henrique Holschuh
Reply to: