On Thu, Aug 26, 2010 at 08:30:26AM -0700, Marc Perkel wrote:
> > Just out of interest, do you use a built-from-source exim, and if so, what do
> > you have EXIMDB_LOCK_TIMEOUT set to?
>
> I'll have to look that up. What should it be set to?
As long as you're prepared to wait for a lock. 60 (seconds, I assume) sounds
plenty to me.
> I am running it out of ram disk but it is a very busy server. In fact I
> have the queue running out of ram disk as well. The server processes
> millions of emails a day. Typically runs 500-800 concurrent connections.
> I'm wondering if this might be a bug in Exim where some better lockup
> detection code need to be added?
As yet, this thread contains no evidence that it's "locked up" (as in,
unrecoverably) - only that you're experiencing timeouts waiting for locks,
presumably because the system is heavily contended.
Can you provide some more specific numbers?
- how many messages per day
- how many connections per day
- how many times does each connection run a ratelimit check
Just trying to get a feel for how much you're hitting the ratelimit db.
"lsof" on the lock file might also be informative (to show how many processes
are waiting at any one time).
My guess is that there's no bug - you're just asking more of your system than
it can deliver.
--
Dave Evans
http://djce.org.uk/
http://djce.org.uk/pgpkey
