On Wed, Oct 22, 2003 at 09:33:46AM -0700, Nick Arnett wrote:
Anybody seeing similar stuff going on with other mail?
I have only seen these delays with brin-l. I run postfix and procmail on my linux box and process 100's of messages a day (most of them spam, unfortunately) and I've never seen a similar problem.
Did these delays only start when you switched to your ISP's MTA?
I think it was much later, actually. And switching back here might be just the wrong thing to do. The real problem *may* be that our server, despite the upgrade a few months back, is overloaded. We're receiving many thousands of e-mails a day (total messages per day for our work is around 20K, but some are from web forums and newsgroups). I'm also wondering if the delays are caused by temporary network outages, when mail is then received by our backup MX at EasyDNS, which then tries to dump it all to us periodically, creating a huge spike. Moving our mail server to the Verio-hosted virtual machine would take care of that, I suspect.
We've added some load and network monitoring to the server, so we've gotten a better idea of what's going on over the last few days. Our virtual server is under-utilized in any case, and has terrific connectivity...
Oh, and the really big consumer of CPU cycles is SpamAssassin. But it does such a good job that I'm not sure we'll find a reasonable alternative. I'm still somewhat floored by the number of dictionary attacks.
... Just looked at the server settings, since we also installed Webmin, which makes it so much easier. We were only allowing Postfix 10 child processes, which I'm sure was waaaay too low. It's at 50 and I suspect it's still too low, but we'll try this for a while.
Glad you asked!
-- Nick Arnett Phone/fax: (408) 904-7198 [EMAIL PROTECTED]
_______________________________________________ http://www.mccmedia.com/mailman/listinfo/brin-l
