At 09:03 AM 4/13/2004, you wrote:
>> Are you sure all the >> blocklists you've chosen are >> working properly and replying in a >> reasonable timeframe? Not positive - I've seen rejections from all of them in the maillog. Other than that I have no clue how to test them. BTW, they are spamcop, spamhaus, and orbl.
we recently had a serious problem with server load - the first thing we did was to remove all blocklists APART from spamcop - it made things better, but we still saw timeouts on spamcop. These appear to have gone away for now, but we only have spamcop left as a blocklist. We also dropped the timeout time limit to 5 seconds on the spamassassin config.
Next, I noticed that a LARGE amount of simultaneous emails to an account could generate upwards of 20+ spamd daemons - the account causing the problem was a mailbox used for search engine submissions - we turned off spamassassin for this user (as all mail is simply thrown away, why bother determining that it's spam?).
I'm currently investigating how to limit the amount of spamd processes - I'm sure I've seen it as a config for mailscanner, but need to find out if it can be done for spamd - and basically the problem arose from having too many processes gobbling all the available RAM and causing the server to swap like a mad thing!
Interestingly, we have a mixture of nuonce spamassassin/mailscanner and bassi's mailscanner - the problem ONLY appear to happen on our nuonce installations - the server RAM requirements appear to be SIGNIFICANTLY higher, and although we have 768mb and 1024mb on each machine, this simply is NOT enough.
fwiw, I've seen server load averages as high as 30 during crunch times, and when I see these, I'll attempt to kill off the spamd processes - better to let some spam through than have clients complain of web sites not responding at all. We've been lucky to spot these as they happen for the most part - but it's only a matter of time before it happens in the middle of the night.