"MM" == Marc MERLIN <marc_news@vasoftware.com> writes:
MM> Ideally it'd be in Defaults.py so that it doesn't get blown
MM> away on upgrade, but that's already a good thing.See my recent checkin. :)
MM> Out of curiosity, is it really faster to have the whole mesage
MM> in a pickle instead of plaintext?I actually don't remember if I timed it or not, but I suspect it has to be faster to do a binary unpickle than to re-parse and re-generate for each queue move. It probably matters little for held messages.
MM> Another titbit (sp?) of info: On sf.net, I stat files held for
MM> moderation and delete everything that's been there for more
MM> than a month. You'd be surprised how many spams held for
MM> moderation and never moderated we had before that :-)Now that would be a cool script to add, although with SpamAssassin, we've been able to /vastly/ reduce the amount of spam coming through python.org. The right tool for the job, etc...
MM> I'll probably also contribute a small script that watches the
MM> number of messages in the mailman queue, and sends a warning
MM> if there are more than X (meaning that you have some serious
MM> backlog, or some other problem)Which reminds me, how's your performance been with Exim4? We're seeing a lot of weird stuff happen that wasn't happening with Exim3, specifically, we seem to exhaust our open file limit about every 3rd or 4th day. It's so bad that Mailman will throw a traceback indicating it couldn't import a module (which is presumably because Python couldn't open the module file to read from). I've seen this wedge Mailman and I've also seen Mailman recover, and I haven't been able to totally correlate this with peak traffic. It's disturbing and we'll have to do something about it, although hopefully not as drastic as reverting to Exim3. I'm just wondering if any other Exim4 users are seeing similar problems.
-Barry