On 04/13/11 20:43, Simon Hobson wrote: > Thanks Nigel for your help. While not directly fixing the problem, > you've certainly given me some good hints that have helped me narrow > things down a bit. > > > Perhaps I should have started with ... > > ... upgrading the software. > > The backend system was up to date as Debian Lenny, today I upgraded > it to Squeeze (which includes upgraded MySQL from 5.0 to 5.1). The > performance is the same, but the errors now appear to have gone. With > the errors gone, there aren't the delays there were before while > things timed out, but the rate still seems to be limited to about 2 > recipients/second. With logging turned down/off, the raw query rate > with policyd is now a little over 60 queries/s. It seems to make no > difference if I do 20 mails to 50 recipients each, or 100 mails to 10 > recipients each. > > I think for my next test I need to generate mails from more than one > system and see if the rate limit is in the mail cluster or the source > machine.
I'm glad you making progress man. Your findings really help and will help alot of people looking for similar info in future. What sort of CPU usage are you seeing from policyd vs. mysql? policyd should only be restricted by your CPU processing power and your DB speed, and I've personally not managed to saturate the CPU portion and its always the DB speed that gets hit first it seems. -N
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Users mailing list [email protected] http://lists.policyd.org/mailman/listinfo/users
