Mark Goodge wrote:
I'm hoping someone can help point me in the right direction for the
solution to this one. I'm in the process of adding a new server to an
existing mail cluster, but for some reason I'm getting really slow
throughput. The problem seems to be delays in the active queue - here's
a sample from qshape:
T 5 10 20 40 80 160 320 640 1280 1280+
TOTAL 253 0 0 0 0 252 1 0 0 0 0
example.com 18 0 0 0 0 18 0 0 0 0 0
example.net 16 0 0 0 0 16 0 0 0 0 0
I strongly suspect that your amavisd-new is the bottleneck.
Examine your logs and the output of the amavisd-nanny program.
Amavisd will log timing breakdowns at log_level 2 or higher;
that can be useful to see where it's spending its time. Slow
DNS can cause SpamAssassin to take longer than expected to
process mail.
amavisd-nanny will show how many amavisd-new threads are
available and how many are being used - maybe you don't have
enough configured or they aren't all being used for some reason.
I believe there was a patch for qshape posted a while back
that changes it to show mail waiting for the content_filter,
rather than the final destination. Search the archives...
Or maybe just a boring old networking problem, such as a
duplex mismatch or flaky cable, that's dramatically slowing
down transmission.
Pick one message and follow the logs for that message from
submission, through the content_filter, to final delivery.
That should help narrow it down.
--
Noel Jones
