On 9 Mar 2005, at 19:41, David Sparks wrote:
I've been running this code for a day now without crashes or signs of memory growth on about 3M messages.
What I've noticed is that certain messages are almost "stuck" on the sending MTA because processing of them stalls (in 15000 out of 3Million). The envelope sender and recipient are always the same in all of the 3M messages so I doubt that is involved.
The plugins I'm running are rcpt_ok and queue/spamtrap. Is there a DNS query based off the Reply-To or From address? I have require_resolvable_fromhost set to 0.
What am I missing?
Nothing - sounds like a bug. Any chance you can find out what phase the messages get stuck in?
As it usually turns out, I was mistaken. After firing up tcpdump on both the local and remote sides it because obvious that the problem is that the remote MTA (sendmail in this case) is the one doing the DNS lookups and timing out (after 56 seconds).
I don't know why sendmail would do a DNS query on the domain in the From: address ... as I though I had turned off checks like that and this doesn't happen in our other spam trapping operations. But I digress ...
In any case, thanks for the help -- I'll get back to testing.
Cheers,
ds
