[Mimedefang] Over 6 min delay before passing messages?

[Matt Bruce]

Hi Les,

> How did you upgrade to sendmail 8.12.x?

I didn't. It was a virgin install - done the hard way from source
(deliberately). :)

> It splits the receive/deliver process into separate runs and you 
> need to start 2 daemons. If you built a non-Redhat version yourself 
> you need to fix the init script to match.

It's starting with SMTP+queueing, based upon a script from the
Sendmail.org site, so I doubt this is the issue. Other than the delay it
works fine. You are spot on about one thing, though: the fault is
definitely with Sendmail (or lower). I removed the MIMEDefang line from
sendmail.mc and the problem's still there.

I reduced the problem by adding the confTO_TIMEOUT definition to the
sendmail.mc file. By adding it with a 75s setting, the time delays in
delivery drop from a regular 6m20s delivery (which appears to be
OS-specific if the definition is omitted) down to 2m30s.

I've just increased the logging in Sendmail to 98 and am able to
determine that the remaining time delay appears to be due to a problem
at OSI Layer 8. Me! :) 

It fails on the first 2 MX records and succeeds on the third. With a
TO_TIMEOUT of 75s, that reduces the actual processing time to under a
minute, and I suspect I'll be able to make it shorter.

Many thanks to the chap who suggested I up the Sendmail logging. :)

Cheers,
Matt