From: Ben Spencer
Date: 2007-11-13 06:21:57 -0500
Subject: Follow-up: 10 second milter timeout
More information..: http://www.milter.info/#Support
For the list archives:
OS: RHEL4 Update 4 (Linux 32 bit)
Impacts: at least several Snertsoft milters (milter-spamc, milter-spiff,
milter-clamav). Note: there is no reason to think that this is Snertsoft
milter specific at this time. Given the workaround, I would think this could
happen with any milter.
In the mail logs, we noticed that sendmail would give up on the milter if
the milter took longer then 10 seconds to process the mail. This is in spite
of the "T=" settings (example: T=C:15m;S:4m;R:4m;E:10m) being set longer
then 10 seconds. The milter did usually return the data even after sendmail
gave up on it. Anthony looked through the various Snertsoft milter code and
couldn't come up with anything there which was causing it. The libmilter
code also didn't seem to have anything in it which looked to cause this.
Given that the milter did return data, it really looked like it was sendmail
giving up on the milter -- and not the milter giving up.
While trying to deal with this, I removed the "F=T" setting (example:
changed "F=T;T=C:15m;S:4m;R:4m;E:10m" to "T=C:15m;S:4m;R:4m;E:10m")
surprisingly, this "issue" went away and things have been running well ever
since with no such timeouts reported (and sendmail is indeed waiting over
the previous 10 seconds for various things to complete).
-- Binary/unsupported file stripped by Ecartis --
-- Type: application/x-pkcs7-signature
-- File: smime.p7s
Copyright 2009, 2012 by SnertSoft. All rights reserved.