jmaimon at ttec
Apr 28, 2004, 4:57 AM
Post #1 of 1
(289 views)
Permalink
|
|
Re: [Clamav-users] [PATCH] clamav-milter --max-child-wait
|
|
This is actualy a lot worse than I thought.
Apr 27 04:25:50 nameserver2 clamav-milter[4431]: ClamAv: private data
not NULL
Apr 27 04:25:50 nameserver2 clamav-milter[4417]: ClamAv: private data
not NULL
Apr 27 04:25:52 nameserver2 clamav-milter[4427]: ClamAv: private data
not NULL
Apr 27 04:25:52 nameserver2 clamav-milter[4429]: ClamAv: private data
not NULL
Apr 27 04:25:52 nameserver2 sendmail[4484]: i3R8PMaZ004484: Milter
(clamav-milter): timeout before data read
Apr 27 04:25:52 nameserver2 sendmail[4484]: i3R8PMaZ004484: Milter
(clamav-milter): to error state
Apr 27 04:25:53 nameserver2 sendmail[4484]: i3R8PMaZ004484:
from=<Gabe [at] attwireless>, size=1043, class=0, nrcpts
=1,
msgid=<354899489-1083054322-cardhu_blackberry.rim.net-1350- [at] engine3>,
proto=ESMTP, daemon=MTA, relay=mx20.blackberry.net
[206.51.26.249]
Apr 27 04:25:54 nameserver2 clamav-milter[4435]: ClamAv: private data
not NULL
Apr 27 04:25:54 nameserver2 clamav-milter[4518]: hit max-children limit
(11253 >= 25): waiting for some to exit
Apr 27 04:25:56 nameserver2 sendmail[4487]: i3R8PO9w004487: Milter
(clamav-milter): timeout before data read
Apr 27 04:25:56 nameserver2 sendmail[4487]: i3R8PO9w004487: Milter
(clamav-milter): to error state
Something obviously went horribly wrong.
Patches at http://www.jmaimon.com/clamav
(the ClamAv: private data not NULL are libmilter errors. I think the
reason they occur is because sendmail places the mail filter into error
state and closes the connection. Therefore I do not yet see how
clamfi_cleanup would get called)
Joe Maimon wrote:
So this morning clamd hung up. But then to add insult to injury
max-children of clamav-milter piled up behind it like a car wreck. This
patch adds the argument --max-child-wait=, which works like this.
* -1 wait 60 seconds for max_children and continue.(Old behavior)
* 0 or no value, no waiting, exit if max_children.(New default)
* >0 means wait that long, then exit if neccessary.(Mix)
Justification:
If there are too many milters, hanging around doing nothing for 60
seconds wont make it any better. Besides, sendmail milter could be
configured to time out before that.
Also, you would expect that saying --max-children actualy limited it to
--max-children.
Joe
|
