Subject: Re: sendmail - the memory hog #1
To: None <email@example.com>
From: Jukka Marin <firstname.lastname@example.org>
Date: 11/20/1996 09:45:30
> > PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND
> > 19539 root 2 0 7540K 7728K sleep 2:50 18.41% 18.41% sendmail
> > 19185 root 18 0 11M 12M sleep 4:29 0.00% 0.00% <sendmail>
> 1) With what arguments did you start sendmail?
The normal -bd -q30m
> 2) Did you remember to rebuild sendmail.cf after the upgrade to 8.8.3?
Still running 8.7.5/8.6.9
> 3) How many SMTP connections to your machine does 'netstat -f inet'
> show? What TCP state are they in?
At the moment, there are 21 sendmail processes, but only 11 SMTP connections
(netstat -f inet | grep smtp |wc -l). Most connections are in CLOSE_WAIT
state, three in TIME_WAIT and one in SYN_SENT.
> 4) Anything interesting result from 'fgrep sendmail /var/log/*'?
Nothing that I can see. Lots of messages about successful deliveries and
normal problems (there was one machine which didn't accept mail for some
time, so I have timeout messages in /var/log/maillog). No errors in
> 5) Use 'ps -l' instead of 'ps -ux'. What are all those sendmail
> processes sleeping on (WCHAN)?
All but one (the original sendmail process) are sleeping on 'lockf'.
The memory hogging processes are trying to deliver queued mail. It
seems every sendmail process that is launched gets in this state now.
The first one was started when I rebooted the machine and the others
at 30 minute intervals. No sendmail process has exitted. Some are
trying to deliver the same message, most are "working" on different
---> http://www.jmp.fi/~jmarin/ <---