Odd behavior in exim
I noticed that I am getting some strange behavior from Exim. While
checking puppet dashboard, I noticed that a few machines were being
changed every run. When I drilled down into these four machines, I saw
the following in puppet:
ensure changed 'stopped' to 'running'
This happened on all four hosts between puppet runs, on 2012-11-18
between 09:56 and 10:10. all sid machines running exim4, exim4-base,
exim4-config, and exim4-daemon-light, version 4.80-5.1. Note that I
have other sid boxes running the same versions of the same packages.
Two of the boxes affected are physical, and the other two are openvz
On these affected boxes, I see the exim process running:
# ps auxww | grep exim
Debian-+ 30804 0.0 0.0 46812 960 ? Ss 21:42 0:00
/usr/sbin/exim4 -bd -q30m
But exim4 status (which is what puppet uses) shows:
# /etc/init.d/exim4 status
[....] checking separate queue runner daemon.../etc/init.d/exim4:
done (not running).
[....] checking combined SMTP listener and queue runner
daemon.../etc/init.d/exim4: invalid arguments
The only other commonality is that during this window, each of these
four hosts have in their logs, which does not appear to be related
(though these being the only four boxes to have it and during that
timeframe...I don't believe in coincidences)
Nov 18 09:56:39 worf init: Trying to re-exec init
Nov 18 09:56:39 worf init: no more processes left in this runlevel
Nov 18 09:56:47 dax init: Trying to re-exec init
Nov 18 09:56:47 dax init: no more processes left in this runlevel
Nov 18 10:05:35 valiant init: Trying to re-exec init
Nov 18 10:07:08 defiant init: Trying to re-exec init
I have since rebooted dax, just to see, and /etc/init.d/exim4 status
still shows the same error.
I also checked the dpkg.logs, and the last exim4 upgrade on valiant
was 2012-10-05, dax was 2012-10-29, on worf and defiant was
So what would cause these four hosts, out of 16, to start giving
errors in /etc/init.d/exim4 status? The configs are managed by puppet,
and a check of all sid boxes shows all the configs match (the working
and non-working, a total of 9 hosts), as do /etc/init.d/exim4 files.
Can someone more familiar with exim4 give me an idea of what could be wrong?