[Mailman-Users] qrunner running but not flushing qfiles/in

[John Dennis]

On Fri, 2004-10-22 at 12:41, tomg at sqlclinic.net wrote:
> Hey Mark, thanks for the reply.  Yeah I figured that the missing qrunners were
> the problem but there was no log entry that was revealing. Then I read in the
> mailmanctl script that restart doesn't kill the parent ps, just the qrunners.
> So I manually stopped and started and that did it.  We'll see what happens on
> the next boot (last one before yesterday was 72 days ago so I'll have to write
> a note to myself).

Hmm ... depending on your host system arguably its preferable to use the
service command to manage services (e.g. invokes the mailman init.d
script, which then in turn invokes mailmanctl). For what its worth the
vanilla init.d script supplied with mailman calls mailmanctl with the
restart argument when invoked with restart. But a more typical restart
implementation in init.d scripts for services is stop followed by start
(currently done in our distribution). It sounds like Tom would not have
run into this problem if that were the case and perhaps the default
init.d script (misc/mailman.in) should do that, but read on ...

However I'm a little bit perplexed by why a stop followed by a start
caused a missing qrunner to be recreated. The master watches for the
death of a child and automatically restarts it unless the norestart
option is given or a handful of other conditions exist including
exceeding the max number of restart attempts, all of which are logged. A
master restart is implemented via sending the watcher a SIGINT. If the
watcher were dead you should have gotten an error on restart. I'm mildly
troubled you had to perform a stop/start sequence without any errors
logged.
-- 
John Dennis <jdennis at redhat.com>