[Pkg-mailman-hackers] Bug#505638: Bug#505638: Bug#505638: mailman: Mailman dies on log rotation

[Jérôme]

Le 2015-01-11 14:24, Thijs Kinkhorst a écrit :

>> Some solutions are proposed here :
>>
>> http://wiki.list.org/pages/viewpage.action?pageId=4030580
> 
> Well, that page contains a collection of approaches to do it, and we
> implement many things that are mentioned on the page.
> 
> The observation in this bug log that it may be a race condition because
> two snippets are executed at the same time may be relevant. Do you also
> experience this bug only on Sundays?

I'm sorry, I can't tell. It happens so rarely. And my server only hosts
a few low-traffic lists, so it can take a few days until I realize
mailman is down.

Indeed, on the 11th of January (last Sunday), I have two qrunner logs :

-rw-rw-r-- 1 list list       212 janv. 11 04:53 qrunner.5.gz
-rw-rw-r-- 1 list list       202 janv. 11 04:53 qrunner.6.gz

It looks like the rotation was made twice :

qrunner.5 :

Jan 11 04:53:21 2015 (14960) ArchRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14968) VirginRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14962) CommandRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14964) NewsRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14969) RetryRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14961) BounceRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14963) IncomingRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14965) OutgoingRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14959) Master watcher caught SIGHUP.  Re-opening
log files.
Jan 11 04:53:21 2015 (14960) ArchRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14961) BounceRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14963) IncomingRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14965) OutgoingRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14968) VirginRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14964) NewsRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14969) RetryRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14962) CommandRunner qrunner caught SIGHUP. 
Reopening logs.

qrunner.6 :

Jan 10 04:52:41 2015 (14962) CommandRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 10 04:52:41 2015 (14963) IncomingRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 10 04:52:41 2015 (14965) OutgoingRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 10 04:52:41 2015 (14968) VirginRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 10 04:52:41 2015 (14959) Master watcher caught SIGHUP.  Re-opening
log files.
Jan 10 04:52:41 2015 (14960) ArchRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 10 04:52:41 2015 (14961) BounceRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 10 04:52:41 2015 (14969) RetryRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 10 04:52:41 2015 (14964) NewsRunner qrunner caught SIGHUP. 
Reopening logs.
Jan 11 04:53:21 2015 (14959) Master watcher caught SIGHUP.  Re-opening
log files.

As a workaround, I just modified /etc/logrotate/mailman into

/var/log/mailman/vette /var/log/mailman/error /var/log/mailman/bounce
/var/log/mailman/digest /var/log/mailman/subscribe /var/log/mailman/post
/var/log/mailman/qrunner /var/log/mailman/fromusenet
/var/log/mailman/locks /var/log/mailman/smtp
/var/log/mailman/smtp-failure {
        su root list
        weekly
        missingok
        create 0664 list list
        rotate 4
        compress
        delaycompress
        sharedscripts
        postrotate
            [ -f '/var/run/mailman/mailman.pid' ] &&
/usr/lib/mailman/bin/mailmanctl -q reopen || exit 0
        endscript
}

/var/log/mailman/mischief {
        su root list
        monthly
        missingok
        create 0664 list www-data
        rotate 4
        compress
        delaycompress
        sharedscripts
        postrotate
            [ -f '/var/run/mailman/mailman.pid' ] &&
/usr/lib/mailman/bin/mailmanctl -q reopen || exit 0
        endscript
}

I don't mind the refresh frequency.

I'll see what happens. But since the error is so rare, not seeing
anything won't mean it is "solved".

-- 
Jérôme