Stefan Siegel <debian <at> sdas.de> writes: > > It happened again. Of course something else must be wrong making an > apache child not exit on SIGTERM, but it should not make apache die > completely every sunday morning. I will now > change /etc/logrotate.d/apache2 to use "apache2ctl graceful". > > Please consider fixing the restart part of the initscript or doing a > graceful restart in the logrotate config, because apache is > unreliable as it is now. > >
We are having the exact same problem. It seems that instead of relying on a timer (sleep), it should test to see if the process(es) are dead before trying to start it again. Apparently 10 seconds is just not enough time for the children to clean up and term. It probably shouldn't take them 10 seconds, but it doesn't seem like we should rely on time. Is there a reason why we can't just use 'apachectl restart' instead of 'apachectl stop' and 'apachectl start' in the restart function of the initscript? It seems the problem is in the initscript, not in the logrotate config. I think we need to fix this in the initscript properly no matter what we end up doing with the logrotate config (restart vs. graceful restart). -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
