HTTP server was dead

Pierre-Marie Baty · **HTTP server was dead -** 05-03-2004

Hi admins and regulars,

This morning I had to log in via SSH in order to restart the web server, which died during the night. I checked with "who" to ensure nobody was working on it, and it seems the service died for no reason.

There's something to investigate here (ahh, if only days were 48h instead of 24 9_9 )

Onno Kreuzinger · **Re: HTTP server was dead -** 05-03-2004

i checked, it died on a SIGUSR1, a gracefull restart, triggert by logrotate...
i could not reproduce it, so wait until tomorrow morning ...
i activated allmessages to be logged...

cheers

(*rolf* me is concentrated on what me can do make me do a good job)
[b.t.w. "I" spoken equals german "egg", so i seem to have a lot of ...]

botmeister · **Re: HTTP server was dead -** 05-03-2004

Is there a way to have such services automatically restart on fail? I can do this in windows, so i assume it can also be done in Linux.

Onno Kreuzinger · **Re: HTTP server was dead -** 05-03-2004

hmm that is a generic statement, i can't realy answer w/o writing a heck of stuff.
so summed up, yes of course, but why? apache should not fail, its apache and a sigusr1 is a feature that is used every day at 6:00 by the logrotate service; and for the win stuff, well i could make a list as long as i want with software not running for ever :-)
regardles what is done (and payed).
i'm no linux evangelist and proud of my win9x systems in production (CTI) and my own w2k desktop i work on for 3+ years w/o reinstall or leaving the gui.

i will get that fixed, or at least trace it down. the last resort is daemon tools.

botmeister · **Re: HTTP server was dead -** 06-03-2004

Of course Apache should not fail and the problem should be found and fixed, but since services do on occasion fail, I think it would be nice to have it restart automatically to keep the down time to a minimum. I expected such a function would be built into the Linux server as a default safety feature. Why are you saying "last resort" ?

Pierre-Marie Baty · **Re: HTTP server was dead -** 06-03-2004

I sorta agree with botmeister... after all it's the job of the system admin to check the logs periodically ?

Terran · **Re: HTTP server was dead -** 06-03-2004

There is a lot that can be done at this field, e.g. use BigBrother or Nagios to check for certain system parameters and processes and alert the administrators if something goes wrong.
Or use simple scripts to check and restart processes using cron.

But as memed already said, apache is not known to crash at all! I've plenty of them running since years and none of them ever crashed without a humans intervention

(e.g. wrong configuration). Not even a failed CGI program can cause a total crash as the main apache process never starts them directly.

The only time I ever had a real problem with apache was when we started using apache 2.0, php and postnuke. This combination caused a system crash with big file downloads...

Onno Kreuzinger · **Re: HTTP server was dead -** 06-03-2004

"last resort", because daemon tools are usually used in big email server systems based on qmail (multi million mails per day), and my professional experience with apache 1.3 (the one we are using) realy indicate that it does not crash .

but i just made a simple watchdog, which checks every 15 mins and restarts apache in case it does not respond with the http code 200.
the logfile entry will be "Apache responded with code 200, all ok." when it did not restart apache. if it does apache will make log entries.

cheers

/edit
and of course there is allready a webmin service monitor running, which also should restart the server, but i could not find a trace of that working or not working, so i had to "wait" until it should have worked, now i found out that it faild bacause inetd was in the watch list. that service is not used, but xinetd so now i expect it to work. any suggestions when to test :-)
edit/