Backups causes the "fcgid process manager" to die

[Andreas Svensson]

We recently got the backups up and running on our VPS, and they have been working as they should.

However, 02:00 this night apache2 stopped responding to requests, 02:00 is also the time we have scheduled for the backups.

Looking at the apache2-log I find this:

[Tue May 25 02:00:40 2010] [error] mod_fcgid: fcgid process manager died, restarting the server
[Tue May 25 02:00:43 2010] [notice] SIGHUP received. Attempting to restart

So obviously something went wrong around the time the backups started (but the backups were correctly made and transferred). Although, it could also have been around the time they had been packaged up and ready for transfer to our "personal repository" as they currently only amount to some 500MB which isn't unthinkable to have been completed in 1-2 minutes.

Tracing back I find that this has been occuring every night since we activated the backups. However with one important exception, this night it didn't restart. Looking at the past entries it is clear that apache2 has immediately restarted as it should, which may be an inconvenience but not really a big issue.

[Sun May 23 02:01:22 2010] [error] mod_fcgid: fcgid process manager died, restarting the server
[Sun May 23 02:01:23 2010] [notice] SIGHUP received. Attempting to restart
[Sun May 23 02:01:35 2010] [warn] No JkLogFile defined in httpd.conf. Using default /var/log/apache2/mod_jk.log
[Sun May 23 02:01:35 2010] [warn] No JkShmFile defined in httpd.conf. Using default /var/log/apache2/jk-runtime-status

However, apache2 becoming unavailable in the middle of the night is obviously a big issue for us. Our current backup plan consists of 4 domains that are (individually) set to start backup at 02:00 (total of about 500MB), they could obviously be spread out but they are all small enough so that the backups finishes quickly.

Is anyone aware of what could be causing the "fcgid process manager" to die?

 

[IgorG]

Seems that configuration file of Apache web-server was altered on this server. First of all make sure that you have User/Group directives which should define the system account to run the server with. Also make sure that MaxClients limit is not set to very high value - it usually has 256 value.
Check with

netstat -nap | grep :80 | wc -l

that you have not more than 1000 connections to apache and with

ps aux | grep httpd | wc -l

that there aren't many child apache processes. Try to clean the semaphores with the commands like following:

ipcs -s|awk '{print $2}'|grep '[1-9]'
ipcs -s|awk '{print $2}'|grep '[1-9]'|xargs ipcrm sem

Contact support team for more detailed investigation if it not help.