non-existent configuration file(s): /etc/exim.conf

[Bazer]

2006-09-18 19:17:05 non-existent configuration file(s): /etc/exim.conf

Exim recently started failing and this was the only error in 'exim_paniclog' that showed up but not always. Sometimes it fails without any error whatsoever.

The way it fails is curious because it stops listening on the standard SMTP port but keeps the secure, encrypted port open.

I wasn't able to track the cause and thought if anyone else had a similliar problem.

 

[mctDarren]

Is /etc/exim.conf there? Weird if it only happens sporadically.

 

[Bazer]

exim.conf is in place of course. It looks like cPanel is updating exim.conf and is being caught by exim trying to read it or cPanel failed to udpate the file alltogether.

 

[chirpy]

Have you checked that you're not running anything via cron that could be touching exim.conf or other exim related files, such as running /scripts/mailperm which I've seen cause havoc with /etc/localdomains

 

[Bazer]

I'm pretty sure none of my scripts I've added don't touch exim.conf. I also haven't added any of the cPanel scripts from /scripts.

I applied a temporary fix by configuring monit to restart exim every time it fails to connect with it but this shouldn't be happening.

It happens under normal load and at a time unrelated to any of the relevant cron jobs.

 

[chirpy]

I've seen monit screw up cPanel servers, as does SIM, since they're duplicating the work that chkservd does which can lead to race conditions.

 

[Bazer]

I have installed monit in the first place because services (pop3 especially) were failing and cPanel didn't bring them up.

No notification from chkservd, no mention of failures in it's log, nothing.
Even now, after problems with exim, chkservd does not have anything in it's logs but:

Service check ....cpsrvd [+]...Done

Out of curiosity: what are the consequences of these chkservd race conditions?

 

[acenetryan]

We don't use monit or SIM but with two applications both trying to restart a service, you can get stuck into a loop of sorts.

For example, let's say a service fails and monit notices first.

Monit trys to restart the service and issues the command.

Before the service can be started by monit, chkservd notices the service is down.

The service comes up because of monit, but chkservd's last status said it was down so it trys to restart. As a result, the service comes down during the restart.

Monit notices the service is down, repeat.

Both applications are racing to restart the service.

This obviously is dependent on the timing of the service checks (race condition), but could be a problem.

 

[chirpy]

No notification from chkservd, no mention of failures in it's log, nothing.
Even now, after problems with exim, chkservd does not have anything in it's logs but:

Service check ....cpsrvd [+]...Done

That suggests that you have not fully configured WHM > Service Manager

 

[Bazer]

That suggests that you have not fully configured WHM > Service Manager

Ah, so that's what the "monitor" checkboxes are for!
Nothing beats hiding a feature under an undocumented checkbox, I always say! Thank you cPanel!

Rants aside, thanks for your help Chirpy.

rking: Thanks. I haven't experienced something like you describe with monit because I always set up a timeout for every monitored service. If any of the monitored services fail, for example 5 times in a row in a time span of 5 minutes, monit will stop trying to start that service.