One restart action for multiple events

[burn1024]

I'm using Zabbix to monitor a service (say, httpd). There are multiple check configured (process running, tcp port open, etc). I want Zabbix to restart the service should one of the checks fail. The action condition looks like this:

(A or B) and C
(A) 	Trigger = "Apache is not responding on server Template_Linux_server"
(B) 	Trigger = "Apache process is not running on server Template_Linux_server"
(C) 	Trigger value = "PROBLEM"

The problem is that when the process is down, every single check triggers a restart, so I end up with multiple restart actions, which I obviously don't want. One possible solution would be to script around it in the actions themselves, but it is not too elegant and with number of checks rising this would become really messy. I'd like to keep that logic in Zabbix. Any advice on how to work around that?

[anpat89]

What are your "Action operations" regarding this Action. I'm guessing Remote Command but what is the command?
Regards,
Andrew

[burn1024]

yes, remote command. "sudo service xxx restart"

[anpat89]

As you have that 1 command to run on all triggers, it seems that if the service is not responding first of all, it restarts.

In the middle of the restart, the service is off, so the 2nd trigger is true, hence another restart and it will loop like this.

Perhaps a way around it is to create 2 actions.
The first entails "Apache is not responding on server Template_Linux_server".
The action is as you have it: "sudo service xxx restart".

The 2nd action is for: "Apache process is not running on server Template_Linux_server"
rather then restart, make it start: "sudo service xxx start".

In both instances you can have the value = "PROBLEM".
Hopefully this fixes the loop that this is causing.
Kind Regards,
Andrew.

[burn1024]

I don't have a loop. It works just fine, except that I have 2 restarts instead of one, which I don't want. What you suggest would make it happen in 50% of cases, depending on the trigger that has fired first.

[anpat89]

It would make it restart if it is not responding.
It would make it start if it is not on.

This will not create the same issue and should replicate the problem. Did you test it out?
Cheers,
Andrew

[burn1024]

No, I didn't. The action is triggered twice now. That means both actions will be triggered with suggested change. So I will end up either with
- start + restart
or
- restart + start(failed)
neither of which I want. Anyway, even if it did work, like I said, the number of checks will grow. And most of them will require a start/restart. That just doesn't fit in here.

[anpat89]

Perhaps it's best I leave you to sort out your own solution as none of my advice is being heard.

If you just have the one action: not responding.
would that fix the problem? if a service is off isn't it also not responding? My expertise on the issue is limited.
I tried the best I could to help.
Regards,
Andrew.

[burn1024]

Perhaps you should try to better understand the problem before offering an advice. The checks could be (and will be) anything - tcp connection failed, http response code is wrong, login failed, certain page(s) does not return expected output, etc. I cannot expect one check to cover all functionality, although if such check existed, it would indeed help in this case.