In another thread, Alexei suggested using cron scripts and zabbix_send to do checks that take long instead of having the agent run them.
I have a particular check (an adapation of http://boinkor.net/debian-security-updates.html) that will take a pretty long time - 60 seconds or more per check. If I want to deploy that on 60 or more hosts, I need to edit 60 or more cron tabs - ugh.
That's where zabbix could come in: it already runs commands periodically by querying the zabbix agent. The only problem with running slow checks seems to be that the zabbix-suckerd waits until the command is finished.
My suggestion is that the zabbix-suckerd should be able to tell the agent to run the program and move on to the next check; the agent should just report that the check was started. The check program will then use zabbix_send to report the status back as soon as it's finished.
Does that make sense to you?
I have a particular check (an adapation of http://boinkor.net/debian-security-updates.html) that will take a pretty long time - 60 seconds or more per check. If I want to deploy that on 60 or more hosts, I need to edit 60 or more cron tabs - ugh.
That's where zabbix could come in: it already runs commands periodically by querying the zabbix agent. The only problem with running slow checks seems to be that the zabbix-suckerd waits until the command is finished.
My suggestion is that the zabbix-suckerd should be able to tell the agent to run the program and move on to the next check; the agent should just report that the check was started. The check program will then use zabbix_send to report the status back as soon as it's finished.
Does that make sense to you?
Comment