Notification when Status Information (not status) changes

[pato]

Hi all,

I've got a service check which reads from a log4net database and reports back on matching rows indicating a particular error. A column from the row (job number) is returned in the notification as status information ($SERVICEOUTPUT$). I encountered behaviour today in which one row is returned, triggering a CRITICAL status change and a notification. However, immediately after resolving this, another row is generated from a new error. The status information changes, as it's a new job number, but the status is still CRITICAL, so no new notification is issued. Aside from configuring periodic repeat notifications, can I trigger additional notifications on a status information change?

[Box293]

No, because it is the state of the service that triggers when it changes from OK / WARNING / CRITICAL / UNKNOWN. Just because the wording of the status changes does not make any difference if the service is already in a hard critical state. When it sends it's next notification the updated text will be sent in the notification.

[eloyd]

You can cheat and make the service trigger an event handler, pass the state and status to the event handler which saves the last copy of the status information somewhere (temp file) and compares the new status to the old status. If it's not the same, then send an email. This way, Nagios still does the scheduling, but you're doing an end run around the notification logic.

[pato]

You can cheat and make the service trigger an event handler, pass the state and status to the event handler which saves the last copy of the status information somewhere (temp file) and compares the new status to the old status. If it's not the same, then send an email. This way, Nagios still does the scheduling, but you're doing an end run around the notification logic.

That's brilliant and looks like exactly what I need. Thank you so much!

[eloyd]

Well, I wouldn't say it's "brilliant" exactly. "Stunning," maybe.

Let me know if you need help.

[sreinhardt]

You could also set the service to have isvolatile on. However it should be noted this will send on every warning\critical result not just when the status messages change. Otherwise I'd agree that eloyd's idea sounds like a pretty killer way of handling it.

[eloyd]

No, "killer" would be to have the event handler figure out if it should notify, and then send the command back to the Nagios command pipe to do the notification. That would be killer.

[abrist]

Proper "Killer" would entail the Pony Express, electrodes, Clint Eastwood, and lasers.