Feature request: Latching status

[sickel]

I would love to have the possiblilty to have a "latching" status. That is, when the status is set to anything else than OK, it stays that way even if the logged entitiy itself goes back to a normal status.

My use case: Some of the systems I am monitoring have an integrated Door Open alarm. I can easily check that in nagios, but I want to know not only if a door is open, but also if a door has been open. I am storing state information, but it would be easier to see the alarm directly in my nagios dashboard.

I am sure this feature also could be useful for logging various "should not happen" cases that may indicate problems even if the problem itself went away,

[everwatch]

Two solutions.

1. Put an event handler on the service that disables the service on a Door Open state. This way, it simply won't check again until someone manually reactivates the check.

2. Send a notification. Let the humans know something happened.

[Rfferrao87]

Isn't that merely a passive check with the freshness interval unset?

[everwatch]

The passive check with no freshness interval would work, but only if the device can send Nagios the information (like an SNMP trap). If Nagios has to perform an active check to get the information, then there is no way to "latch" the state and keep it that way unless you disable further checks of the service after it hits a warning/critical state.

That's why my recommendation is for an event handler to fire off that uses the Nagios API (or CMD file) to disable further checks of the service if the service is in a non-OK state. But a passive check could work, since we're talking about security events, which are hard to poll for, but easy to submit passive results.

Personally, I think the real solution to "I want to know if an event has occurred, even if it's not currently occurring," is to send a notification. That is, after all, what they're for. Plus, it requires zero extra work to make a janky latch state.