Triggers Disappear after Agent Stops Responding

[nnephew]

We have several hosts which become mostly unusable when a specific volume approaches capacity. We are monitoring the used % of the volume and have triggers at various thresholds. The problem we have experienced is that when we reaches the highest trigger threshold, the associated action is executed but the triggers then disappear from Monitoring -> Triggers and the host appears normal in the front end.

When we reach this threshold, the host pretty much stops responding and most commands fail but we would still like the triggers to show a problem in the Zabbix front end.

Any ideas?

[nick0909]

My guess is that the agent is actually not responding so there is no data. If a trigger is set to alert when something is at 90% (for example), and the agent returns nothing, then the evaluation will be (x>90%) which is false, because x is null. There is a nodata() check that you could add, so something like (x>90% OR X=nodata() = alert), or making an independent trigger for no data from the agent.

You can verify if this is the case before modifying your triggers by going to MONITORING > LATEST DATA and seeing what value is coming in from the agent at the time there should be an alert and there isn't. Then create a trigger based on that value.

[Heilig]

nnephew, all of you have described is true. The item passed to the status of "Not supported" and the trigger was gone (can not calculate the expression).

Do you have problems with certain items (you can not understand why they go to not supported state) or you surprised that if the server is down you do not get notified of problems with disk space, CPU usage, etc?

[nnephew]

My thought was that if we hit a threshold that causes a trigger to go to a "problem" state AND THEN the device becomes unresponsive before the trigger is resolved, the trigger should still be a problem. I would not expect the agent to send any additional data but I would expect existing triggers to remain in the "problem" state.

[Heilig]

In zabbix another logic - if the trigger expression can not be calculated, the trigger goes into unknown status.
In such a situation, it is necessary to use the maintenance mode for server or/and see into items and trigger expressions which this problem have.