alert that NEVER clear even if trigger is ok on trapper items

[ghislain]

hi,

I use a trigger to alert me if i have any errors appearing on my Raid arrays. Errors are sent to a trapper item. The trigger is:

count(/backuppalais.aqserver.com/mdadmevent,24h)>0


We are the 3rd of mars, the mdamevent is :

2023-02-28 17:51:06	RebuildFinished /dev/md/HOST mismatches found: 3456 (on raid level 1)
2023-02-28 17:50:33	RebuildFinished /dev/md/SWAP1
2023-02-28 17:50:33	Rebuild99 /dev/md/SWAP1
2023-02-28 17:50:33	RebuildStarted /dev/md/HOST

as you see the last event is 4 days old, but the alert is still here and do not want to clear.

i tried recovery expressions, changing time delay, and a bunch of other thing but those alerts never clear of.

I that a fact now that trapper item trigger never get refreshed unless new data is sent ? it was working some time ago but now it seems that the trigger is not at all working unless some new data come in.

best regards,
Ghislain.

[LenR]

I think your problem is that with SNMP trap (and log) items, it's difficult to have a "cleared" condition. Typically a nodata() is used in the trigger to clear after some amount of time, for example event at 17:50, then nodata(30m) to clear after 30 minutes. Have your notification sent within the nodata() time. Another option is to use manual close, the procedure being you needed to know that an event happened, you verify it's clear, then close the event.

Even when events resolve themselves, you sometimes need to be notified, for instance, when an array uses a hot spare, someone needs to replace a failed drive and set it up as an available spare for the next failure.

[ghislain]

well i think the real issue is that zabbix should NOT allow a count() or nodata() on a trapper item as they will never really be taken into account if there is no value that comes. At least nodata should be prevented.

best regards.

[cyber]

well i think the real issue is that zabbix should NOT allow a count() or nodata() on a trapper item as they will never really be taken into account if there is no value that comes. At least nodata should be prevented.

That is wrong approach... I can probably find many cases, where those functions are absolutely OK to use with trapper items.. Simplest one, just make sure your script somewhere is OK and returns values regularly.

I am convinced, that user should know the way how items/triggers work and adapt accordingly.

Trigger calculation based on arrived data has always been like that https://www.zabbix.com/documentation...lculation-time