Problem with agent.ping and nodatata()

[tobster]

Hello forum

we currently have a weird issue with our monitoring. Every sunday we get an alert which gets fired far too late according to the config.

We have an agent ping item which is checked every 30s with a nodata(300) trigger. On sunday morning the zabbix agent of one host goes down. Around 0710

So in my understanding of Zabbix the alert should be fired on 0715 --> 300 Seconds no data

In fact the alert comes after 1200. I checked all the maintenance windows and only one applies for the host in question. From 0800 to 1000 on sunday.

But even with this window the alert should be fired before 0800. Even if we assume that the time when the agent goes south is covered by the maint window then the alert should be triggered 5min after the window ends at 1000.

Any idea what going wrong here? Other nodata() triggers work fine. It seems the problem occurs only if a maintanace window is involved

We use Zabbix server v2.4.6 on a Centos 6

[tobster]

I think I found the root cause
According to the logs not only the zabbix agent was down but the whole host where the agent runs on. As the host was unavailable the Zabbix Server disabled the checks on the whole host after 3x15s not reaching the host.
That explains why the nodata(300) trigger fired after the host was up again.

So whats the best way to catch such an issue? Lowering the nodata value to be below the time zabbix server waits until de-activate a non-up server? For example nodata(30)
Or going even lower and set the item interval to 10s and then nodata(10)?

[batchenr]

i think lower the item interval and then it checks more frequently and refreshes faster.