Indeed, function check_escalation() in src/zabbix_server/escalator/escalator.c does not recheck whether host is in maintenance and neither does it check other action conditions like "Trigger severity", etc.:
/****************************************************************************** * * * Function: check_escalation * * * * Purpose: check whether the escalation is still relevant (items, triggers, * * hosts, and actions are still present and were not disabled) * ...
We shall discuss whether this is by design or not.
that is by design. "maintenance does not stop active escalations", citing from the zabbix training 
Thanks, richlv! Indeed, it is already documented at https://www.zabbix.com/documentation/2.4/manual/config/notifications/action/escalations .
Hi,
Thanks very much for the answer.
Now we have a few months of zabbix usage and we have acquired a fairly good expertise.
Still, the way manteinance is handled in regards to how the actions are evaluated, is annoying.
Last night again, we had the following :
- Trigger switched to Problem
- Action sent an e-mail
- Then the maintenance started (it was the end of the working day, maintenance is setup everyday from 7:00pm to 7:30am with no data collection)
- all night long, the action escalation step incremented and sent e-mails <-- during maintenance period, the trigger has no value (it isn't Problem, or OK), so it doesn't make any sense to continue escalation. Besides, there is a condition "maintenance is not mainenance" in the action, which isn't respected when the host is in maintenance
When you setup a maintenance, you don't want the action to continue sending e-mails.
Would it be possible to upgrade this case to feature/improvement request ?
There is another similar request at ZBXNEXT-1565. Maybe we can consider this issue as a duplicate of that.
I agree, the case seems similar.
I will follow-up with the other case.