Trigger status

[elkor]

I've given some thought to this, and I've worked with the current method for a while, and I just don't think it's the right way to do things.

Many (myself included) have been vexed by the trigger "unknown" status for quite some time and I think we need to make a change to make it a little more logical.

Zabbix sets all the trigger status flags to unknown when it looses (or never had) communication with an agent, which is fine and what should be done. But the practice of keeping that value until the trigger changes is just causing more problems than it's worth. For instance if you (or your client) looks at a chart showing trigger statuses over time it has the potential of showing a large percentage of "unknown"

To correct this without a kludge I propose the following: we change the logic so that not only does Zabbix set all the trigger values to false upon loss of communication, but upon resumption of those commuications it re-evaluates all the trigger statuses. Alternatively, if it's easier, set all the triggers to false and let nature take it's course with the normal evaluation procedure; even that would be a little better than what it is now.

[James Wells]

Greetings,

Makes a lot of sense.

The way I see it, the Zabbix server, on start-up, would scan through the DB and set all triggers to unknown, and scheduled all items associated with those triggers for a near future check. Once the items are checked, the triggers are updated and remain in their new status until such time as the server is stopped or the actual status is changed again based on item evaluation.

The real problem though comes in the item evaluation. At present, when an item is checked, the Zabbix server records the value then performs a trigger check against that item value. Unless told to do so, via trigger dependencies, the item and trigger are evaluated singularly, meaning that the server does not take into account any other items or triggers when determining the status of the current trigger.

One way around this, at least for Zabbix agents, in non-active mode, is to depend all triggers on the pseudo item status. This way, if the server loses communication with the agent, it will not evaluate any other triggers, thereby leaving them at their current value.

Unfortunately, I am not really sure how to make something like this work for Zabbix agents in active mode, or for agentless checks.