Checkmk 2.5 is out!

[Niklas (Niklas Pulina)]

Hi GarthH,

Thank you for sharing your experiences. At the moment, you can only create one new special agent rule per each “query block” in the custom graph editor; as our special agents follow a “first matching rule wins” approach, you can therefore only transfer one “query block” into an active special agent rule per each host.

This limitation however only applies to the “quick path” from inside the custom graph editor. If you want to add additional queries to a special agent rule, this is still possible. For this, please open the “active” (i.e., matching) rule from the rule set “Other integrations > Metric backend (custom query) > Metric backend (custom query)”, and add another query via “Add new entry”. With this approach, you can associate multiple queries with one and the same special agent rule, enabling you to have more than one query result converted into Checkmk services.

Please be assured that we’re already looking into making alerting on OpenTelemetry metrics more convenient in future iterations of Checkmk. We’re on it.

Being one of the product managers working on these topics, I’d be super interested in further details about your experience. Maybe we could have a meeting anytime soon. I’ll contact you via PM.

Best, Niklas

[GarthH (Garth Horsburgh)]

I wasnt using the quick create, I was making them manually but I misunderstood and was trying to make a new rule for each metric, but I see now, I just keep adding metrics to the one rule for that host. I didnt see the add new entry button thanks!

[GarthH (Garth Horsburgh)]

If I enable this on the host under agents,

It appears to add every single thing that it can find as services, many of which have multiple metrics and it works

But If I try to do that via the special agent rule, you cant have multiple metrics, you have to choose a specific one

How is the host agent method doing it and can I recreate some of them manually?

[martin.hirschvogel]

In this case, as you want system.memory.utilization, I assume filtering for data point attribute state:used would be the right choice. That results in one metric.
What would have been the expected outcome for you? One service per metric instead of the current “UNKN”?