Metric for Dynatrace Host Health Tile?

[mycelium]

Hello, 

I’m working on recreating a Dynatrace dashboard externally so it can be displayed on external screens that cannot directly access Dynatrace. I’m currently trying to replicate the Host Health tile from the Classic Dashboard, but I’m having trouble identifying the underlying data source.

So far, I’ve tried using the Problems API to retrieve hosts impacted by active problems. However, this approach seems to miss certain cases: for example, when a host shows as unhealthy (red) due to a network issue affecting one of its processes. In these situations, the host does not appear as an impacted entity in the Problems API results.

I’ve also explored the built-in host availability metrics, but they don’t seem to fully match the behavior of the Host Health tile.

Is there a specific metric or combination of metrics that this tile relies on?

 

[AntonPineiro]

Hi,

I think entities api tells you host status (running, offline, ...).

But maybe you can Grant anonymous access and they can see dashboard without Dynatrace access.

Best regards

[t_pawlak]

Yes, in my opinion, Entities API is a better direction than Problems API here.

the Host health tile shows the number of hosts in the environment and the number of hosts currently affected by problems. At the same time, Dynatrace exposes a host-level healthState (HEALTHY / UNHEALTHY) on the entity itself, and host health is determined more broadly based on detected anomalies and alerts. Because of that, it does not always match a simple impactedEntities lookup from the Problems API.

I would approach it like this:

Use Entities API v2 to get the host state, for example with:
type("HOST"),healthState("UNHEALTHY")
Treat builtin:host.availability.state only as a supporting signal for offline / unavailable / unmonitored cases, because that metric reflects host/OneAgent availability, not the full health evaluation behind the tile.
But treat builtin:host.availability.state only as a supporting signal for offline / unavailable / unmonitored cases, because that metric reflects host/OneAgent availability, not the full health evaluation behind the tile.

That also explains the case you described: a host may appear red because of a related anomaly or alert (for example a process or network issue), even if it does not show up as expected in the Problems API impacted entities output.

[mycelium]

Thank you for your suggestion. I’ve tried using the Entities API with type("HOST"),healthState("UNHEALTHY"), but the results still match what I get from filtering hosts impacted by problems. That is, the host related to the network issue is still missing. 

It seems that the Entities API does not evaluate host health in the same way that the tile does. The only workaround I've found is digging up the host via fetching the problem's properties. Is there a more direct/comprehensive method for retrieving the missing host?