Experience with GKE Autopilot Observability?

[r_weber]

Hello community,

I'd like to get some feedback regarding GKE Autopilot and Dynatrace K8s monitoring (esp. for the new K8s integration, logs, events, metrics).

As far as I can see now, we are only able tu use application-only monitoring with autopilot and also the K8s API isn't available, which means there is no use of the new K8s UI in Dynatrace. For logs and events you have to use fluentbit and forward these to Dynatrace (limitations) and the whole user experience in the Dynatrace UI is impacted.

Is my perception right? Is it worth using GKE Autopilot, considering that the Dynatrace integration/experience is lacking a lot of the nice features? Any feedback from other users with experience with GKE Autopilot & Dynatrace would be really appreciated!

Cheers!

[r_weber]

Adding a few more specific questions while working with the first GKE Autopilot integration...

Logging
I noticed that forwarding logs from GKE Autopilot clusters should be done via fluentbit. This seems to not cover any process/pod logs out of the box and/or those logs are not associated with the actual process/service.

Events
No (k8s) events are associated with application workloads/pods. I guess that is the abstraction of GKE Autopilot, loosing this kind of visibility? 

New Kubernetes App

Should hte GKE Autopilot integration also work with the new Kubernetes App? In our first attempt we can only see the application-only pods as hosts, their processes and services but no "highlevele" information about the cluster in the k8s app (well that is exactly what Autopilot is abstracting from the user, so kind of expected).

Maybe Mr. Kubernetes @alois_mayr can shine some light on it 😉