topic Re: Deep monitoring for processes doesn't get auto-enabeled after regular updates in Container platforms

Deep monitoring for processes doesn't get auto-enabeled after regular updates

nanand3 — Tue, 07 Feb 2023 09:47:19 GMT

Hi team,

We have deployed OneAgent using Dynatrace Operator (latest) automatically for Kubernetes monitoring.

As a part of our regular updates, whenever a new image (AMI) is pulled and new K8 nodes spin up, its processes (Java, nginx etc.) do not get deep monitoring enabled automatically and a message shows up "Container needs to be restarted to get deep monitoring".

We have also tried to implement PriorityClass and also observed OneAgent process has an older age than others, still, it does not auto-inject into the process whenever we rotate the nodes.

We need yours expert suggestions here. Expectation is to get Dynatrace OneAgent getting injected automatically whenever any new nodes spins up.

Current Manager Server version: 1.256

@pahofmann

Thanks,

Nitesh Anand

Re: Deep monitoring for processes doesn't get auto-enabeled after regular updates

ChadTurner — Thu, 13 Apr 2023 11:47:06 GMT

Just to confirm, you don't have any process rules that are preventing the deep monitoring correct? And deep monitoring will work after you restart the process/container in particular? And this is a new container that is spun up? If that is the case i think that is a normal function when a new host if observed, a recycle of the processes/Pods on the node is required for some technologies.

Re: Deep monitoring for processes doesn't get auto-enabeled after regular updates

nanand3 — Thu, 13 Apr 2023 12:00:34 GMT

We moved to cloudNative approach of integrating oneagent to the applications that are on AWS.
And with this cloudNative we are not seeing the issue with process asking for a restart when new container is spun up.

Why there is a race condition with classic method compare to cloudNative?

Thanks,

Nitesh Anand

Re: Deep monitoring for processes doesn't get auto-enabeled after regular updates

Julius_Loman — Sun, 16 Apr 2023 15:43:52 GMT

@nanand3 yes, the race condition is a limitation of the classicFullStack deployment option and it is not present in cloudnativeFullStack. This is, I believe, mostly due to OneAgent pod start time (it takes some time to get the pod running).