Process Group splitting in k8s is based on...?

[JoSk]

Hi,

we have to merge some processes at a customer k8s system BUT we still want to have them split based on namespace or cluster.

Right now we have process detection rules that look for k8s container name (but these are the same for both PROD and TEST clusters) and then use container name or base pod name depending on what we want.

Now I wonder if processes from a different cluster would also fall into these rules once the processes are restarted. For PROD this could be a problem as restarts are only possible in rare and late maintenance windows. I guess this depends on if process splitting is also done on cluster level or not. I somehow think not (but correct me if I'm wrong) and if workload name doesn't matter or is the same anyway we would get all processes from PROD and TEST into 1 Process Group which would be bad.

Can we simply create a second Detection Rule in addition to our existing one (so both are executed) or would only 1 work correctly and is the order important?

[Dant3 · Accepted solution]

The rules for Splitting are defined in "Cloud application and workload detection" in Process and Containers.

There you can define what data is used to set apart different workloads. In addition to this, you have to consider HostGroups in case of different clusters. 

If Both clusters are in diff HG, the workloads and eventual services will always split away. Even if the the whole dataset have the same name, they are in different HG. 

If Test and Prod are in different Clusters, having different HG is the way to go. In fact I would also recomend having a Lower Environment for Non production components.