Orchestra CPU consumption

[karbyshevds]

Hi there. Searching for help because already read all docs and even a source code a bit but can't find any clue.
I'm experiencing with high CPU consumption of openunison-orchestra POD version 1.0.40. Deployed to k8s cluster and with 2 CPUs limit and it constantly throttling.

No any any suficient load from users is provided but I it produces hundreds of such messages in log in a minute.

[2025-11-26 15:08:34,175][Thread-17] INFO  K8sWatcher - watching https://kubernetes.default.svc/apis/openunison.tremolo.io/v1/namespaces/openunison/groupmetadatas?watch=true&timeoutSecond=25&allowWatchBookmarks=true
[2025-11-26 15:08:34,176][Thread-16] INFO  K8sWatcher - watching https://kubernetes.default.svc/apis/openunison.tremolo.io/v1/namespaces/openunison/groupmetadatas?watch=true&timeoutSecond=25&allowWatchBookmarks=true
[2025-11-26 15:08:34,176][Thread-24] INFO  K8sWatcher - watching https://kubernetes.default.svc/apis/openunison.tremolo.io/v1/namespaces/openunison/trusts?watch=true&timeoutSecond=25&allowWatchBookmarks=true
[2025-11-26 15:08:34,178][Thread-14] INFO  K8sWatcher - watching https://kubernetes.default.svc/apis/openunison.tremolo.io/v1/namespaces/openunison/customazs?watch=true&timeoutSecond=25&allowWatchBookmarks=true
[2025-11-26 15:08:34,183][Thread-17] INFO  K8sWatcher - watching https://kubernetes.default.svc/apis/openunison.tremolo.io/v1/namespaces/openunison/groupmetadatas?watch=true&timeoutSecond=25&allowWatchBookmarks=true
[2025-11-26 15:08:34,183][Thread-24] INFO  K8sWatcher - watching https://kubernetes.default.svc/apis/openunison.tremolo.io/v1/namespaces/openunison/trusts?watch=true&timeoutSecond=25&allowWatchBookmarks=true

As I understand it's a CRD watching threads but is it ok that they spawns so often? Is there any way to reduce the number of threads or maybe checking requency?

Thank you in advance.

[mlbiam]

To answer your question about the number of threads: each OpenUnison type has its own watch, and its own thread. For instance, in the logs you provided trusts uses thread 24. I doubt this is the reason for your CPU issues.

If you want to see what's doing work, you can take a thread dump by execing into the orchestra pod:

k exec -ti openunison-orchestra-5b899c85c7-vmcmq -n openunison -- sh -c 'kill -3 $(pgrep -f java)'

then look in the std out logs, you'll see a thread dump. pull that out of the logs and you can use one of the free thread dump analyzers to see where the issue is.

Also, 1.0.40 was released in March 2024, you should consider getting up to 1.0.43.