Add workqueue to retry failed informer events in WorkerPoolSyncer #728
Description
If an informer event in the WorkerPoolSyncer fails, the event is never retried. In the case of delete is causes permanently stale workers / actors. We should use a workqueue to retry in the case of failures.
Activity
EItanya commentedon Aug 4, 2026
I have a slightly orthogonal, but related question. Why do we use controller-runtime managers for reconciling in the
atecontrollerand not for this?sfunkenhauser commentedon Aug 4, 2026
Good question :)
I'm not sure what the history is here, but ultimately we should move the WorkerPoolSyncer out of the apiserver (#729). This was meant to be a stop gap and should be the smallest change possible.
EItanya commentedon Aug 4, 2026
That’s fine, but I think we should align on a single k8s reconcile pattern if possible
sfunkenhauser commentedon Aug 5, 2026
Fixed by #674