If you use native kubernetes with HA Jobmanager pointed at cloud storage, does it clean up old jobs? I'm running this setup and the job-result-store seems to just grow, even after erasing a FlinkDeployment and spinning it back up.

Our

job-result-store

does get truncated after some time. There should be a directory next to that one with the name of your job and that one contains the pointer to the last committed checkpoint and job - that one does get cleaned up when the FlinkDeployment is deleted. Not sure if it helps. (just for interest) We’ve been trying to figure out what the canonical way of migrating a job to a new kubernetes cluster is and we came up empty handed. Every operation seems to delete that pointer to the last checkpoint and without that there are only convoluted solutions. That’s why we investigated how those HA artifacts are affected/managed.