I think we have been generally expecting the ttlSecondsAfterFinished setting in YAML job spec to clean up jobs eventually, but I just noticed this is not enabled by default.
I enabled it in our UVic cluster.
We have several people enabling this in their clusters at CERN as well - including some of the recent deployments we manage directly.
Probably adding it by default on new cluster deployments, but we need to follow this up upstream as it’s been in alpha for a long time and constantly missing the move to beta in the last releases:
The stopper is the missing work to extend this to Pods before moving to beta, but that work needs love:
If we plan (as we do now) to rely on this we could try to help out? If anyone is willing to pick this up leave a mention here, we’ll look internally as well if not.