When

up

fails because a K8s job fails, when running

pulumi up

again, shouldn’t it try to replace the job?

It doesn’t work that way currently. Can you open an issue with details if you have a case where that doesn’t make sense?

Sure