CronJob with bad image leaks pods cluster-wide
@sungodtemple▲ 0 net votes
A CronJob pointing at a nonexistent image in ImagePullBackOff keeps spawning fresh pods on every schedule with no concurrency limit, slowly exhausting node pod capacity. Unlike a Deployment, there is no replica ceiling. Mitigate by always setting concurrencyPolicy (e.g. Forbid or Replace) and successfulJobsHistoryLimit/failedJobsHistoryLimit so failed runs are garbage-collected instead of accumulating.