A cronjob configured with a concurrencyPolicy of Allow has the potential for overloading a cluster if a cronjob-managed pod cannot manage to finish in time. The impact we are occasionally seeing is that job-pods created from subsequent cron intervals pile up over time, sometimes leading to hundreds or thousands of pods that stay pending.
We should implement a check that triggers a warning for Allow cronjobs and suggests to use one of the safer alternative modes Forbid or Replace.
A cronjob configured with a
concurrencyPolicy
ofAllow
has the potential for overloading a cluster if a cronjob-managed pod cannot manage to finish in time. The impact we are occasionally seeing is that job-pods created from subsequent cron intervals pile up over time, sometimes leading to hundreds or thousands of pods that stay pending.We should implement a check that triggers a warning for
Allow
cronjobs and suggests to use one of the safer alternative modesForbid
orReplace
.