Nature and description of request:

A way to control the amount of events created by objects/namespaces, this could come in a multitude of ways: deleting events beyond a specific age. Deleting older events once an amount quota has been reached, throttling recurring events or anything that seems fit to handle this problem.

Why does the customer needs this:

Faulty applications and especially misconfigured cronjobs can create an abundance of events which can bloat up the etcd causing severe performance degradation. While a proper fix for these issues would be a patch or fixing the misconfiguration, administrators may only notice the event spam once it had already become a problem, which then would need to be resolved by fixing the fault in question and deleting the event keys using etcdctl (which is unsafe), this feature brings much needed resilience against the aforementioned situation, while not negating the problem it does makes sure the cluster cannot damage itself while trying to handle a user error.

The business requirement:

A way to make sure user provisioned workloads won't damage the performance (and eventually the availability) of the cluster more than they should. The solution would ideally be minimally limiting towards the user, this means relying on job quotas / disabling cronjobs for users is not ideal.

Affected components: k8s core/v1/events And it's controller