Ayoub Mrini

Hello, I ended up implementing a little tool https://github.com/machine424/kube-hpa-scale-to-zero that simulates this for Kube clusters that cannot enable the feature gate, awaiting for the feature gate to go GA.

Hello, I'm sorry, but we do not use OPA and we do not want to rely on a mutating webhook for this (no problem if topolvm's mutations webhooks are bypassed,...

Hello @cupnes, > From my understanding, in the case of postgres-operator' delete protection, you can DELETE but you can undo it from the log. It seems not a good idea...

Hello @cupnes, Could you please share the reasons? We can help with the implementation and I don't think it'll be something hard to maintain. This is really critical, this is...

Hello @cupnes, Yes, I am aware of the failure policy, but these webhooks may get unregistered (ValidatingWebhookConfiguration deleted or misconfigured) by mistake or for any other reason. You don't want...

Hello @cupnes, Cool, thank you for the link, I'll watch it and see what I can do to help. This shows that deletion protection is a big deal not only...

Hello @llamerada-jp, Thanks for your fast reaction. I should have emphasized it more but the panic occurs when the pod to be scheduled has a priorityClassName (a high one, see...

Hello @llamerada-jp, You can use the shipped-with priority class `system-cluster-critical` for `huge-pod`.

Hello @llamerada-jp Yes, as I've already mentioned, it seems to be related to preemption, I reproduce this with: - node `N` with size of`S` lvm disk. - node `N` hosts...

If I really understand how tokens refreshing is implemented, new fetched tokens (at least for oidc) are not stored in `~/.kube/config`, that means, if a refresh occurs, `~/.kube/config` becomes obsolete....