Open Miles-Garnsey opened 2 years ago
Can you provide some more details such as:
Sure.
Kubernetes distro: kind Storage class used (or even better a K8ssandraCluster manifest): Attached (I don't think storage class is relevant given this it is a secret which is taking a while to be created). Maybe kubelet logs if anything appears relevant: events attached. Going to each Kubelet seems like overkill, let me know what you're expecting to see?
I've measured the timings a bit more carefully, it looks like this issue slows the cluster start down by 1.5 - 2 minutes or so in most cases. It is noisy however, as we see the error occurring 7+ times.
This isn't huge next to the 14-15 minutes Cassandra frequently takes to come up, but it still seems like it might be a useful optimisation if we can move <DATACENTER_NAME>-keystore
secret creation forward in the reconciliation cycle to avoid it causing any pause at all.
What happened?
When spinning up new clusters I'm seeing them stuck with the following error for quite some time:
Did you expect to see something different?
Yes, I wouldn't expect encryption materials creation to be quite so slow, can we look at how this is sequenced so that it doesn't delay Cassandra pod creation?
┆Issue is synchronized with this Jira Story by Unito ┆Issue Number: K8OP-197