Vlastimil Holer

Nice, if it's hard to convince your colleagues, it might help that I could contribute this feature (add lvm tools into an image, modify CSI to **conditionally** use local tools...

@remidinishanth We are using AWS EKS with Bottlerocket OS, which is a limited system focused only on running Kubernetes without common tools (like LVM). Kubelet is not running in container...

@ushitora-anqou Thank you, it's working for me at the end, I really appreaciate it! I only found an issue with Helm chart not indenting properly the `lvmd.additionalLVMDYamlContent` while creating the...

I'm not sure know which "rest pods" you mean. I'm using the portable OSDs/Mons (backed by PVC), so they are migrated to other nodes by Kubernetes even when operator is...

IDK about crashcollector, I'm not using this one. But CSI pods are managed by a daemonset and Kubernetes ensures that the pods run or don't run on nodes where they...

The complain is about the Rook Ceph exporters, I don't see the issue with any other component in our environment. The CSI provisioners are deployments, but not explicitly scheduled on...

Yes, it's very edge case. NOTE: If one disables the monitoring on running cluster, it deletes properly all Deployments for rook-ceph-exporters, even for nodes which don't exist anymore (and are...

> If you do change anything from the old node like its spec, you can see the exporter pod will get deleted because rook will reconcile for that node. That...

I can't easily try, nodes are dynamic with dynamic (but IP-based, e.g. `ip-A-B-C-D.eu-central-1.compute.internal`) hostnames. Once node is gone, it's deleted from Kubernetes and there is low probability for me that...

Still valid.