11/13/2022, 8:39 PM
it was just a temporary lab cluster so I deleted it in the end and recreated the pool with more CPU and memory nodes. Would be great to understand in the future how to remove those instance-manager pods


11/14/2022, 10:46 AM
I guess you mean the old instance manager pods? If there is no replica running (
means the related volume is detached) insides a non-default instance-manager-r pod, the pod will be cleaned up automatically. Similar logic applied to the instance-manager-e pods, but a running engine means the related volume is attached on the node. In other words, after the Longhorn system upgrade, Longhorn will directly deploy a set of instance manager pod as the default one. But at this moment, all attached volumes have running engines/replicas in the old instance managers. If you do live upgrade for the running volumes, the replicas will be migrated to the new instance-manager-r pods then Longhorn can clean up the old instance-manager-r pods automatically. But if you want to clean up the old instance-manager-e pods, you need to detach the old running volumes from the related nodes. In the future, Longhorn may merge the instance-manager-r and instance-manager-e pod on a node into one single pod. And we will consider the improvement of the engine upgrade part.