Longhorn instance-manager upgrade in a Homelab setting
My kubernetes based homelab was recently borked by a myriad of issues plaguing the Longhorn release 1.11.0.
- https://github.com/longhorn/longhorn/issues/12578 - borked validating webhook
- https://github.com/longhorn/longhorn/issues/12573 - memory leak
Both requires manual bumping Longhorn to a hotfix image.
When bumping the instance-manager, existing volumes still using the old instance-manager will persist until volume is detached.
The official Longhorn document suggested scaling the workload down to achieve that outcome.
Personally I find that suggestion flawed, while I do have ArgoCD which helps reconcile the configuration, I don't want to manually update the deployment count.
Turns out, I can manage the same result through node drain - which forces workload to be moved to a different node.
Granted - I don't have a fool-proof disruption budget set, so some workloads are interrupted; and all of my volumes are RWX. But I'll take this much simpler approach to manually scaling down/up each workload.
Comments
Post a Comment