This message was deleted.

I solved this.

Cool .. you should also seriously consider using the Helm Chart

But I preferred to use kubectl apply -f longhorn.yaml. Longhorn Chart is sub-chart in my Helm APP.

I am curious what the root cause was and how you solved it?

I found the reason of the problem is about backing image. The Longhorn cluster has a node that is unavailable. But the VM image stored by the node. The YAML of the engine image record the vm's image. Now when you updated the longhorn, you would found the longhorn manager log displayed the image don't found.

your Helm App … so you package everything into one app? Sounds hairy and unnecessarily convoluted.