This message was deleted.

I think all volumes still need to be detached. This is because the tolerance setting applies to multiple components, such as the manager, UI, and driver deployer. And these components will fail to restart because there are still volumes running. Ref. https://longhorn.io/docs/1.3.2/advanced-resources/deploy/taint-toleration/#setting-up-taints-and-tolerations-after-longhorn-has-been-installed

Right… What’s the reason everything has to restart? If those components are already running on nodes that don’t have a specific taint I’m trying to get around they shouldn’t need to restart? I have a node that runs a different architecture so it has a NoSchedule taint to avoid pods being scheduled there unless they have a toleration. I just need the components to be scheduled there as well as everywhere else.

? If those components are already running on nodes that don’t have a specific taint I’m trying to get around they shouldn’t need to restart

Not possible as this is out of control of Longhorn. When you set a toleration for a Kubernetes deployment/statefulset/deamonset, Kubernetes will restart all pods if these controllers. Also, bare pods like instance-manager-xxx pods cannot modify the toleration so Longhorn has to delete it and recreate it.