Cordoning and then deleting the pods doesn't seem to have the same intended result as cordoning the node and setting eviction requested to true on the Longhorn node. @faint-sunset-36608 (You're on the git blame for the line I'm about to link 😅) Do you know if the Replica's new .spec.evictionRequested[1] field is appropriate for Longhorn clients to use? Or is it mainly a Longhorn implementation detail. For context, I'm working on a Harvester bug[2] and I'm trying to figure out if there's a more elegant way to do this instead of marking the entire node as eviction-requested. Though after typing all of that out I'm starting to think maybe just evicting all replicas on the node is probably the most desirable state to be in when putting it into maintenance mode anyway 🤷‍♂️ [1] https://github.com/longhorn/longhorn-manager/blob/266210f7c2ac6e4d8b80381d25daa673dd9577d5/k8s/pkg/apis/longhorn/v1beta2/replica.go#L50 [2] https://github.com/harvester/harvester/issues/3378