This message was deleted.

Yes, apparently, Replica Zone Level Soft Anti-Affinity: Required. Allow scheduling new Replicas of Volume to the Nodes in the same Zone as existing healthy Replicas. Nodes don't belong to any Zone will be treated as in the same Zone. Notice that Longhorn relies on label

<http://topology.kubernetes.io/zone=|topology.kubernetes.io/zone=><Zone name of the node>

in the Kubernetes node object to identify the zone.

By default, "zone level soft anti-affinity" is true. If there are zone labels on the cluster nodes, Longhorn will try to avoid replicas sharing a zone, subject to other constraints. If that is not possible, Longhorn is permitted to place multiple in the same zone (the anti-affinity is "soft"). If the setting value is "false", the anti-affinity is then "hard" and no sharing will be done, possibly at the cost of not scheduling a required replica at all and leaving the volume degraded.