fix(reconciler/volume): disown replica from disowned volumes nexus

[abhilashshetty04]

While we disown a replica from volume we used to validate if the replica is owned by any nexus. In some cases involving Outage on volume target node we observed that the new nexus do get created after the volumeattachments are deleted from the node. However, we werent able to disown the Unknown replica from volume since it was still owned by old target.

Now we check if replica is owned by the current target of the volume rather then if it is owned by any target.

[abhilashshetty04]

@dsharma-dc , Before the disown_unused_replicas we run disown_unused_nexuses which will clear old nexus's owner id as the new nexus is already in the volume.state.target.

So we will receive only new nexus when we call from disown_unused_replicas

We can improve on docs i suppose. Its kind of race condition. Suppose we create new nexus and its published to the volume and reconciler already ran disown_unused_nexuses. So we will receive both nexuses in that scenario and the Unknown replica will not be removed. But in the next iteration we can guarantee that disown_unused_nexuses will be run before disown_unused_replicas and we get only active nexus of volume.

We have PR on develop. I will see how we can change docs in that.

[abhilashshetty04]

bors merge

[bors-openebs-mayastor]

Build succeeded:

• continuous-integration/jenkins/branch