When a node comes back up after being rebooted and attempts to rejoin the cluster, the join process automatically changes the node's state from LEFTCLUSTER to DOWN so that it can rejoin the cluster. When this occurs, the join server initially sees the node that is attempting to join the cluster as being in the LEFTCLUSTER state. The join server signals the joining node that it is busy because the joining node is not in the DOWN state. It then notifies all of the remaining nodes in the cluster that the joining node is DOWN and to start the node-down processing, which must be completed before the node is allowed to rejoin the cluster. The joining node continues retrying to join the cluster until the node-down processing is completed on all of the cluster nodes at which time the joining node is allowed to rejoin the cluster.
If SF is not running on all the nodes, or if SF is unable to shut down the node which left the cluster, and the LEFTCLUSTER condition occurs, then the system administrator must manually clear the LEFTCLUSTER state. The procedure for doing this depends on how the LEFTCLUSTER condition occurred.
