During a node or NIC failure, it is possible that of all the recorders in the failed node might reconnect to any other single available node. This connection issue is specific to CA enabled shares. If this connection issue occurs you can manually re-balance the cluster.
Manually re-balancing recorders across nodes
After any activity that causes recorders to move between PowerScale nodes, the recorder to node ratio can become unbalanced.
A recorder can be moved manually from the existing node to another node in the cluster to re-balance the node. You might have to perform the following procedure multiple times to get the recorder to the desired node.
When manually rebalancing a cluster, you can change the connection load balancing algorithm to Connection Count. When using Connection Count, use theisi smb sessions command to force the systems that you want to move. It is important to use a time interval greater than 1 minute between each forced server reconnection. Once the cluster is back in balance, change the connection load balancing algorithm back to Round Robin. Round Robin is the best algorithm for unattended reconnect scenarios.
Enter the following commands:
isi smb sessions list
isi smb sessions delete -f <computer name>