Problem

How to system automation when moving from stand-alone CFT to multi-node?

Resolution

Once moving from a Stand-alone to a multi-node configuration, system automation will have to be revisited in regards of Transfer CFT.

Task monitoring is no longer to be done on the CFT monitor(s) (that are the nodes for such a configuration).

The CFT nodes are managed directly by the node Managers (Copilot).

It leads the automation to monitor only the node manager and eventually, to issue commands to re-balance the CFT nodes.

All actions on the CFT nodes are to be managed through the node managers, this include the start and stop of nodes when needed.

At the initialization of an LPAR, the automation can start the node manager and, once initialized, issue the restart command to re-balance the nodes in regards of the available node managers.

Fine action can also be automatized to act on single node and avoid a full nodes restart if preferred.

Currently, there is no automatic node re-balancing when a node manager is bring back up (other than the restart command that shut and restart all the nodes) and such a feature, requiring a kind of 'node affinity' could become available in a future version.