Internal IP address ranges

The number of IP addresses assigned to the internal network determines how many nodes can be joined to the cluster.

When you initially configure the cluster, you specify one or more IP address ranges for the primary InfiniBand switch. This range of addresses is used by the nodes to communicate with each other. It is recommended that you create a range of addresses large enough to accommodate adding additional nodes to your cluster.

While all clusters will have, at minimum, one internal InfiniBand network (int-a), you can enable a second internal network to support another Infiniband switch with network failover (int-b/failover). You must assign at least one IP address range for the secondary network and one range for failover.

If any IP address ranges defined during the initial configuration are too restrictive for the size of the internal network, you can add ranges to the int-a network or int-b/failover networks, which might require a cluster restart. Other configuration changes, such as deleting an IP address assigned to a node, might also required the cluster to be restarted.