You did not tell us what your cluster's topology looks like and you think the HA release is 5.4. So let me make some assumptions based on what it should look like when you use IPAT with aliasing which is the default for 5.4.

Your cluster has two physical adapters which are configured during boot with a non routed IP address each. E.g. after node boot your adapters are configured with sth. like 192.168.10.1 and 192.168.20.1 on one node and 192.168.10.2 and 192.168.20.2 on the other. These addresses are just used for cluster communication and the nodes are not reachable via TCP/IP from your production network. Only after a Resource Group is online you have a (Service Label) IP address active and the default route. At least this is the presuppostion HA makes. If you understand this you now know why xntpd becomes active after a RG came online only: HA 5.4 deferrs the start of all services that rely on a working production network.

To avoid the nodes (without an active RG) being not reachable via TCP/IP from production network IBM introduced the Persistent Address. Such an address is configured during node boot as an alias and while it is accessible right from the start it is not part of a resource group. Opposed to an aliased boot address such a Persistent IP may be in the same subnet as the Service IP Label. Because the default route is usually tied to this subnet the node is then always reachable just as any other non-clustered node. With a working TCP/IP environment you could start it from the /etc/rc.tcpip automatically with

Code:

start /usr/sbin/xntpd "$src_running"

and/or manually with

Code:

# startsrc -s xntpd

You may read the HA 5.4 Installation Guide and the HA 5.4 Admin Guide. Then follow up with the sg246769 redbook aka HACMP cookbook.