As far as I know, using DNS for hostname resolution does not "conflict" with Serviceguard in any way.
If the HP engineer insists on it, please ask him/her to identify a written document (e.g. a link to document and a PDF page number) where this restriction is documented.
I think the engineer won't find such a document, and might learn something new trying :-)
We have many Serviceguard clusters, all using DNS successfully.
There are some things you should do for maximum reliability, though:
- you should have the IP addresses and names (in both fully-qualified and short forms) of all the cluster members listed in the /etc/hosts files on each cluster node
- in /etc/nsswitch.conf, you should have both "hosts" and "ipnodes" configured to look up first "files", then any other sources you want (e.g. "dns" and maybe "nis", in whichever order you need).
As you should know, any Unix system should be configured so that it always can resolve _its own hostname_ quickly and successfully, even if external name resolution services (e.g. DNS and NIS) are not available. Otherwise some very basic things can slow down if the system has network connectivity problems. This can make it harder/slower to find and fix the actual problem, so you don't want that.
In a cluster, this rule should usually be extended to other cluster nodes too: each node should be able to resolve the names of itself and any other node in the same cluster. Adding the package IP addresses and their names to /etc/hosts is useful too.
If your node or package IP addresses ever change, you must already update the configurations on each cluster node: updating the /etc/hosts at the same time requires very little extra effort.
MK
MK
