Hi,
( I guess you are in active/passive HACMP resource configuration )
Haven't Tried that before with TSM, but I Would Recommand:
Use three nodes:
one for each basic node excluding the shared directories.
one for the resource group service address (new node name), including only
the shared directories, keeping the password file ( and all other tsm client
files )
in a shared directory. (that way the node will go to the same place nomatter
on which
host you are on).
You should, probably, bring up the tsm scheduler for this node
as part of you start script, and bring it down with the stop script.

In case you have more than one resource group, implement that for
each one, assuming each could be brought up on a separate host.

In case of failover, the dsm node wail fail over with the resource group...

Good Luck.

Ran.