hi chip,

i'am faced with this problem sometimes too.

i solved it always using the follwing way.

1. stop the service configured in the cluster as a "generic service" .... 
normaly already done by the failure.
2. within the configuration for this "generic service" -> delete the 
configured registry-key replicated by the cluster
3. on the cluster-node, where the resource-group is currently running, in 
which the generic service is configured start the tsm-cli by
        dsmc q mg -optfile=<opt-file which is used>     (any other command 
is working also)

        now you should be prompted for the user/node-id and the according 
password -> type them in

4. now the password should be stored within the registry.
5. start the configured generic service wihtin the cluster administrator
        now the service should start and stay online
6. now update the configuration of this generic services with the recently 
deleted registry for registry replication

-> now you can switch the resource group from this cluster-node to another 
and the service should start on the other node and stay online too.


Mit freundlichen Grüßen / Best Regards 
Uwe Schreiber