Del,

I fixed this in the following way:

1.      Two services need to be installed on each node of the Cluster with
the batch file;  dsmcutil install /machine:%1 /name:"%2 TSM Agent"
/node:"SQL_%2" /password:"SQL_%2" /optfile:%3:\tsm\msSQL\dsm.opt
/clientdir:C:\tsm\BACLIENT /clusternode:yes /clustername:<clustername>
/autostart:no
a.      Install the first service on Node 1 using the parameters
<machine_name_1> <Virtual_Sql_Name_1> <shared_disk_resource_name_1>
b.      Fail all Node 2 resources over to Node 1
c.      Install the second service on Node 1 using parameters
<machine_name_1> <Virtual_Sql_Name_2> <shared_disk_resource_name_2>
d.      Fail Node 2 resources back to Node 2
e.      Install the first service on Node 2 using the parameters
<machine_name_2> <Virtual_Sql_Name_2> <shared_disk_resource_name_2>
f.      Fail all Node 1 resources over to Node 2
g.      Install the second service on Node 2 using parameters
<machine_name_2> <Virtual_Sql_Name_1> <shared_disk_resource_name_1>
2.      The result is that each node is running both of these two services,
both set to manual.
a.      <Virtual_Sql_Name_1> TSM Agent
b.      <Virtual_Sql_Name_2> TSM Agent
3.      Now, when the nodes fail over, the service restarts and any
scheduled event that is still within the window will restart.

I've been unable to find this documented anywhere in the TSM literature, or,
if it's there, it's not clear that both services are required on both nodes.

Glen