| Hi Andrew,
You right it is either X or Y type of answer I have to blame it on the
cold morning here well everyone needs an excuse :-).
What I ment was that you need to the FCS running on the nodes where you
have ALL-IN-1 installed and not on all the nodes. Your users will be
connecting to the two nodes on the cluster to use ALL-IN-1, so you do
not require it on the other nodes.
Regards,
Sunil
|
|
re .3
It may not be quite that simple. For the non-DSO case it is as .3
says, you only need the FCS on the nodes that are running ALL-IN-1.
Where it could get a little confusing is if you have the DSO installed
on your cluster, and remote users have named the drawers by cluster
alias, or a non-ALL-IN-1 node. The two cases are different:
Drawers named via cluster alias:
In this case when the FCS is requested to get info on the drawer
(talking about on the remote system now), the FCS will notice that
it does not serve the drawer on cluster XXX, and will broker to
XXX::"73=". If XXX is a cluster alias, the exact node to connect
to is decided by DECnet, in which case, if all the nodes are
enabled for incoming connects, you could connect to any of them.
There are two ways around this: 1) Don't use cluster alias as
the partition name when creating a drawer, use the real node
name you want and make sure it is one of the nodes that runs
ALL-IN-1. 2) Only have those nodes that are running ALL-IN-1
enabled for incoming connects.
Drawers (ok partitions) named as the node not running ALL-IN-1.
Well if this happens, the drawer create would fail miserably and
the user would be told FCS is not available. To get around this
either use cluster alias (and see above) or don't use the
non ALL-IN-1 nodes as partition names.
If you are not on the net, other than your cluster, and there will not
be a DSO installed, then as .3 says, you only need the fCS started
on the nodes that are running ALL-IN-1.
--bob
|