| >> I think this behavior is a limitation of NT Cluster
>> (no bug, a feature :-)
>>
>> 1. is this right ? is it a limitation ?
>> 2. is there anything we can do immediately to make a
>> cluster running under this network environment ?
>> 3. What (and when) will make NT Cluster work in a routed
>> network (these are very common in large companies !!!)
This issue has been discussed many times in this conference. The
product is defined as a LAN solution, not a WAN solution. There are
two options to get it working in a WAN.
1. allow broadcasts through routers. You, as everyone else, have
found this to not be the answer.
2. connect to the individual nodes instead of the cluster alias. This
does not give you the client failover, but it does give you the 'fastest
field service in the world' for accessing the data on the backup
system. Everything on the cluster works the same. It is just that
the client has to know to reconnect; just as if they were Mac, DOS, or
OS/2 users.
tgc
|
|
rep. 0
> via routers, no broadcast. Name resolution is done
> via WINS.
rep. 1
> 1. allow broadcasts through routers. You, as everyone else, have
> found this to not be the answer.
Actually, if the customer exclusively utilizes WINS, meaning
all clients are H-mode, then the issue of broadcast traffic
(across routers) becomes of little consequence.
The other alternative is wait for NT Cluster V1.1. WAN support
is addressed via IP failover. Network file share support will
hopefully be addressed in a service pack.
And then, there's Wolfpack migration sometime this summer.
|