I have 2 2003 std servers with multiple HP NICs installed. The configuration established sent the onboard NIC to the user's network while the other two NICs went to completely separate subnets for the sole purpose of iSCSI. Using the MSoft iSCSI initiator I have configured failover between the two iSCSI NICS that resolve on said separate subnets and it works beautifully - hops across with no service interruptions.
user's subnet is 192.168.0.x
iSCSI 1 subnet is 192.168.3.0
iSCSI 2 subnet is 192.168.4.0
All the subnets are electrically isolated save for the 2003 servers in the middle.
So far so good. However, I am experiencing a problem on the user's subnet as follows:
The only items on the iSCSI subnets are iSCSI devices, so users have no need to access them for any purpose. Recently we noticed some users experiencing what seemed to be random slowdowns for network activities. A trace shows that the NIC IP addresses for the iSCSI interfaces are being pushed to clients via AD DNS.
I subsequently removed the "update dns" option from each iSCSI NIC and manually removed the 3.x and 4.x entries in the DNS servers but 20 minutes later they are back and messing with client PCs.
Apparently I have overlooked some aspect of this - how can I prevent those two iSCSI interface NICs from being handed out in DNS requests, short of manually entering the correct subnet server IP address (192.168.0.5) in lmhosts?
Seems like I should be able to at least prevent DNS entries on an as-needed basis. I think that is all it would take really.