DC and multiple NICs

[danomac]

Okay, I have a couple of DCs that have two NICs. Initially I assigned each NIC to be a part of the same subnet (thinking for reliability reasons - more on that below.)

So each server is something like this:
x.x.aaa.2/24
x.x.aaa.3/24
x.x.bbb.2/24
x.x.bbb.3/24

DNS and DHCP are working correctly (listens on both and is dishing out both IP addresses. One is a forest controller and the other is a child domain in the forest. Trust traffic works with no issues.)

However, sometimes when a user logs on it says that their profile/folder redirection doesn't work. I suspect this has something to do with netbios, as manually typing it in in windows explorer shows what I want to see. There seems to be a 1/4 chance that it won't work when logging in.

I've read elsewhere that DNS and netbios can cause issues like this, so I debated teaming the NICs as one IP address. However, I've read real mixed results doing this and apparently Microsoft doesn't support it. So what's one to do on a DC? I'd like the load balancing if I can get it as sometimes a dozen people like to seemingly log on at the same time, leaving one of the gigabit nics unused.

It appears that both servers can/will support teaming provided I install the Intel software to do so. Now I wonder if I should just disable one adapter on each machine instead of teaming.

Anyone have any experience with this?

 

[itsp1965]

multihomed DCs (multiple active NICs) are more trouble than they're worth unless as you said you enable teaming for performance purposes. If your DC is not overly taxed you would be better off just disabling the one NIC.

 

[danomac]

Ironically enough, when I was pondering what to do exactly one of the servers involved "lost" it's network connections. I don't know why, there's nothing in the event logs, but I took it as a fast opportunity to install the latest drivers and set up a team. After reading elsewhere that network teams are working well in several environments I felt a little better and tried it out.

The box in question is also an Exchange box (we don't have financial means of running a server for each service.) I've noticed the throughput to it is much higher now.

I'll just let it chug along and monitor the event logs for a few days to see if any oddball issues come up.