Server 2008 odd network card behaviour

[TheCrust]

We've got a bit of an odd one here.

Over the summer we started upgrading our servers to 2008 server. The DC's have all gone across fine (were DCpromo'd down, rebuilt from scratch rather than upgraded, and DCpromo'd back up) and the fileservers went likewise.

Some stuff is still on 2003 server, for instance our Exchange box, but the core domain infrastructure and file servers are all now migrated to 2008 server.

Now last week we had one of the student file servers - a Stone Computers "intel clone" rack job with dual Intel Pro 1000MT server adaptor and Windows Server 2008 standard - reporting that both network cards were disabled even though they weren't. The affected server could access the network and surf the internet, ping other hosts, open up shares, worked a treat as you'd expect BUT try to access it from any other device and it refused to respond to a ping and you couldn't access any shares on it or even start a RDP session. Just as if the firewall had reenabled itself (it hadn't - it's disabled via a GPO).

Right-click the card, select "enable" as normal, and the server showed the enabling dialogue box but then re-reported the card as disabled. All very odd. Two reboots finally cured it.


Today we've just had the same thing but on another server. This time a HP Proliant DL180 G5, brand new this summer, single Broadcom gigabit card onboard - running Windows Server 2008 enterprise (it's one of our DCs)

The server stopped responding on the network and investigation showed the NIC to have "disabled" itself - even though the server could still access resources on the network.

As with the other server, a reboot cured it - but we shouldn't have to be doing this sort of thing to a server!


Neither server has any power management features enabled on the NIC (why would you want a server to switch a NIC off?) but they are connected into the same stack of Cisco 3750G switches.


Anyone else had this sort of problem or behaviour with their Win 2008 boxes?

[Zimmer]

Not seen this issue with 2008 unfortunately.

However I have seen something almost identical on 2003 R2 (64bit) on our file servers last year.

Do you aggregate the dual Intel NIC's using the Intel driver software?

We configure our file servers with Intel dual NIC's for added throughput and for a bit of failover should one of the switches in the stack fail. Anyway, we found it was down to the latest drivers from Intel, we rolled back to a previous version and the problem was solved. We have since upgraded to 2008 and we are using the latest drivers from Intel's support site and we haven't had a problem yet (touch wood!).

[Valith]

I'm glad I've found someone else with similar issues. I did three server 2008's for a client back in the summer. The first DC was an existing 2U Dell Poweredge and the second two were both 1U IBM servers. It doesn't happen often but every once and awhile the Dell DC flakes out and begins having severe network issues. I can connect to it remotely through the firewall but it'll do one of the following odd behaviours:

1. No network machines can connect to it or open it's file shares. However it can browse the network/internet just fine without an issue.

2. Network machines can connect to it and use it's file shares. However it cannot connect to any network machine/server in the organization. One time when this happened I tried to open up DNS/DHCP console and it couldn't find any DNS/DHCP servers (itself being one and the services were running.) When I launched AD console it couldn't find a domain controller (again it is the primary DC.)

Each time reboots have resolved the issues.

It'll go for two weeks or two months having no issues then it'll flake out. It's only happened maybe five times since the install but it really does concern me quite a bit. I haven't been onsite in a couple of months so after reading this post it makes me think I should really take a look at updating the drivers on it and seeing if that helps.

The two IBM 1U servers haven't experienced a single problem during the entire time.

If you have any resolutions or anything else you run into let me know I'd be more than interested to hear about anyone else with similar problems!

[Michael]

I have never seen this issue on 2003 Server. It's interesting you've seen the problem on two sets of hardware with 2008 Server. It could well be a bug which could be fixed with SP2, whenever that will be released.

[IanT]

you havent got any NIC Software running have you? Teaming? Failover? Load Balancing setup on the NIC?

[TheCrust]

Hi guys - thanks for the replies, sorry for my lack of.

@IanT and Zimmer, no - we don't use any teaming software - for instance the Proliant DL180 only has a single NIC so it can't be teamed with anything and the Intel-based machine had it's second NIC disabled after Win2k8 decided it was going to refuse to bind any protocol stacks to it.

Since I posted up, both machines have behaved themselves to date - and we haven't seen the problem on any other Win2k8 box on the network.

It's very odd. A bit like one of our Win2k8 terminal servers reporting it had run out of disc space last week when it hadn't. Again, a reboot cured it but quite why it was needed when the same hardware under Win2k3 would have run perfectly for months I have no idea....

I wonder if Win2k8 R2 will be just as entertaining?