Troubleshooting Experts Only --- Dual-port NIC issue...

Soulstice

[H]ard|Gawd
Joined
Jun 26, 2001
Messages
1,104
Hey guys,

This is the first time I've ever run into a problem like this.

The situation is as follows:

O/S: Windows 2K8 R2 Core --- Enterprise Edition

NIC(s): 2 x Broadcom NetXtreme II BCM5709C Dual-Port Controller




I am able to get both NICs to install, but only a single interface on each card reports to be working properly in Device Manager (when I remotely connect).

I originally used pnputil.exe to install the drivers; however, in order to further troubleshoot the issue, I installed the most updated version of devcon x64, from the Driver Development Kit.

For those unfamiliar, devcon is the command-line equivalent of the Device Manager GUI.

I've tried everything... removed the NICs using devcon, re-installed the drivers, removed the drivers, forced an update, removed the Windows built-in drivers (infs), forced another update, etc, etc, etc.

For some odd reason, I can only get a single interface on each card to function. The other interface reports the typical "Exclamation mark" in the Device Manager GUI, specifies the following error: This device is not working properly because Windows cannot load the drivers required for this device. (Code 31)

If anyone of you have dealt with this sort of issue in the past (e.g. multiple identical devices on a system, yet, only one is able to load), please help me out.

I'm pulling out my hair. For the record, I had both NICs working on a separate system using an identical motherboard.... ...both interfaces on each of the nics worked. It was only after I Ghosted the system.

However, I've removed all hidden devices using devcon and a small script I wrote, so it's not a conflict of any sort. I'm beginning to think it's something registry-related. I've been working on this problem for two straight days, and I can't afford to waste anymore time. I'm in the process of deploying a Hyper-V Cluster using CSVs, and setting up the NICs was supposed to take half a day at most.

Any advice would be much appreciated. If there are places in the registry where I should be looking, or if there are specific settings that I need to check, please let me know.


Thanks in advance.
 
I used Symantec Ghost Suite 2.5


I mapped a drive over the network, and used a custom-made WinPE DVD to boot the system.

Then ran the following command:

ghost32.exe -z5 -auto -split=2048


Pretty standard. I've never had a problem with imaged systems in the past.
 
Nope, the system was not SysPrep'd! In fact, I hate SysPrep.

In the past, I would use NewSID. However, as of Windows 7 / 2K8 R2, NewSID no longer functions properly and wrecks the system. Since then, I have used the second-best option, Symantec's GhostWalker.

However, in the case above, I've tried imaging and not using GhostWalker, to see if it was GhostWalker causing the issue somehow... ...same problem.

I have since installed and built my second cluster node from scratch, and all NICs are working as expected. However, I haven't attempted making an image yet, so it's possible that this issue will arise again if I need to add a third system with identical hardware to the cluster, or if one of my machines crashes and I need to restore / rebuild from scratch.

I've never run into a problem like this before. I think it must be registry related somehow... or a specific inf or driver file that I was unable to get rid of. Although, I deleted and cleared out any suspect prior to attempting to re-install the Broadcom drivers again.
 
We ran into this same problem using Syspreped images for 2k8 R2 server core DCs late last year.

We actually had a Microsoft Engineer on site assisting with our build process. The ONLY way we could get things to work reliably was to use an unattended automated install. We could not get images to work period. Didn't matter which imaging tool we used. Imagex wouldn't work, and neither would anything else we tried.

We never were able to find a fix process either. it could not be corrected without reloading from scratch using a normal server cor isntall sequnce of some sort to do the install (manual or an unattended automated).

We were on a really tight timeline and it was killing us because we took 2 weeks fighting it.

Working with the full version of 2k8 R2 like we were for 6 of our 60 some odd DCs, we could use sysprepped imagex images with no issues, but not with server core.
 
I've had a lot of fun with Server 2008/Hyper-V and Broadcom NICs also - remember the TCP Offload/Chimney problems they had with the RTM version of R1? As of late I've also had a lot of trouble with the RSS (receive side scaling) functionality also. Both of these issues were resolved with firmware updates - so while it really isn't very good troubleshooting practise - are you sure you're running the latest firmware available from the vendor?

To be honest, after the above issues, we just made the switch to Intel NICs which have been as reliable as NICs should be.
 
Yep, I'm still unable to figure out the problem, but I have been able to reproduce it.

I loaded Windows 2008 R2 Enterprise on a completely separate Opteron server using nVidia nForce NICs.

Windows used its own drivers, and loaded them without a hitch.

I then installed the nVIDIA drivers, and again, they installed without a problem.


However, I then decided to uninstall the nVidia drivers, and reboot, assuming the system would find the NICs and load the Windows drivers. Nope... ...only one of the two nVIDIA NICs comes up, while the other remains with an error 31 (Cannot load / find the drivers). No matter what I do, I am unable to get the second NIC to load.

Holy shit... ...this is obviously not an isolated incident. I love Server Core, but someone from Microsoft better comment on these ridiculous driver issues. At the very least, make the Remote Device Manager work in Read/Write mode, instead of Read-Only, because being forced to use DevCon to remove, update, and re-install *.inf driver packages is painful, especially when troubleshooting an issue such as this.

If anyone else has comments or troubleshooting advice, I'm all ears. :(
 
I have come across EXACTLY the same problem but with a different NIC manufacturer, which indicates to me that it is not driver based (which were my initial thoughts too). I have run up a Hyper-V 2008 R2 server with an onboard Broadcom NIC to which I then added an Intel ET Dual network adapter post O/S install, i.e., PnP. I have one of the two ports working and the other one is showing the Exclamation Mark in Device Manager with the same error code of "This device is not working properly because Windows cannot load the drivers required for this device. (Code 31)".

I have raised a call with Intel who have suggested a resource conflict and asked me to disable to the onboard NIC before reinstalling the drivers but given the DevCon hassle I think it may be just as quick to reinstall the entire O/S with the NIC in place from the get-go. I will let you all know how I go.

BTW I second (and third, and fourth!!) the point that Soulstice made about the read-only device manager issue. Given that Hyper-V (and server core) don't have any real means of administering devices (I am refusing to accept DevCon as a solution) the least they could do is remove the read-only restriction from the remotely connected MMC.

Anywho, I will report back later on this one....
 
Ok, after updating the bios and changin a few settings to no avail I opted for a rebuild based upon Soulstice's experience of rebuilding the opteron server and having both NICs operational. The early indication is good - I have rebuilt the Hyper-V machine and have both NICs operational - I am about to install the intel drivers and expect it should be fine. If you don't hear back from me you can assume it all worked.
 
Back
Top