[sccm 2012] Network drops during OSD

[clareq]

This is strange. OSD is working, and we have successfully PXE booted and built almost 1000 machines. However, Toshiba Tecra A9 laptops are confusing me. We can PXE boot them, and boot into WinPE. We put in the media password, and select the task sequence. If I set up a ping to the System Centre server it responds perfectly until I select the TS. At that point the ping times out and the build fails with a 0x80072EE2 error, which i believe is a time out.

 

I've checked the drivers in the boot image, but it can't be them, as we can see the list of TS I've set up. Typing this I've just thought it could be a problem with the TS I'm selecting, so I'm about to try a different one.

[FN-GM]

Have you tried using different network card drivers? Maybe try some vendor drivers instead of the built in PE drivers.

[clareq]

Dell only have vista drivers available, but I've tried them and the Intel ones. A different TS didn't help. Its strange - I can start a ping to the SCCM server, and it's fine, until I select a TS, then it just drops off.

[FN-GM]

Dell only have vista drivers available, but I've tried them and the Intel ones. A different TS didn't help. Its strange - I can start a ping to the SCCM server, and it's fine, until I select a TS, then it just drops off.

 

Did you stick those drivers in the Boot Image?

[clareq]

Yes, and I've tried the appropriate drivers in both the 32bit and 64bit boot images.

[clareq]

Strange. I've set the OSDComputername variable on the Unknown Computer Collection, so we are prompted for a computer name at the very start of the deployment. If I OK that box without inputting a name, the build continues. On all other models, the variable works as intended.

[clareq]

I'm wrong. It just crashes out later. The logs show it loading the correct network card, then later, without a reboot, it looses network connection. It only happens with the Toshiba Tecra A9 machines.

[TheCrust]

We have something similar - not necessarily the same - with our new Dell 3011 AIO PCs.

 

Our SCCM 2012 R2 build process involves PXE-booting into the 32-bit WinPE image and selecting the task sequence (including entering machine name if 'unknown'). The first step of the TS is usually a reboot into WinPE relevent to the target O/S - more as a historical thing than because it's actually needed - before formatting the drive and downloading/applying a thick image produced from a reference machine.

 

There's about 25-50% of our Dell 3011s that run through these task sequences without issue. They connect, select TS, stage WinPE and reboot, download and apply image just fine.

 

The rest don't want to know - they connect, select TS, stage WinPE and reboot - but then refuse to pick up a DHCP address after the reboot. This prevents them from connecting to SCCM to download/apply the image and the task sequence eventually fails with a (IIRC) 8007002 error.

 

There is something about the hardware that - despite being configured identically to the other Dells from the same batch, right down to BIOS version and settings - doesn't seem to want to collect an address via DHCP after the reboot. Opening a command prompt and running IPCONFIG /release or /renew has no bearing - they set themselves up with an APIPA address regardless and won't let it go until powered off and rebooted. The SMSTS.LOG on the clients didn't shed any light on the matter beyond telling us what we already knew.

 

In the end we took the reboot out of the TS to see what effect that had - and the machines built perfectly. Once imaged with the target 8.1 load they run just fine and connect without problem 100% of the time - but this reboot at the start of the TS just seems to make their network connectivity do odd things.

 

As said, none of the other Dells in the batch are affected by it and we've not seen this behaviour in any other machine in the rest of our estate (a mix of HP desktops and Toshiba laptops of varying pedigrees)

 

I had assumed it was down to the new Dells being inconsistent somehow - we have very little experience of their hardware - but having read your issues I'm not sure about that anymore...

Edited August 24, 2014 by TheCrust

[clareq]

They were dropping the network for us without a reboot. I'd F8 and start a ping going to our SCCM server as soon as it PXE booted, and it was fine - I could ping the server by name or IP address, and the laptop had picked up an IP address correctly. I would then work though the build process - select a TS, input the OSDComputerName variable. Regardless of the TS chosen, or whether or not a name was put into the variable, the ping stopped sometime before the first reboot. In the end I created offline build media, and they are building fine that way - at the end of the build they have successfully added themselves to the domain, with the correct name and in the correct OU.

[TheCrust]

You know when the ping drops, does the NIC still have a valid IP address bound to it? One that will work on your LAN rather than an APIPA address. Edited August 29, 2014 by TheCrust

[clareq]

Yes, it does. I can't understand what is happening with them They are built now, using the off line media option, but it's a strange one.

[TheCrust]

I wonder if there's something in the WinPE images that causes the network stack to fault under certain conditions....? Could you still ping the IP address assigned to the card and the loopback address when you were unable to access the wider network?

[clareq]

Didn't try that, just found an alternative way to build them, as we were against the clock. I might re-visit it when the madness of the next two weeks are over.