Migrating to the SCCM UDI for OSD, part 5: Quirks

Continued from part 4:
http://blog.uvm.edu/jgm/2015/03/10/sccm-udi-4-apps/

In the previous parts of this blog series I have shown how PowerShell and VBScript can be used to automate Driver, Application, and Operating System selection and installation in MDT/UDI. In this final post I will discuss a few final UDI quirks that we worked though without using scripts.

UDI Boot Media:

We experienced an odd problem during our OS Deployments when booting from SCCM/UDI media. The deployment process starts, as documented, with a dialog allowing the user to configure the networking environment for OS deployment, followed by a Task Sequence selection dialog. After selecting the task sequence, we expected some scripts to run, followed by the initiation of the UDI Wizard dialogs. Instead what we found was that UDI would first format an partition the local drive, copy a new WinPE environment to the drive, and then restart into WinPE running from local disk before brining up the UDI Wizard. This created a substantial delay between Task Sequence selection and the start of the Wizard. My impression was that users would find this delay unacceptable.

Our MS consultant suggested that this was happening because the revision of the UDI/WinPE environment on our boot media did not match boot image referenced by our active task sequence. To test this, we rebuilt all of the boot images in our environment from the Windows 8.1 ADK sources, then created a new MDT boot image from the ADK image, created a new task sequence referencing the ADK image, and finally generated new UDI Boot Media with the MDT boot image as a source. After completing these actions and updating our distribution points, we were able to get the UDI Wizard to start after Task Sequence selection without first requiring a reboot.

Improper OS Drive Letter Assignment

In our first deployments, we found that at the end of the Task Sequence, our operating system was left running on a D: or E: drive, instead of the usual “C:” drive. Knowing full well that this could cause massive end user confusion, I set out to correct this problem.

I was able to find a few potential solutions in Michael Neihaus’s always dependable TechNet blog:
http://blogs.technet.com/b/configurationmgr/archive/2014/04/28/how-to-ensure-that-windows-installs-on-c-during-a-system-center-2012-configuration-manager-osd-task-sequence.aspx

After weighing the options, I chose to try the “Delete Mounted Devices from registry” option. This required me to import a custom task sequence into our SCCM instance, and then copy/paste the required Task Sequence steps into our production Task Sequence. This approach instantly was successful, so I am sticking with it. Thanks again, Michael.

Failure of UserExit Script Processing

Under MDT/LTI, we made use of the “User Exit Script” capabilities of MDT to generate unique default computer names for UDI clients. Cursory inspection of the docs suggested that User Exit script functionality also is supported under SCCM/UDI, so I decided to transplant our script. However, after doing so I found that the default presented in the UDI Wizard was #GenUniComp()#, rather than a generated computer name.

#GenUniComp()# is the value assigned in our CustomSettings.ini file for the OSDComputerName variable. However, MDI/UDI is supposed to “exit” any code placed between the hash tags (#) out to the Windows Scripting Host. “GenUniComp” is the name of a VBScript function that I defined in our “ZUVMUserExit.vbs” User Exit script. The function works well enough under MDT/LTI, so what is going wrong in UDI?

To diagnose this problem, it helps to know a bit about how User Exit scripts are processed. “UserExit” is set as a variable in the CustomSettings.ini for your environment. This variable is detected and processed by the ZTIGather.wsf script that is run several times during each MDT/LTI and SCCM/UDI deployment. To troubleshoot problems with the script, you should look to the SCCM/UDI log files generated by ZTIGather.wsf. When I examined the ZTIGather logs, I could find no reference at all to the UserExit variable getting processed by the ZTIGather script. The next logical question is, why not?

Logically, the problem is not likely to be caused by an error in the script. First, the same script works just fine under MDT/LTI. Additionally, the script uses the MDT logging functions, so if it had executed, we would have expected to see at least some log output. I thus assumed that the script was not getting executed at all. I started looking at the ZTIGather logs from the top, and found reported “processing [Settings]…” followed immediately by “processing [Default]…”. Interestingly, I had defined the UserExit variable under the [Settings] section of CustomSettings.ini, which worked find under MDT/LTI. However, the SCCM version of ZTIGather appears to process only “Priority” and “Properties” settings under the [Settings] section. My UserExit definition was getting ignored because it was in the wrong place. After moving the “UserExit” definition to the [Default] section, UDI started to process the GenUniComp() function of my User Exit script just as it had under MDT/UDI.

While our UserExit script is not really involved in driver handling, I am including it below just for the sake of completeness:


In the final section of this series, I will present operational instructions for the use of these scripts in our environment. These documents are tailored for the UVM environment, but might serve as a template for instructions at other sites.

Next:
http://blog.uvm.edu/jgm/2015/03/12/sccm-udi-6a-opsdrivers/


Series Index:

Advertisements