You use a System Center Configuration Manager PXE service point (Configuration Manager 2007) or a Pre-Boot Execution Environment (PXE) distribution point (Configuration Manager 2012) to perform PXE boots. In this situation, the operation first appears to work successfully, but then the process stops running. When you examine the server on which the PXE service point and Windows Deployment Services (WDS) are installed, you discover that WDS has crashed.
When you restart WDS on the server, this does not resolve the issue. When you restart both the Windows Management Instrumentation and WDS, or when you restart the server itself, this may temporarily resolve the problem. However, the issue eventually recurs, and WDS crashes again.
If you try to reproduce this issue by continuing to perform PXE boots, you discover that although the issue may occur frequently, it cannot be reproduced on a consistent basis. The crash behavior occurs randomly.
To work around this issue, use one of the following methods:
· Disable the PXE IP Helpers in the backup, redundant router so that duplicate PXE requests are not sent. For more information about PXE IP Helpers, see the following TechNet article:
· Configure the Configuration Manager WDS provider to be single-threaded instead of multithreaded. This will limit WDS processing of PXE requests to one at a time and will prevent the second, duplicate PXE request from conflicting with the original request. To configure the Configuration Manager WDS provider for single-threading, create the NumberOfThreadsregistry key with a DWORD value of 1 in the following location:
Configuration Manager 2007 32bit WDS server:HKEY_LOCAL_MACHINE\Software\Microsoft\SMS\PXE
Configuration Manager 2007 64bit WDS server: HKEY_LOCAL_MACHINE\Software\Wow6432Node\Microsoft\SMS\PXE
Configuration Manager 2012 DP/WDS server:HKEY_LOCAL_MACHINE\Software\Microsoft\SMS\DP
Doing this does not typically affect server performance for PXE requests except in environments where a large number of PXE requests are performed on a consistent basis. In these environments, we recommend that you use the first workaround.