Disjoint namespace causes Windows HPC Server 2008 R2 compute nodes to show error node health

View products that this article applies to.

Symptoms

The node health of all compute nodes in a Windows HPC Server 2008 R2 environment change from an initial node health of "OK" to "Error". Restarting the HPC Management Service on the compute nodes or rebooting does not resolve the issue.

↑ Back to the top

Cause

Windows HPC Server 2008 R2 utilizes the Environment.UserDomain property to build the fullyqualified path to the compute nodes. In a disjointed namespace, the result is the environment variables USERDNSDOMAIN and USERDOMAIN are different, resulting in an incorrect query to the DNS server.

↑ Back to the top

Resolution

A fix for this issue is scheduled to be included with Windows HPC Server 2008 R2 Service Pack 3.

↑ Back to the top

More Information

To identify an environment with this issue, open an elevated command prompt and type "set U". Compare the values for USERDNSDOMAIN and USERDOMAIN. If the value for USERDOMAIN does not match the first portion of the string for USERDNSDOMAIN, install Service Pack 3 to all machines in the HPC environment and allow 10-15 minutes to for the Node Management Service to update the Node Health status of the compute nodes.

↑ Back to the top