5.2. Integrating RSM with a Microsoft HPC or Windows-Based Cluster

You can configure RSM to submit jobs to a Microsoft HPC or Windows-based cluster by creating a configuration in RSM, as described in Defining RSM Configurations.

Following are additional considerations you may need to make when integrating with a Microsoft HPC or Windows-based cluster.

Prerequisites Installation

The Ansys product installation requires the installation of several prerequisites. The installer will check for these prerequisites on the machine where the installer is launched (for example, the head node). If you plan to have a network installation of Ansys products in your Microsoft HPC or Windows-based cluster, you must also install the prerequisites on each execution node. If you do not install the prerequisites on each cluster node, job execution may fail.

You can install the prerequisites separately by running ProductConfig.exe from the top-level directory as an administrator. You can also install the prerequisites silently using the following command:

ProductConfig.exe -silent -prereqs

For more information, refer to the Ansys Installation and Licensing Documentation.

For a Microsoft HPC cluster, you may also be able to install the prerequisites on all nodes using the clusrun utility that is part of the Microsoft HPC Pack installation. For more information, refer to the Microsoft HPC documentation.

Passwords

RSM does not require users to manually cache their Windows password with Microsoft HPC. Each RSM job runs the hpcutils.exe tool prior to submitting the job to the cluster. This tool programmatically does the equivalent of cluscfg setcreds.

However, if you still see the error messages regarding the password in the RSM log, such as "Failed to cache password with HPC" or "Account password MUST be cached with MS Compute Cluster," you may need to verify that the Service Packs for Microsoft HPC Pack and Windows Server have been properly installed. If you have not installed the Service Packs, you may still need to run cluscfg setcreds command from cluster head node to cache the HPC password.

Mixed Domains

You can use RSM when the client computer and the cluster are different domains. The assumption is that the client computer and user account are on the corporate domain and the cluster is its own domain. In this case, the cluster domain must be configured to have a ‘one-way trust’ with the corporate domain. That is, the cluster domain trusts the corporate domain but not vice-versa. Corporate domain users must be able to use cluster resources (login as CORPORATE\user into a cluster node). If the cluster administrator can add corporate domain accounts as cluster users, then this trust has likely been configured when the cluster domain was created.

Multiple Network Interface Cards

Cluster nodes, especially the head node, generally have multiple network interface cards (NIC) to facilitate separate public and private networks. When configuring the network topology for Microsoft HPC with RSM, be sure to select either Compute nodes isolated on a private network or Compute nodes isolated on private and application networks. Otherwise, client-server communication difficulties may arise and additional manual configuration will be required. Refer to Configuring a Computer with Multiple Network Interface Cards (NICs) for configuration instructions.

Network Path Configuration

If the RSM working directory or Ansys software installation is referenced using a UNC path specification (for example, \\nodename\path), refer to Network Installation and Product Configuration for special considerations related to network drives. Note that both the working directory and Ansys software installation must be have "Full Trust" set on all compute nodes.

Setting the AWP_ROOT242 Environment Variable on Execution Nodes

To avoid job script errors, it is recommended that you set the AWP_ROOT242 environment variable on all execution nodes.

Run the following command from the head node:

REM Share the installation directory so that compute nodes can use it, replacing the share name 
REM "AnsysInc242" with the share you want to use and the v242 installation path shown with 
REM that of your actual R19 installation directory
net share "AnsysInc242"="C:\AnsysInstalls_R242" /grant:everyone,full

REM Set AWP_ROOT242 on all nodes so that they can use the shared installation, replacing the share name 
REM "AnsysInc242" and install root directory "v242" with your actual share and installation root directory
clusrun setx AWP_ROOT242 "\\%COMPUTERNAME%\AnsysInc242\v242" /M