To configure Moldflow Insight to run on a cluster

This is a general guide only, based on a typical Portable Batch System (PBS) installation. Users with customized variations of PBS may require their system administrators to help configure the system.

The following instructions assume that you have a thin-client version installed on a filesystem that is shared between the cluster nodes.

System setup

  1. Enable SSH keyless authentication between the computer nodes, to permit the primary execution node to launch the Simulation Job Manager (SJM) on all nodes.
  2. Set all environment variables that a study requires to execute in the mfclient.env file, using the format variable=value. The mfclient.env file is in the root directory of the Moldflow installation file. Example environment variables include
    • the location of license servers
    • the location of required libraries
    Note: Transfer any variables that you have set via a 'module load' into the mflcient.env file, because modules only work for the first process launched. Subsequent child studies launched from the SJM do not inherit this module mechanism and fail.

Testing the installation

To test the installation, run the example.pbs script provided, with the diagnostsic option enabled. To enable the diagnostic option:
  1. Using a text editor
    • Change line 30 from DIAGNOSTICS=0 to DIAGNOSTICS=1
    • Adjust the PBS settings to match your system configuration
    • Alter the study file name (line 23) to match a study located in the same folder as the example.pbs script
  2. Run the script.
Note: When the diagnostic option is enabled, the SJM files are written to a sub-folder of the submission folder. You can use this information to diagnose potential issues. Do not enable the diagnostic option during the normal course of execution as it may cause the simulations to run slower.