11.4. Known Issues and Limitations

The following are known issues at the time of release, as well as system and software limitations:

When submitting distributed Ansys Mechanical and Mechanical APDL solutions to a multi-node Ansys RSM Cluster (ARC) on Windows, each calculation is limited to using the cores of a single machine, regardless of the number of cores requested. The cores on the remaining machines remain idle for the duration of the calculation. This limitation does not apply on Linux.
When submitting a design point update from Ansys Workbench to a SLURM cluster via RSM, the job may fail with a 'Segmentation fault' error. This is caused when subtasks are executed one by one instead of in parallel, resulting in some subtasks timing out while waiting to start (the default timeout is 5 minutes).
To work around this issue, add the --exclusive argument to the runSubTask command in the {installdir}/v242/RSM/Config/xml/hpc_commands_SLURM.xml file as shown below:
```
<runSubTask>
    <primaryCommand name="runSubTask">
      <application>
        <app>srun</app>
      </application>
      <arguments>
        <arg>--exclusive</arg>
        <arg>--ntasks=1</arg>
```
For details about the --exclusive argument, see https://slurm.schedmd.com/srun.html.
When using an application's My Computer, Background setting to solve a SMART Crack Growth problem on the local machine, the solution may fail as a result of a failed remeshing attempt. Installing ARC services as daemons resolves this issue.
Currently the RSM Job Monitoring application shows the jobs of the current user only. It does not show the jobs of all users.
For System Coupling jobs:
- Coupling jobs must be submitted as design point updates.
- Coupling jobs must be run in the foreground, so all coupling participants must be configured for foreground solution updates.
- Coupling jobs that are submitted to Microsoft HPC 2019 may fail if the cluster staging directory is not shared from the head node.
- The coupling process may fail to start for coupling jobs that are submitted to Microsoft HPC Pack 2019.
A solution submitted to RSM may fail to start if the name of a folder in the project path contains special characters, or characters with special marks such as accents. Removing these characters from the folder name resolves this issue.
Currently, the RSM Cluster Load Monitoring application is supported on Windows only.
When running remote design point updates through RSM to a Linux cluster, some jobs may be shown as failed in the RSM Job Monitoring application, even though the design point calculation completed and all results were retrieved. This will be accompanied with an error such as "[ERROR] FATAL UNHANDLED EXCEPTION: System.ExecutionEngineException: ExecutionContext_ExceptionInAsyncLocalNotification ---> System.Threading.ThreadAbortException" in the RSM job log. In this situation, the results that have been retrieved are correct, and the failure status can be ignored.
Job submission from Windows RSM clients to a Linux cluster may fail at the user authentication stage if user accounts do not follow the account@machine format.
Passwords that are cached by RSM and used to submit jobs to HPC resources cannot contain special characters, or characters with special marks such as accents.