Cluster Configuration

Good Day,

I'm trying to configure my calculations over a cluster proceding as explained on the website, but once I launch a calculation on a case I already realized under Caedium locally, I get the following error :

volume [www.1702.5] RANS Flow: Failed
Unexpected process exit.

What should I do?

Thank You for any answer you can give me.


WIndows HPC Server Cluster Configuration Diagnostics

I'll assume you are following the instructions at "Microsoft Windows HPC Server 2008 Cluster Configuration".

Caedium has to be installed on a node with job submission rights for the head node.

Specifically did you successfully complete assumption 2:

"You have saved your user ID and password with the Microsoft HPC Cluster Manager. Try submitting a Tests->System->MPI Performance test (e.g., MPI Ping-Pong: Latency) under the Diagnostics group in the HPC Cluster Manager. Then check the state of the latest job. If the job state is configuring, then right-click on the job and select Submit Job..., provide your user ID and password, check the option Remember my password and click OK. "


Yes this step was successfully completed.

More HPC Diagnostics

It might be helpful to switch on OpenFOAM diagnostics, if you haven't already, and monitor the log window:
File->Preferences, Physics->OpenFOAM->Diagnostics = on

I'll assume you have successfully set up the shared folders too, right?

Try running a case with the Shared Memory option from Caedium with the shared folder settings. Does this work?

When you try to run a case from Caedium using the Cluster option does the job appear in the HPC Cluster Manager? If so does it provide any diagnostics as to why the job failed?

No job appearing

My diagnostics were on while running, I had the mesh decomposition and then the error message. And, yes I shared forlders too.

Indeed, the jobs are not appearing in the HPC Cluster Manager, and so I have no diagnostics about the job failing.

User Privileges

Are you using Caedium as a user with HPC job submission privileges?

Try issuing a test job using the command-line job submit syntax.


Yes we are using it with submission privileges. We use the same machine for other software and it works usually. I'll try with the other software to see if the problem comes from the machine or the cluster.
Thank you for your help.