Running HFSS on a slurm-based machine/RSM cannot be accessed

Hello everyone,

I am trying to run a HFSS simulation at a remote machine (SciNET) that uses SLURM as the scheduler.

I have made a script with the below commands:

 

The file jobscript.sh and I run it from the $SCRATCH (=/gfps/fs0/scratch/g/gelefth/vatalogg/) using "sbatch jobscript.sh". I have also changed the directories to point to the $SCRATCH folder, because I could not write in the default directories.

The problem is when I run it, the job is submitted and it creates a "results folder", but it is not solved and I get an error:

"[error] Project:helical_antenna, Design:HFSSDesign1 (DrivenModal), Unable to locate or start COM engine on 'nia0042' : Unable to reach AnsoftRSMService. Check if the service is running and if the firewall allows communication. (11:04:21 AM  Jul 15, 2020)"

Here is the relative file with the batch info:

 

Can someone help me with this problem and guide me to solve it? I am not familiar with running HFSS on linux through a command line, so I apologize if it is something obvious.

Comments

  • KremellaKremella Admin
    edited July 17

    Hello,

    I'm moving this to the Installation and Licensing thread for better visibility. I hope this helps.

    Thanks

    Karthik

  • mmadoremmadore Forum Coordinator
    edited July 17

    Do you know if SciNET has configured the SLURM custom scheduler proxy for ANSYS Electronics products?  Are other users running successfully on this SciNET cluster with HFSS?

     

    Thanks,
    Matt

  • vataloggvatalogg Member
    edited July 17

    Thank you Karthik for moving the post and Matt for your response.

    To be honest, I do not know if Scinet has configured their scheduler to run for AnsysEDT. I will ask them, but can you please provide me some more information about what they should do for this configuration?  They definitely know more about how their system and the scheduler works, but they have told me before that they do not have anyone in the support team who is familiar with HFSS.

     

    Regards,

    Vasilis

  • tsiriakstsiriaks 3240 El Camino Real #290, Irvine, CA 92602Forum Coordinator
    edited July 22

    Hi Vasilis,

    Your script does the request to SLURM first to get the node list and put that to HFSS batchsolve command. This would need the first node on that list to run Ansoft RSM service, and hence you are seeing this error. There was a discussion about this, ref:

    https://studentcommunity.ansys.com/thread/could-not-start-the-memory-inquiry-solver-when-trying-to-do-multinode-hfss/?order=all#comment-e062d205-0b5a-4b8b-b9db-abc201448788

    and the user brought up the method described in

    https://chpc.utah.edu/documentation/software/ansys-edt.php#ub

    which dynamically starts ansoft RSM on the fly, then stops once the job is done.

    However, as stated in that post, we do have our own customized script to integrate AEDT with SLURM, so you won't have to manually query nodes then start Ansoft RSM accordingly like this. This is the SLURM custom scheduler proxy that Matt mentioned above.

    Let us know if you are interested and we will reach out and provide it to you.

    Thanks,

    Win

  • vataloggvatalogg Member
    edited July 22

    Thank you Win for the information. Things seem to make more sense now...

     

    For the solution with the dynamic allocation, I tried it based on the script from Utah Univ. I can ssh into the assigned node, while it is running, but I have some problem locating the file (or service?) called ansoftrsmservice that I need to start. I run the script from a personal $scratch folder where I have the rights to write files. But, of course, there is no $scratch/Linux directory. Then, I tried to find this file into the installation folder of the software ~/ansys.19.2/AnsysEM19.2/Linux64/, where there was an "ansoftrsm" but no ansoftrsmservice. The other file for registering the node can be found in that directory, but I get an error of:

    "Registering product-specific engines with RSM...

    Error: RSM is not running on this node."

     

    For the other solution with the custom script, I would be interested and it may make things less complicated for me or other users in the future. I also contacted SciNET and they are eager to configure it.

     

    Best regards,

    Vasilis

     

  • tsiriakstsiriaks 3240 El Camino Real #290, Irvine, CA 92602Forum Coordinator
    edited July 23

    Hi Vasilis,

    Just to answer your question, the Ansoft RSM is a separate installation. You have to rerun the installer again and select the option on the first page to install the RSM.

    Let me reach out and send the custom script to you

    Thanks,

    Win

  • vataloggvatalogg Member
    edited July 24

    Hello Win,

    Thank you for your answer. Indeed it seems that RSM is not installed for ansys/19.2 in their system. I was able to start rsm and register nodes for a previous version (19.0), where it is installed and there is a separate rsm folder in the installation directory... So, this seems resolved.

    And also thank you for the scripts!

    Best regards,

    Vasilis

  • tsiriakstsiriaks 3240 El Camino Real #290, Irvine, CA 92602Forum Coordinator

    Hi Vasilis,

    Great to hear ! I'm glad things are working for you now.

    Thanks,

    Win

  • Hi Win;

    Could you send me the customized scripts for SLURM please. We have same error "...Unable to locate or start COM engine on <computenode> : Unable to reach AnsoftRSMService...."

    Thanks,

    Taner

  • tsiriakstsiriaks 3240 El Camino Real #290, Irvine, CA 92602Forum Coordinator

    Hi Taner,

    I believe you are a commercial customer according to your email address. Unfortunately, due to certain support policy, I can't reach out to you directly.

    Could you create a support request (SR) ticket with us, then let me know here that you have created it in our systems ? Please do not post the SR ticket number here. Just let me know it's created and I will track it down.

    Thank you,

    Win

  • Hi Win;

    I have created SR. But by using different account name: hpc_admins at <same domain>


    Best Regards;

  • tsiriakstsiriaks 3240 El Camino Real #290, Irvine, CA 92602Forum Coordinator

    Thanks Taner. I've seen your SR's.

    I'm inquiring this internally on how to handle them. You will hear from us soon.

    Thanks,

    Win

Sign In or Register to comment.