Learning Forum

[Atul Singh]

Hi,

First off, yes I do have an HPC license and a research license. Here are my starting questions.

1) Opening the fluent session, If I have to run parallel, I am asked to enter GPGPUs, and then I can sleect run types where I could select "Shared memory on local machine " and "distribuited memory on cluster". In the later case, does the GPGPU number needs to be on what will be on the "distributed memory on cluster" or the one of my local machine?

2) The instruction manual gives instructions for "Starting parallel ansys on windows system" and "Starting parallel ansys on linux system", however, If I want to run from windows machine to linux cluster, first, is that even possible, and second, which instruction do I follow on it (version22).

3) The section requires me to attach a file containing machine names. How do i get this from the linux cluster? The pdf only gives examples Machine1, Machine2 etc, but what is the command to find the name of the cluster that I will push my job to? Or rather how do I know my cluster's machine name?
also, there is a queue in place, which assigns jobs as per availability, so how can I know beforehand what will be the "machine names"? If you know what i mean? Or is this something else?
Is this name, should be the login node, or the result of some other command. 
Please do give elaborate answer for this?

4) The instructions manual also writes (p4006) that a password less ssh needs to be set by enabling the "Remote cluster head node". Doesn't expalain how!!!
Does it mean, that from my cmd (for windows) I should be able to login to my cluster passwordless, and after it, Fluent will not require any other settings for this? Or something else?

5)Same questions for the scheduler.
We use a slurm job scheduler. The settings required here are "Slurm Submission host". Instructions does leave the possibility to leave this empty, however the question is similar. How can I know the details of my host, before I have been assigned the resource. Shouldn't this be automatically set somewhere once the resources request has been approved?

Please help

[Swathi V. V.]

Hi, 

To answer your first question, you can keep GPGPU input as 0 if you are not interested in using GPU solver. For more infromation in GPGPU, please refer to 38.6. Using General Purpose Graphics Processing Units (GPGPUs) With the Algebraic Multigrid (AMG) Solver (ansys.com)

2. You can share case and data file from simaultion done in windows and transfer it to the Linux system. These files can then be used to run simualtion in Linux system. There are many other ways to submit job from your local system to remote system. Please go through the user manual for the same.

Please check with your systems engineer to know the details of your cluster and machine specs. 

 

[Atul Singh]

It took you 6 days, to answer with "read through the manual"????????????????

I specifically quote that in the manual those things are not given properly. (see p4006 is written in the brackets), which should tell you , that I already have. !!

[Kalyan Goparaju]

Hello, 

The following reponse is assuming that you have access to an IT maintained cluster, with a job scheduler. If these are not true, the below response might not be valid. 

1. Selecting GPGPU's is not mandatory. If you don't have a reason and/or resources to use them, you can set this value to 0.
2. Please refer to User's guide (section 38.2 - Starting Parallel Ansys Fluent Using Fluent Launcher) - This section discusses how to setup Fluent to run on remote Linux nodes.
3. Based on your point 5, you have a job schueduler on your cluster. If you are running by turning on the 'Use Job Scheduler' option, you don't have to specify the machine names file. 
4. Please contact your IT for setting up the passwordless SSH. This is not Ansys or Fluent specific and hence is not a part of the manual.
5. Slurm Scheduler Host should be the login node for your cluster that can communicate with the SLURM job scheduler. Do not confuse this with the simulation nodes that are assigned when you request for resources. 

Much of what I have outlined above is covered in great detail in the User's Guide (38.2).

Thanks,

Kalyan