fluent on remote linux cluster
 

Hi all
I need to use parallel fluent on remote cluster...SHARCNET if you are aware...
I want to use my fluent on windows to connect to the cluster and submit jobs from windows without using the command prompt like putty.exe


I do not know how to set up parallel fluent launcher and I would appreciate if someone could guide me through the process

This may not be as easy as you think. It depends on the cluster setup. Most clusters will run a job scheduler, so you cannot just "run" programs as you wish. You can usually make a request for an interactive job, which will dedicate nodes to you to use interactively.

You may be able to connect to the nodes from your local machine, but there are usually security firewalls in place to stop attacks, and these firewalls will usually stop Fluent child processes as well (rsh).

See if your head node has VNC open. If you can VNC into the head node with an interactive job, you will be able to run Fluent with all the pretty GUI buttons.

Good Luck,

Stu

Stu,
I have been using the NX client for gui access to these nodes to run fluent and icemcfd in standalone mode.....The problem is that these nodes do not have workbench installed on them...and I want to submit fluent jobs....so I was thinking if I could do all the preprocessing in workbench and then use nodes for simulation only...since they have fluent installed

I am aware that some systems allow remote access....
when I connect to these systems from putty or nx client...the addresses are like...

for example: saw.sharcnet.ca

I have tried giving this as the remote cluster head node....but its not working...
I also do not know what to give in the field of Remote fluent root path....

I would appreciate of you could guide me through this....

This is a very inefficient way of running your simulation. You will have to reserve as many nodes as you need, and they will be sitting idle while you setup your problem. You would be better to specify your problem beforehand and then upload the .cas and .msh files to the cluster, or otherwise write a script that will automatically setup your problem from your mesh.

From the sharcnet.ca website, I can see that they are running some custom job scheduler (SQ!?!?) If this is the same as PBS, you can easily open an interactive job. "sqsub -I -l 16" should give you 16 cores to work on. The compute nodes may not run X11 though, so you may still need to do it command line style anyway.

I think your sysadmin would be alot more helpful on this topic.


Stu

Stu,

My setup does not take a lot of time ...maybe as I use journal file...I have used the command line submission but I just too much hassel....I need to upload my case first...

Well thanks for the comments and advice anyways:) I have contatced the guys from sharcnet and ANSYS..waiting for their reply!