Learning Forum

[andrea.riva]

Good morning,

I have to generate a big mesh (370 million elements) on fluent meshing. I can launch the mesh generation on a remote linux Cluster. The cluster has several computing nodes, each of which has certain number of cores and a certain RAM. When I submit my job, a queue manager assign my job to the free nodes and the job starts. I have the possibility to choose the number of cores and the distribution of cores among nodes. Eg: I can choose to run a job on 30 cores belonging to a single node OR to run a job on 30 cores distrubuted on 3 computing nodes (10 cores for each node). With the second option I should have more available RAM, since each computing node has his own ram (about 250 GB). However, what I see is that when I run my job the largest part of the RAM is allocated to a single node (and not divided between the three nodes). So, the mesh generation fails with an "out of memory" error.

Is there a string to allocate better the RAM and distribute it on different computing nodes? Is there a way to do that from a command line?

Following, my job-submission file

#!/bin.bash # use bash as command interpreter

#$ -cwd                 # currentWorkingDirectory

#$ -N HCT_mesh        # jobName

#$ -j y                 # merges output and errors

#$ -S /bin/bash         # scripting language

#$ -l h_rt=168:00:00    # jobDuration hh:mm:ss

#$ -q nmec.q            # queueName

#$ -pe mpi_10 30        # cpuNumber

#________________________________________________________

/software/ansys2023R2/v232/fluent/bin/fluent 3ddp -t30  -meshing -cnf=machinefile.$JOB_ID -g < inputfile.txt > outputfile.txt 

echo End Parallel Run

Thank you very much 

 

Andrea 

 

[Federico]

Hello, parts of the meshing workflows in Fluent Meshing are not yet parallelized, such as the Surface Mesh task, and can therefore only be run in serial