Installation on cluster

All questions about installation
Forum rules
Please read the forum usage recommendations before posting.
Puneeth
Posts: 11
Joined: Thu Oct 18, 2018 8:37 am

Re: Installation on cluster

Post by Puneeth »

Hello,

The admin recompiled Code-Saturne as per your recommendations and this has fixed the issue!

Thank you very much!

Sincerely,

Puneeth
Puneeth
Posts: 11
Joined: Thu Oct 18, 2018 8:37 am

Re: Installation on cluster

Post by Puneeth »

Hello,

After reinstaling CSv5.1.6, there is a new error relating to the srun command (SLURM Workload Manager).
The slurm-jobid.out file contents are as follows:
Code_Saturne
************
Version: 5.1
Path: /gpfslocalsup/pub/code-saturne/5.1.6
Result directory:
/gpfs7kw/linkhome/rech/genlma01/upz85tg/CH590_A1_alpha/CASE1/RESU/20200923-1111
****************************************
Compiling user subroutines and linking
****************************************
****************************
Preparing calculation data
****************************
Parallel Code_Saturne on 160 processes.
***************************
Preprocessing calculation
***************************
**********************
Starting calculation
**********************
srun: error: Invalid numeric value "40.0" for ntasks-per-node.
****************************
Saving calculation results
****************************
solver script exited with status 1.
Error running the calculation.
Check Code_Saturne log (listing) and error* files for details.
Error in calculation stage.

The ntasks-per-node should ideally be an integer and not a floating point value as seen here.
There are no listing or error files created.

Could you please guide me on how to fix this?

Thanks and regards,

Puneeth
Attachments
summary.txt
(33.58 KiB) Downloaded 51 times
Puneeth
Posts: 11
Joined: Thu Oct 18, 2018 8:37 am

Re: Installation on cluster

Post by Puneeth »

Hello,

Found the bug that generated the error relating to srun.
It is from the function to compute n_procs_per_node within the resource_info class at line 1086 in file cs_exec_environment.py. The division operation is not floored (using //) and instead uses "/" which returns a floating point value in python 3.x.

Thanks and regards,

Puneeth
Yvan Fournier
Posts: 3424
Joined: Mon Feb 20, 2012 3:25 pm

Re: Installation on cluster

Post by Yvan Fournier »

Hello,

Thanks for the feedback. This was fixed November 2018 for the v6.0 and master branches, but was not backported to v5.0.

I'll fix it in the next v5.0 release (but there might be other Python 3 bugs in the GUI), which will be maintained until spring of 2021 (until v7.0 is released).

Best regards,

Yvan
Post Reply