Welcome, Guest
Username: Password: Remember me

TOPIC: Installation problem on Cluster with SLURM-scheduler

Installation problem on Cluster with SLURM-scheduler 7 years 1 week ago #28184

  • Karl
  • Karl's Avatar
  • OFFLINE
  • Fresh Boarder
  • Posts: 6
  • Thank you received: 1
Hello,

I tried to install Telemac-2d on our Linux-Cluster at TU Munich and failed. The cluster Coolmuc2 uses a SLURM-scheduler.

Serial execution of Telemac-2d is possible. When I try to split the domain Partel gives the following error: partel-crash.png

I used the following Config-file: systel.COOLMUC2.cfg. In Partitioner_para.f I removed the Comm-Varaible from P_ALLGATHERV_I and in Partel.f I changed the format FMT4 from I7 to I9.

The following modules are loaded:

Currently Loaded Modulefiles:
1) admin/1.0 3) mkl/11.3 5) lrz/default 7) gcc/.4.7 9) geos/3.3.3 11) pythonLib/2.7.6
2) tempdir/1.0 4) mpi.intel/5.1 6) intel/17.0 8) obspy/0.9.2 10) mpi4py/1.3.1 12) python/2.7.6

Does anyone know how I can avoid MPI usage to get Partel working?

Thanks a lot in advance!

Karl
Attachments:
The administrator has disabled public write access.

Installation problem on Cluster with SLURM-scheduler 7 years 4 days ago #28192

  • konsonaut
  • konsonaut's Avatar
  • OFFLINE
  • openTELEMAC Guru
  • Posts: 413
  • Thank you received: 144
Hello Karl,

for scalar partitioning I would try to remove the parmetis libraries in your config file. For assistance, have a look on the config file enclosed which we used at the Vienna Scientific Cluster VSC one year ago. It uses the SLURM scheduler, too. Please note that some parts are adapted specifically to the VSC.

At that time other users and me tried the parallel partitioning but I was not successfull, see this forum thread: www.opentelemac.org/index.php/kunena/12-...support?limitstart=0
Maybe in the meantime some problems have been resolved by the developers..

Hope this helps,
Clemens
Attachments:
The administrator has disabled public write access.
The following user(s) said Thank You: Karl

Installation problem on Cluster with SLURM-scheduler 6 years 11 months ago #28244

  • Karl
  • Karl's Avatar
  • OFFLINE
  • Fresh Boarder
  • Posts: 6
  • Thank you received: 1
Hello Clemens,

thank your very much for your help. It solved half of my problem. The other part was caused by a wrong choosen library for Partel. Now my program is running.

Thanks a lot!

Karl
The administrator has disabled public write access.
Moderators: pham

The open TELEMAC-MASCARET template for Joomla!2.5, the HTML 4 version.