radical-collaboration / hpc-workflows

NSF16514 EarthCube Project - Award Number:1639694
5 stars 0 forks source link

Supporting ORNL Andes #128

Closed mturilli closed 3 years ago

mturilli commented 3 years ago

https://www.olcf.ornl.gov/olcf-resources/compute-systems/andes/

wjlei1990 commented 3 years ago

In radial.saga, the self.rm.host() returns localhost. So saga is not aware of the system.

After some modification on the source code, the radical.saga now works on andes.

Plan to move to radical.pilot and radical.ensemble soon.

andre-merzky commented 3 years ago

After some modification on the source code, the radical.saga now works on andes.

What modifications are those?

wjlei1990 commented 3 years ago

The code change I made are in the file of adaptors/slurm/slurm_job.py.

Since self.rm.host are localhost. So I manually set the value to andes:

self.rm.host = "andes"

There are a few changes to the header information.

if 'andes' in self.rm.host.lower(): 
    script += "#SBATCH --chdir %s\n"   % cwd    # instead of --workdir

if 'andes' in self.rm.host.lower(): 
    script += "#SBATCH -N %d\n" % n_nodes
wjlei1990 commented 3 years ago

pull requests submitted https://github.com/radical-cybertools/radical.saga/pull/823