I’m trying out fireworks with a cluster which uses the multi-cluster feature of the SLURM queuing system.
Though I can customize the script and to include the “clusters” option.
It seems that I have to tweak the CommonAdapter to work with this setup.
- When submitting jobs with the “-M” option, SLURM reports the job id like this,
sbatch -A snic2018-1-33 -M snowy test
Submitted batch job 257467 on cluster snowy
I think this makes CommonAdaptor fail to retrieve the job id:
- When querying job info with the “-M” option, a extra line is printed,
squeue -u yunqi -M snowy -h
257471 core test yunqi R 0:05 1 s156
I think this will cause the adaptor to get the wrong number of jobs:
Will it be of interest to modify the CommonAdaptor to support multi-cluster?