<> ------ = Submitting jobs = ---- Whatever you read here may need to be adjusted to fit your specific case. Do not hesitate to ask for some help when needed. == Filesystems == Some but not all partitions are available to the compute nodes. Compute nodes will not be able to access any data from filesystems that are not listed here. /work /scratch /home/ == Slurm partitions == There are currently 2 partitions, '''normal''' and '''bigmem'''. The '''normal''' partition is the default partition if you submit a job without precising witch partition should be used you job will be placed in one of the normal partition. The normal partition has limited RAM of 250GB, in case you need more than that please use the '''bigmem''' partition. Use ''-p'' option can be used to specify the needed partition. == Load necessary software == By default only some software will be available when login. To be able use other software scripts you should first load then. The command '''module''' will help you to manage the modules dependencies. To to check which software are installed, can be used after importing, {{{ module list }}} Expected output is, {{{ Currently Loaded Modules: 1) autotools 2) prun/1.3 3) gnu8/8.3.0 4) openmpi3/3.1.4 5) ohpc }}} to check which software are available, ready to be used without importing anything. The same command can be used to search a specific package. {{{ module avail }}} Expected output is, {{{ --------------------------------------------------- /opt/ohpc/pub/moduledeps/gnu8-openmpi3 ---------------------------------------------------- adios/1.13.1 hypre/2.18.1 netcdf-cxx/4.3.1 petsc/3.12.0 py2-scipy/1.2.1 scorep/6.0 trilinos/12.14.1 boost/1.71.0 imb/2018.1 netcdf-fortran/4.5.2 phdf5/1.10.5 py3-mpi4py/3.0.1 sionlib/1.7.4 dimemas/5.4.1 mfem/4.0 netcdf/4.7.1 pnetcdf/1.12.0 py3-scipy/1.2.1 slepc/3.12.0 extrae/3.7.0 mpiP/3.4.1 omb/5.6.2 ptscotch/6.0.6 scalapack/2.0.2 superlu_dist/6.1.1 fftw/3.3.8 mumps/5.2.1 opencoarrays/2.8.0 py2-mpi4py/3.0.2 scalasca/2.5 tau/2.28 -------------------------------------------------------- /opt/ohpc/pub/moduledeps/gnu8 -------------------------------------------------------- hdf5/1.10.5 metis/5.1.0 mvapich2/2.3.2 openblas/0.3.7 pdtoolkit/3.25 py3-numpy/1.15.3 likwid/4.3.4 mpich/3.3.1 ocr/1.0.1 openmpi3/3.1.4 (L) py2-numpy/1.15.3 superlu/5.2.1 ------------------------------------------------------------- /tools/modulefiles -------------------------------------------------------------- MEGAHIT/1.2.9 ---------------------------------------------------------- /opt/ohpc/pub/modulefiles ---------------------------------------------------------- EasyBuild/3.9.4 clustershell/1.8.2 gnu7/7.3.0 llvm5/5.0.1 pmix/2.2.2 valgrind/3.15.0 autotools (L) cmake/3.15.4 gnu8/8.3.0 (L) ohpc (L) prun/1.3 (L) charliecloud/0.11 gnu/5.4.0 hwloc/2.1.0 papi/5.7.0 singularity/3.4.1 Where: L: Module is loaded Use "module spider" to find all possible modules. Use "module keyword key1 key2 ..." to search for all possible modules matching any of the "keys". }}} To search for a module, {{{ module avail <> #OR module spider <> }}} To load a module do, {{{ module load <> }}} Loading a module can be done following those 3 steps, 1. Locate the module, module avail 2. Check how to load it, module spider <> 3. Load you module using the instructions from step 2 Read more about '''module''' usage https://lmod.readthedocs.io/en/latest/010_user.html == Prototype of batch script == This prototype should be in a script file, for example, my_first_script.sbatch {{{#!highlight bash #!/bin/bash #SBATCH -J test # Job name #SBATCH -o /work/<>/job.%j.out # Name of stdout output file (%j expands to jobId) #SBATCH -e /work/<>/job.%j.err # Name of stderr output file (%j expands to jobId) #SBATCH -p normal # Partition to use, another possible value is bigmem #SBATCH -N 1 # Total number of nodes requested #SBATCH -n 16 # Total number of cpu requested or total number of mpi tasks #SBATCH -t 01:30:00 # Run time ([d-]hh:mm:ss) - 1.5 hours # Load your software/command module load CMD/version # Run your command CMD [OPTIONS] ARGUMENTS }}} To run a sbatch script use {{{ sbatch <