A yml and a set of instructions to build a functioning environment for the Research Computing

clone this repo and update with the job composer

Copy and paste the following job script into a job composer job on rc.uab.edu

#!/bin/bash
#SBATCH --partition=express
#SBATCH --mem-per-cpu=4000
module load cuda10.0/toolkit
module load Anaconda3

FOLDER=/data/user/$USER/slurm-ds
URL=https://gitlab.rc.uab.edu/rc-data-science/createandparsesacct.git
if [ ! -d "$FOLDER" ] ; then
    git clone "$URL" "$FOLDER"
conda env create -f /data/user/$USER/slurm-ds/environment-slurm-ds.yml
else
    cd $FOLDER
    git pull "$URL"
    conda env update -f /data/user/$USER/slurm-ds/environment-slurm-ds.yml
fi

Configuring the environment

After the environment is created, you can start up an interactive Jupyter notebook session through rc.uab.edu to check if the environment works.

Under environment setup, specify

# Load required modules
module load cuda10.0/toolkit
module load Anaconda3/2019.10

Under Extra jupyter arguments, specify

--notebook-dir=/data/user/$USER/slurm-ds

For partition, set partition to

express

for time up to 2 hours. Also make sure to specify the number of hours field to match. For up to 12 hours, the

short

partition can be used.

After the Jupyter notebook is started, click on the blue "Connect to Jupyter" button.

Once the Jupyter session is active, select the slurm-2sql notebook. Then change the kernel, via Kernel->Change kernel->Python [conda env:.conda-slurm-ds]

Verify the environment loaded correctly by running the first cell of the slurm-2sql notebook (with the library imports)

Creating a text version of sacct output

If we have to create a database from sacct

directoryToUse="/data/user/$USER/group"
sacct -P -u $USER --starttime=2019-01-01 --format user,start,jobid,jobname,state,partition,maxrss,reqmem,reqcpus,node,nnodes,elapsed >> "$directoryToUse"group.txt