slurm related

aliases

## query my jobs

```
alias fsa='sshare -A users -a |rg -v 0.000000 |sort -k7nr | grep -E "dzhou|santra|lwang|bier|cwu|jacobson|$"'
alias pi='srun --partition=compute-16core-64gb-preemptible --job-name=interactive --nodes=1 --ntasks-per-node=1 --pty bash'
alias q='squeue -u $USER | tee >(tail -n +2 |lc)'
#alias qq='squeue -p gpu-t4-4x-ondemand -t pd |tee >(tail -n +2 |lc)'
alias qqq='squeue -p gpu-t4-4x-ondemand-high-pri -t pd |tee >(tail -n +2 |lc)'
alias qd='scancel `q | tail -n +2| fzf | ff 1`'
alias qa='sacct --format=User,JobID,Jobname,partition,state,time,start,end,elapsed,WorkDir,MaxRss,MaxVMSize,nnodes,ncpus,nodelist -j'
alias qr='squeue -u $USER |rg'
alias si='sinfo | rg -v driver |rg -v session|rg -v gpu-t4-4x-preemptible |rg -v gpu-t4-4x-ondemand-h'

export SQUEUE_FORMAT="%.10i %.50j %.10u %.10T %.10M %.11l %.6D %R"
```

ffa

awk '{ sum += $'$1' } END { if (NR > 0) print sum / NR }'

ffs

awk ${@:2} '{ s += $'$1' } END { print s }'

mm

#!/bin/bash

wait_time=20
user=${2-$USER}

while true
do
    got=$(squeue -u $user | grep $1)
    n_running=$(echo "$got" |grep -c 'RUNNING')
    n_bad=$(echo "$got" |grep -cv 'RUNNING')
    [ -z "$got" ] && n_bad=0
    status=${n_running}-${n_bad}
    #echo $status
    if [[ $status != $old_status ]]; then
        echo $n_running running
        echo $n_bad not running
        echo '-----------------' `date +%R`
    fi
    [[ $status == '0-0' ]] && exit 0;
    old_status=$status
    sleep $wait_time
done
echo done

qq

t=${2-running}

case $1 in
gpu) p=gpu-t4-4x-ondemand;;
gpuh) p=gpu-t4-4x-ondemand-high-pri;;
gpup) p=gpu-t4-4x-preemptible;;
cpu) p=compute-16core-64gb-ondemand;;
driver) p=driver-4core-standard;;
*) p=compute-16core-64gb-preemptible;;
esac

squeue -p $p -O UserName -t $t -h| sort | uniq -c