run.sh

#!/usr/bin/env bash

set -xe

# check input arg(s)
if [ $# -lt "3" ]; then
  echo "Usage: $0 [ECHO run] [MODE {s,p}] [SCHEDULER {gridengine,slurm}]"
  exit 1
fi

ECHO=$1
MODE=$2
SCH=$3
CALLERS=(manta delly lumpy gridss)
STR_CALLERS="[$(printf "'%s'," "${CALLERS[@]}"|sed 's/,$//')]"
JOBS=() # array of job IDs
JOBS_LOG=jobs.json # job accounting log
SAMPLES=samples.csv
USE_CONDA=$([ "$ECHO" -eq "0" ] && echo "--use-conda" || echo "")
MY_ENV=wf

monitor () {  # monitor a job via Xenon CLI
  xenon -v --json scheduler $SCH --location local:// list --identifier $1
}

# activate conda env
eval "$(conda shell.bash hook)"
conda activate $MY_ENV
conda list

# run workflow
cd snakemake && ls -alh
echo "Selected callers: $STR_CALLERS"
snakemake -C echo_run=$ECHO samples=$SAMPLES mode=$MODE \
  enable_callers="$STR_CALLERS" $USE_CONDA \
  --configfile analysis_test.yaml \
  --latency-wait 60 --jobs \
  --cluster "xenon -vvv scheduler $SCH --location local:// submit \
  --name smk.{rule} --cores-per-task {threads} --inherit-env \
  --max-run-time 15 --working-directory . \
  --stderr stderr-%j.log --stdout stdout-%j.log"

# check SV callers' output in VCF files
echo "-----------------"
echo "VCF output files:"
if [ "$ECHO" -eq "0" ]; then
  for caller in "${CALLERS[@]}"; do
    VCF_FILE="$(find data -mindepth 6 -name $caller.vcf)"
    BOOL=$([ -e  "$VCF_FILE" ] && echo 0 || echo 1)
    INFO=$([ $BOOL -eq 0 ] && echo "$VCF_FILE" || echo "None")
    echo " $caller: $INFO"
  done
  echo " merge:" $(find data -name all.vcf)
fi

# show logs
echo "----------"
echo "Log files:"
for f in *.log; do
  echo -e "\n### $f ###\n"
  cat $f
  JOB_ID=$(echo $f | grep err | sed 's/\w*-\([0-9]*\).\w*/\1/')
  JOBS+=($JOB_ID)
done

# collect job accounting info
for j in ${JOBS[@]}; do
  monitor $j >> $JOBS_LOG
done
cat $JOBS_LOG

# exit with non-zero if there are failed jobs
[[ $(jq ".statuses | .[] | select(.done==true and .exitCode!=0)" $JOBS_LOG) ]] \
  && exit 1 || exit 0