WAL-G Full backup with healtchecks.io

gistfile1.txt

#!/usr/bin/env ksh93

# This script determines if it is running on the primary PostgreSQL node and then starts a full backup. It does nothing
# when running on PostgreSQL nodes that are replica's.

# Keep PID detection at the top of script to prevent it from breaking (getting wrong values).
OWN_PID=$$
PARENT_PID=$(ps -o ppid= -p "${OWN_PID}")

SHELL_NAME="ksh93"
SCRIPT_NAME=$0
# Optional UUID of the healthchecks.io check to report status to.
HEALTHCHECKS_IO_UUID=$1
HEALTHCHECKS_IO_URL="https://hc-ping.com"

WALG_BIN="/opt/walg/wal-g"
WALG_CONF_DIR="/etc/walg"
POSTGRESQL_DATA_DIR="/var/lib/patroni/postgres"

# We put the shell name in front of the script name so pgrep doesn't find the parent shell.
for RUNNING_PID in $(pgrep -f "${SHELL_NAME} ${SCRIPT_NAME}"); do
  if [ "${RUNNING_PID}" != "${OWN_PID}" ] && [ "${RUNNING_PID}" != "${PARENT_PID}" ]; then
    echo "ERROR: Process already running: ${RUNNING_PID}"
    exit 1
  fi
done

# Get the current role (master, replica) of the node.
ROLE=$(curl --connect-timeout 20 -s http://127.0.0.1:8009/primary | jq --raw-output .role)

# We only want to run on the master node. Not on replica's.
if [ "${ROLE}" = "master" ]; then
  : # Simply continue.
elif  [ "${ROLE}" = "replica" ]; then
  # Nothing to do. Only the master must make the full backup.
  exit 0
else
  echo "ERROR: Server is in an unknown role ${ROLE} (not 'master' or 'replica')."
  exit 1
fi

if [[ -n "${HEALTHCHECKS_IO_UUID}" ]]; then
  # If a healthchecks.io check UUID is provided signal that the job has started.
  curl -fsS -m 10 --retry 5 -o /dev/null "${HEALTHCHECKS_IO_URL}/${HEALTHCHECKS_IO_UUID}/start"
fi

# Start the actual backup.
${WALG_BIN} --config ${WALG_CONF_DIR}/walg.json backup-push ${POSTGRESQL_DATA_DIR}

BACKUP_EXIT_CODE=$?
if [[ -n "${HEALTHCHECKS_IO_UUID}" ]]; then
  # If a healthchecks.io check UUID is provided report the exit code of the backup process.
  curl -fsS -m 10 --retry 5 -o /dev/null "${HEALTHCHECKS_IO_URL}/${HEALTHCHECKS_IO_UUID}/${BACKUP_EXIT_CODE}"
fi

# Exit with the code of the backup process so cron also knows if the process failed or succeeded.
exit ${BACKUP_EXIT_CODE}