A PostgreSQL backup between nodes in a streaming replication cluster

postgresql_backup.sh

#!/bin/bash

DEST_HOST=${1}
DATA_DIRECTORY=${2:-"/srv/data/postgresql-data"}
ARCHIVE_DIRECTORY=${3:-"/srv/data/postgresql-archives"}

[[ -n "${DEST_HOST}" ]] || exit 1

##
# Prepare for a local backup
/usr/bin/psql --command "SELECT pg_start_backup('Streaming Replication', true)" --dbname postgres

##
# Prepare remote postgresql server for slave-ification
/usr/bin/ssh postgres@${DEST_HOST} "/etc/init.d/postgresql stop; rm ${DATA_DIRECTORY}/FAILOVER; mv ${DATA_DIRECTORY}/recovery.done ${DATA_DIRECTORY}/recovery.conf"

##
# Copy local data to remove postgres server
/usr/bin/rsync -C -a --delete -e ssh --exclude pg_log --exclude pg_xlog --exclude recovery.conf --exclude recovery.done ${DATA_DIRECTORY}/ ${DEST_HOST}:${DATA_DIRECTORY}/

##
# This archives the the WAL log (ends writing to it and moves it to the $archive dir
/usr/bin/psql --command "SELECT pg_stop_backup()" --dbname postgres

##
# this rsyncs the WAL archives that are written after pg_stop_backup is called.
/usr/bin/rsync -C -a --delete -e ssh ${ARCHIVE_DIRECTORY}/ ${DEST_HOST}:${ARCHIVE_DIRECTORY}/

##
# Start postgres (and replication) on the remote slave database
/usr/bin/ssh postgres@${DEST_HOST} "/etc/init.d/postgresql start"