#!/bin/bash
#
# Nagios plugin to check Postgresql streamin replication state
#
# Could be use on Master or on standby node
#
# Requirements:
#
#     Some CLI tools: awk, sed, bc, psql and pg_lscluster
#
#     On master node: Slaves must be able to connect with user from recovery.conf
#                     (or user specify using -U) to database with the same name
#                     (or another specified with -D) as trust (or via md5 using
#                     password specified in ~/.pgpass). This user must have
#                     SUPERUSER privilege (need to get replication details).
#
#     On standby node: PG_USER must be able to connect localy on the database
#                      with the same name (or another specified with -D) as trust
#                      (or via md5 using password specified in ~/.pgpass).
#
# Author: Benjamin Renard <brenard@easter-eggs.com>
# Date: Wed, 04 Nov 2020 15:31:13 +0100
# Source: https://gogs.zionetrix.net/bn8/check_pg_streaming_replication
# SPDX-License-Identifier: GPL-3.0-or-later
#

DEFAULT_PG_USER=postgres
DEFAULT_PG_VERSION=9.1
DEFAULT_PG_MAIN=/var/lib/postgresql/$PG_VERSION/main
DEFAULT_PG_PORT=5432
PG_USER=""
PG_VERSION=""
PG_MAIN=""
PG_MASTER_USER=""
PSQL_BIN=/usr/bin/psql
PG_LSCLUSTER_BIN=/usr/bin/pg_lsclusters
RECOVERY_CONF_FILENAME=recovery.conf
RECOVERY_CONF=""
PG_DEFAULT_PORT=""
PG_DEFAULT_APP_NAME=$( hostname )
PG_DB=""
CHECK_CUR_MASTER_XLOG=1
REPLAY_WARNING_DELAY=3
REPLAY_CRITICAL_DELAY=5

DEBUG=0

function usage () {
	cat << EOF
Usage : $0 [-d] [-h] [options]
	-u pg_user		Specify local Postgres user (Default: try to auto-detect or use $DEFAULT_PG_USER)
	-b psql_bin		Specify psql binary path (Default: $PSQL_BIN)
	-B pg_lsclusters_bin	Specify pg_lsclusters binary path (Default: $PG_LSCLUSTER_BIN)
	-V pg_version		Specify Postgres version (Default: try to auto-detect or use $DEFAULT_PG_VERSION)
	-m pg_main		Specify Postgres main directory path (Default: try to auto-detect or use
                                $DEFAULT_PG_MAIN)
	-r recovery_conf	Specify Postgres recovery configuration file path
				(Default: [PG_MAIN]/$RECOVERY_CONF_FILENAME)
	-U pg_master_user	Specify Postgres user to use on master (Default: user from recovery.conf file)
	-p pg_port		Specify default Postgres master TCP port (Default: same as local PostgreSQL
	                        port if detected or use $DEFAULT_PG_PORT)
	-D dbname		Specify DB name on Postgres master/slave to connect on (Default: PG_USER, must
	                        match with .pgpass one is used)
	-C 1/0			Enable or disable check if the current XLOG file of the master host is the same
				of the last replay XLOG file (Default: $CHECK_CUR_MASTER_XLOG)
	-w replay_warn_delay	Specify the replay warning delay in second (Default: $REPLAY_WARNING_DELAY)
	-c replay_crit_delay	Specify the replay critical delay in second (Default: $REPLAY_CRITICAL_DELAY)
	-d			Debug mode
	-h 			Show this message
EOF
	exit 0
}

while getopts "hu:b:B:V:m:r:U:p:D:C:w:c:d" OPTION
do
	case $OPTION in
		u)
			PG_USER=$OPTARG
		;;
		b)
			PSQL_BIN=$OPTARG
		;;
		B)
			PG_LSCLUSTER_BIN=$OPTARG
		;;
		V)
			PG_VERSION=$OPTARG
		;;
		m)
			PG_MAIN=$OPTARG
		;;
		r)
			RECOVERY_CONF=$OPTARG
		;;
		U)
			PG_MASTER_USER=$OPTARG
		;;
		p)
			PG_DEFAULT_PORT=$OPTARG
		;;
		D)
			PG_DB=$OPTARG
		;;
		C)
			CHECK_CUR_MASTER_XLOG=$OPTARG
		;;
		w)
			REPLAY_WARNING_DELAY=$OPTARG
		;;
		c)
			REPLAY_CRITICAL_DELAY=$OPTARG
		;;
		d)
			DEBUG=1
		;;
		h)
			usage
		;;
		\?)
			echo -n "Unkown option"
			usage
	esac
done

function debug() {
	if [ $DEBUG -eq 1 ]
	then
		>&2 echo -e "[DEBUG] $1"
	fi
}

debug "Starting options (before handling auto-detection/default values) :
PG_VERSION = $PG_VERSION
PG_DB = $PG_DB
PG_USER = $PG_USER
PSQL_BIN = $PSQL_BIN
PG_LSCLUSTER_BIN = $PG_LSCLUSTER_BIN
PG_MAIN = $PG_MAIN
RECOVERY_CONF = $RECOVERY_CONF
PG_DEFAULT_PORT = $PG_DEFAULT_PORT
PG_DEFAULT_APP_NAME = $PG_DEFAULT_APP_NAME
CHECK_CUR_MASTER_XLOG = $CHECK_CUR_MASTER_XLOG
REPLAY_WARNING_DELAY = $REPLAY_WARNING_DELAY
REPLAY_CRITICAL_DELAY = $REPLAY_CRITICAL_DELAY
"

# Auto-detect PostgreSQL information using pg_lsclusters
if [ -x "$PG_LSCLUSTER_BIN" ]
then
	PG_CLUSTER=$( $PG_LSCLUSTER_BIN -h 2>/dev/null|head -n1 )
	if [ -n "$PG_CLUSTER" ]
	then
		debug "pg_lsclusters output:\n\t$PG_CLUSTER"
		# Output example:
		# 9.6 main    5432 online,recovery postgres /var/lib/postgresql/9.6/main /var/log/postgresql/postgresql-9.6-main.log
		[ -z "$PG_VERSION" ] && PG_VERSION=$( echo "$PG_CLUSTER"|awk -F ' +' '{print $1}' )
		[ -z "$PG_DEFAULT_PORT" ] && PG_DEFAULT_PORT=$( echo "$PG_CLUSTER"|awk -F ' +' '{print $3}' )
		[ -z "$PG_USER" ] && PG_USER=$( echo "$PG_CLUSTER"|awk -F ' +' '{print $5}' )
		[ -z "$PG_MAIN" ] && PG_MAIN=$( echo "$PG_CLUSTER"|awk -F ' +' '{print $6}' )
	fi
else
	debug "pg_lsclusters not found ($PG_LSCLUSTER_BIN): parameters auto-detection disabled"
fi

# If auto-detection failed, use default values
[ -z "$PG_USER" ] && PG_USER="$DEFAULT_PG_USER"
[ -z "$PG_VERSION" ] && PG_VERSION="$DEFAULT_PG_VERSION"
[ -z "$PG_MAIN" ] && PG_MAIN="$DEFAULT_PG_MAIN"
[ -z "$PG_DEFAULT_PORT" ] && PG_DEFAULT_PORT="$DEFAULT_PG_PORT"

# Check PG_USER
[ -z "$PG_USER" ] && echo "UNKNOWN : Postgres user not specified" && exit 3
id "$PG_USER" > /dev/null 2>&1
[ $? -ne 0 ] && echo "UNKNOWN : Invalid Postgres user ($PG_USER)" && exit 3

# Check PSQL_BIN
[ ! -x "$PSQL_BIN" ] && echo "UNKNOWN : Invalid psql bin path ($PSQL_BIN)" && exit 3

# Check PG_MAIN
[ ! -d "$PG_MAIN/" ] && echo "UNKNOWN : Invalid Postgres main directory path ($PG_MAIN)" && exit 3

# Check RECOVERY_CONF
[ -z "$RECOVERY_CONF" ] && RECOVERY_CONF="$PG_MAIN/$RECOVERY_CONF_FILENAME"

# Check PG_DEFAULT_PORT
[ $( echo "$PG_DEFAULT_PORT"|grep -c -E '^[0-9]*$' ) -ne 1 ] && "UNKNOWN : Postgres default master TCP port must be an integer." && exit 3

# If PG_DB is not provided with -D parameter, use PG_USER as default value
[ -z "$PG_DB" ] && PG_DB="$PG_USER"

function psql_get () {
	sql="$1"
	debug "Exec 'echo \"$sql\"|sudo -u $PG_USER $PSQL_BIN -d \"$PG_DB\" -w -t -P format=unaligned"
	echo "$sql"|sudo -u $PG_USER $PSQL_BIN -d "$PG_DB" -w -t -P format=unaligned
}

function psql_master_get () {
	sql="$1"
	debug "Exec 'echo \"$sql\"|sudo -u $PG_USER $PSQL_BIN -U $M_USER -h $M_HOST -w -p $M_PORT -d $PG_DB -t -P format=unaligned"
	echo "$sql"|sudo -u $PG_USER $PSQL_BIN -U $M_USER -h $M_HOST -w -p $M_PORT -d $PG_DB -t -P format=unaligned
}

debug "Running options :
PG_VERSION = $PG_VERSION
PG_DB = $PG_DB
PG_USER = $PG_USER
PSQL_BIN = $PSQL_BIN
PG_LSCLUSTER_BIN = $PG_LSCLUSTER_BIN
PG_MAIN = $PG_MAIN
RECOVERY_CONF = $RECOVERY_CONF
PG_DEFAULT_PORT = $PG_DEFAULT_PORT
PG_DEFAULT_APP_NAME = $PG_DEFAULT_APP_NAME
CHECK_CUR_MASTER_XLOG = $CHECK_CUR_MASTER_XLOG
REPLAY_WARNING_DELAY = $REPLAY_WARNING_DELAY
REPLAY_CRITICAL_DELAY = $REPLAY_CRITICAL_DELAY
"

# Set some stuff to PostgreSQL version
if [ $( echo "$PG_VERSION < 10" |bc -l ) -eq 1 ]
then
	pg_last_wal_receive_lsn='pg_last_xlog_receive_location()'
	pg_last_wal_replay_lsn='pg_last_xlog_replay_location()'
	pg_current_wal_lsn='pg_current_xlog_location()'
	pg_wal_lsn_diff='pg_xlog_location_diff'
	sent_lsn='sent_location'
	write_lsn='write_location'
else
	pg_last_wal_receive_lsn='pg_last_wal_receive_lsn()'
	pg_last_wal_replay_lsn='pg_last_wal_replay_lsn()'
	pg_current_wal_lsn='pg_current_wal_lsn()'
	pg_wal_lsn_diff='pg_wal_lsn_diff'
	sent_lsn='sent_lsn'
	write_lsn='write_lsn'
fi

# Postgres is running ?
if [ $DEBUG -eq 0 ]
then
	psql_get '\q' 2> /dev/null
else
	psql_get '\q'
fi
if [ $? -ne 0 ]
then
	echo "CRITICAL : Postgres is not running !"
	exit 2
fi
debug "Postgres is running"

RECOVERY_MODE=0
[ "$( psql_get 'SELECT pg_is_in_recovery();' )" == "t" ] && RECOVERY_MODE=1

if [ -f $RECOVERY_CONF ]
then
	debug "File recovery.conf found. Hot-standby mode."
	
	# Check recovery mode
	if [ $RECOVERY_MODE -ne 1 ]
	then
		echo "CRITICAL : Not in recovery mode while recovery.conf file found !"
		exit 2
	fi
	debug "Postgres is in recovery mode"

	LAST_XLOG_RECEIVE=$( psql_get "SELECT $pg_last_wal_receive_lsn" )
	debug "Last xlog file receive : $LAST_XLOG_RECEIVE"
	LAST_XLOG_REPLAY=$( psql_get "SELECT $pg_last_wal_replay_lsn" )
	debug "Last xlog file replay : $LAST_XLOG_REPLAY"


	# Get master connection informations from recovery.conf file
	MASTER_CONN_INFOS=$( egrep '^ *primary_conninfo' $RECOVERY_CONF|sed "s/^ *primary_conninfo *= *\(.\+\) *$/\1/" )
	if [ ! -n "$MASTER_CONN_INFOS" ]
	then
		echo "UNKNOWN : Can't retreive master connection informations form recovery.conf file"
		exit 3
	fi
	debug "Master connection informations : $MASTER_CONN_INFOS"

	M_HOST=$( echo "$MASTER_CONN_INFOS"| grep 'host=' | sed 's/^.*host= *\([0-9a-zA-Z.-]\+\) *.*$/\1/' )
	if [ ! -n "$M_HOST" ]
	then
		echo "UNKNOWN : Can't retreive master host from recovery.conf file"
		exit 3
	fi
	debug "Master host : $M_HOST"

	M_PORT=$( echo "$MASTER_CONN_INFOS"| grep 'port=' | sed 's/^.*port= *\([0-9a-zA-Z.-]\+\) *.*$/\1/' )
	if [ ! -n "$M_PORT" ]
	then
		debug "Master port not specified, use default : $PG_DEFAULT_PORT"
		M_PORT=$PG_DEFAULT_PORT
	else
		debug "Master port : $M_PORT"
	fi

	if [ -n "$PG_MASTER_USER" ]
	then
		debug "Master user provided by command-line, use it : $PG_MASTER_USER"
		M_USER="$PG_MASTER_USER"
	else
		M_USER=$( echo "$MASTER_CONN_INFOS"| grep 'user=' | sed 's/^.*user= *\([0-9a-zA-Z.-]\+\) *.*$/\1/' )
		if [ ! -n "$M_USER" ]
		then
			debug "Master user not specified, use default : $PG_USER"
			M_USER=$PG_USER
		else
			debug "Master user : $M_USER"
		fi
	fi

	M_APP_NAME=$( echo "$MASTER_CONN_INFOS"| grep 'application_name=' | sed "s/^.*application_name=[ \'\"]*\([^ \'\"]\+\)[ \'\"]*.*$/\1/" )
	if [ ! -n "$M_APP_NAME" ]
	then
		debug "Master application name not specified, use default : $PG_DEFAULT_APP_NAME"
		M_APP_NAME=$PG_DEFAULT_APP_NAME
	else
		debug "Master application name : $M_APP_NAME"
	fi

	# Get current state information from master
	M_CUR_REPL_STATE_INFO="$( psql_master_get "SELECT state, sync_state FROM pg_stat_replication WHERE application_name='$M_APP_NAME';" )"
	if [ ! -n "$M_CUR_REPL_STATE_INFO" ]
	then
		echo "UNKNOWN : Can't retreive current replication state information from master server"
		exit 3
	fi
	debug "Master current replication state:\n\tstate|sync_state\n\t$M_CUR_REPL_STATE_INFO"

	M_CUR_STATE=$( echo "$M_CUR_REPL_STATE_INFO"|cut -d'|' -f1 )
	debug "Master current state : $M_CUR_STATE"
	if [ "$M_CUR_STATE" != "streaming" ]
	then
		echo "CRITICAL : this host is not in streaming state according to master host (current state = '$M_CUR_STATE')"
		exit 2
	fi

	M_CUR_SYNC_STATE=$( echo "$M_CUR_REPL_STATE_INFO"|cut -d'|' -f2 )
	debug "Master current sync state : $M_CUR_SYNC_STATE"
	if [ "$M_CUR_SYNC_STATE" != "sync" ]
	then
		echo "CRITICAL : this host is not synchronized according to master host (current sync state = '$M_CUR_SYNC_STATE')"
		exit 2
	fi

	# Check current master XLOG file vs last replay XLOG file
	if [ "$CHECK_CUR_MASTER_XLOG" == "1" ]
	then
		# Get current xlog file from master
		M_CUR_XLOG="$( psql_master_get 'SELECT pg_current_xlog_location()' )"
		if [ ! -n "$M_CUR_XLOG" ]
		then
			echo "UNKNOWN : Can't retreive current xlog from master server"
			exit 3
		fi
		debug "Master current xlog : $M_CUR_XLOG"

		# Master current xlog is the last receive xlog ?
		if [ "$M_CUR_XLOG" != "$LAST_XLOG_RECEIVE" ]
		then
			echo "CRITICAL : Master current xlog is not the last receive xlog"
			exit 2
		fi
		debug "Master current xlog is the last receive xlog"
	fi

	# The last receive xlog is the last replay file ?
	if [ "$LAST_XLOG_RECEIVE" != "$LAST_XLOG_REPLAY" ]
	then
		debug "/!\ The last receive xlog is NOT the last replay file ('$M_CUR_XLOG' / '$LAST_XLOG_RECEIVE')"
		REPLAY_DELAY="$( psql_get 'SELECT EXTRACT(EPOCH FROM now() - pg_last_xact_replay_timestamp());' )"
		debug "Replay delay is $REPLAY_DELAY second(s)"
		if [ $( echo "$REPLAY_DELAY >= $REPLAY_CRITICAL_DELAY"|bc -l ) -gt 0 ]
		then
			echo "CRITICAL : last receive xlog file is not the last replay file ('$LAST_XLOG_RECEIVE' / '$LAST_XLOG_REPLAY') and replay delay is $REPLAY_DELAY second(s)"
			exit 2
		fi
		if [ $( echo "$REPLAY_DELAY >= $REPLAY_WARNING_DELAY"|bc -l ) -gt 0 ]
		then
			echo "WARNING : last receive xlog file is not the last replay file ('$LAST_XLOG_RECEIVE' / '$LAST_XLOG_REPLAY') and replay delay is $REPLAY_DELAY second(s)"
			exit 1
		fi
		debug "Replay delay is not worrying"
	fi
	debug "Last receive xlog file is the last replay file"

	echo "OK : Hot-standby server is uptodate"
	exit 0
else
	debug "File recovery.conf not found. Master mode."

	# Check recovery mode
	if [ $RECOVERY_MODE -eq 1 ]
	then
		echo "CRITICAL : In recovery mode while recovery.conf file not found !"
		exit 2
	fi
	debug "Postgres is not in recovery mode"

	# Check standby client
	STANDBY_CLIENTS=$( psql_get "SELECT application_name, client_addr, sent_location, write_location, state, sync_state, current_lag
	FROM (
	  SELECT application_name, client_addr, sent_location, write_location, state, sync_state, current_lag
	  FROM (
	    SELECT application_name, client_addr, $sent_lsn AS sent_location, $write_lsn AS write_location, state, sync_state,
	      $pg_wal_lsn_diff($pg_current_wal_lsn, $write_lsn) AS current_lag
	    FROM pg_stat_replication
	  ) AS s2
	) AS s1" )
	if [ ! -n "$STANDBY_CLIENTS" ]
	then
		echo "WARNING : no stand-by client connected"
		exit 1
	fi
	debug "Stand-by client(s):\n\t$( echo -e "$STANDBY_CLIENTS"|sed 's/\n/\n\t/' )"

	STANDBY_CLIENTS_TXT=""
	STANDBY_CLIENTS_COUNT=0
	for line in $STANDBY_CLIENTS
	do
		let STANDBY_CLIENTS_COUNT=STANDBY_CLIENTS_COUNT+1

		NAME=$( echo $line|cut -d '|' -f 1 )
		IP=$( echo $line|cut -d '|' -f 2 )
		SENT_LOCATION=$( echo $line|cut -d '|' -f 3 )
		WRITE_LOCATION=$( echo $line|cut -d '|' -f 4 )
		STATE=$( echo $line|cut -d '|' -f 5 )
		SYNC_STATE=$( echo $line|cut -d '|' -f 6 )
		LAG=$( echo $line|cut -d '|' -f 7 )
		STANDBY_CLIENTS_TXT="$STANDBY_CLIENTS_TXT\n$NAME ($IP): $STATE/$SYNC_STATE (Location: sent='$SENT_LOCATION' / write='$WRITE_LOCATION', Lag: ${LAG}b)"
	done

	echo -e "OK : $STANDBY_CLIENTS_COUNT stand-by client(s) connected\n$STANDBY_CLIENTS_TXT"
	exit 0
fi