Written by Claire Nakakeeto20Dec, 2022

Setting up zimbra redundancy – live sync

It is important to have a failover of every system as it improves the availability of the system and reduces data loss. In this articale, we describe how you can have near-live synchronization between two Zimbra servers so that one of them is live and the other is kept in a warm or very warm standby state. The sync can work in reverse when the mirror (redundant) server becomes the active server. This allows easy fall-back to the original server once the failover condition is resolved.

Zimbra employs several different databases to store messages, message indexes, meta-data, account information and configuration. Although it is possible to synchronise two Zimbra servers at the disk level using DRBD or VSphere, the amount of disk operations from all these databases that need to be replicated would probably take up a lot of bandwidth which may be debilitating and/or expensive to implement if the two servers are in remote locations.

Fortunately, Zimbra keeps a log of almost all it’s transactions in the **redolog**. The only thing not logged here are changes to the LDAP database. An incremental backup is made up of an LDAP dump and a collection of redologs. An incremental backup can be used to bring a backup server up to date if the last full backup of the backup server was more recent than the oldest log in the redolog.

Redolog

If the redolog can be piped to a mirror server in real time, then all the mirror server has to do is keep replaying the logs every so often and it will keep the same state as the live server. The only other thing to keep up to date is the LDAP database. Fortunately, the LDAP database doesn’t change that often so it is quite easy to keep it synced on a directory level.

The easiest way to transfer the redologs is to use rsync. The only problem with that is that rsync does not run continuously. It also won’t handle the archiving of redo.log very efficiently. When redo.log is renamed and moved to the archive rsync will delete it at the remote location then transfer it all over again to its new location in the archive. If we can catch this move taking place then we can move and rename the file on the mirror server before running rsync. Then rsync has very little to do, in theory, nothing except delete any files that have been purged. Another issue with rsync is that the file may be in the process of being written to when it is copied. This results in an incomplete file at the mirror. However, redologs are only ever appended to and so only the last record will be corrupted. Zimbra is designed to be tolerant of redolog corruption otherwise it would be of limited use as a disaster recovery tool.

To keep the redolog live, the `tail -f` command is used over ssh to pipe the file to the mirror. By calling `tail -f -c +0` it tails right back to the zeroth byte of the file, effectively a *copy-then-stream* command.

Redolog purging

If a Network edition is detected and incremental backups are enabled, then the redologs are replayed before any rsync is performed as well as after. This ensures that everything is replayed before the files all disappear to the backup directory.

For the Open Source Edition, or Network Edition with no incremental backups scheduled, the redologs are purged if they are more than a day old and have been replayed. If the mirror server is down then the redologs will just accumulate on the live server to be replayed when the sync process is restarted before being purged.

LDAP

LDAP stores its data in `/opt/zimbra/data/ldap`. This can be copied using rsync to the mirror as long as no changes take place during the copy. On versions of Zimbra older than 8.0, the directory is monitored for this during the rsync operation and repeated if there was any change during that time. For Zimbra 8+, the directory is monitored as before but changes are transferred using an LDIF export and import. This is necessitated by the long time it takes to rsync the sparse files that LDAP now uses to store data.

Known issues

* If the connection breaks at the very moment that the live stream of the redo.log starts, before the tail command reaches the point where it is tailing instead of cataloging the file, then some of the redo.log will not make it to the mirror resulting in some loss of transactions. Fortunately, this is only ever likely just after the log has rolled over so the worst-case losses should be minimal.

* LDAP is only checked every ten minutes so some losses are possible if the connection breaks in that time. However, LDAP isn’t expected to change very often unless something major like a batch account migration is taking place.

* If any redologs go missing, or can’t be replayed successfully for any reason, then there will be gaps in synced email events. Mail may go missing on the mirror server. Check the *live_sync* logs for any log sequence numbers that appear not to have been transferred and processed. In the case of suspected data loss, stop all services and repeat the initial rsync process.

Preparation

Exactly the same version of Zimbra must be installed on both the live and mirror server. To start with, we work on the live server. There is no need to stop Zimbra for most of the install. Only a short amount of down-time will need to be scheduled later to perform a final rsync operation between the two servers.

The mirror server should ideally have the same operating system as the live server and must have exactly the same version of Zimbra installed. The hostname must also be exactly the same.

Live Server

  • Install inotify-tools

inotify tools are required. For RHEL/CentOS, inotify-tools is provided by the epel-release repository.

As user root: (Centos)

yum install epel-release -y
yum install inotify-tools -y

while for Ubuntu, as user root:

apt install inotify-tools -y
  • Create log rotation

The script will create a log file which can be handled by logrotate. As user root:

echo "/opt/zimbra/live_sync/log/live_sync.log {
    daily
    missingok
    copytruncate
    rotate 7
    notifempty
    compress
}">/etc/logrotate.d/zimbra_live_sync
  • Create application directory

The script will live under the /opt/zimbra directory. As user root:

mkdir /opt/zimbra/live_sync
chown zimbra.zimbra /opt/zimbra/live_sync
  • SSH Key

Create the SSH key and just press return every time you are prompted for a passphrase. As user zimbra:

sudo su - zimbra
cd /opt/zimbra/.ssh
ssh-keygen -b 4096 -f live_sync
echo "command=\"/opt/zimbra/live_sync/sync_commands\" $( cat live_sync.pub )">>authorized_keys
exit
  • Main Script

The following script should be saved as live_syncd in the /opt/zimbra/live_sync directory. This should be owned by user zimbra and made executable.

sudo vim /opt/zimbra/live_sync/live_syncd
#!/bin/bash
#	This program is free software: you can redistribute it and/or modify
#	it under the terms of the GNU General Public License as published by
#	the Free Software Foundation, either version 3 of the License, or
#	(at your option) any later version.
 
#	This program is distributed in the hope that it will be useful,
#	but WITHOUT ANY WARRANTY; without even the implied warranty of
#	MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
#	GNU General Public License for more details.
 
#	You should have received a copy of the GNU General Public License
#	along with this program.  If not, see <http://www.gnu.org/licenses/>
 
##########################################################################
# Title  	:  live_syncd
# Author 	:  Simon Blandford <simon -at- onepointltd -dt- com>
# Date   	:  2013-03-12
# Requires   :  zimbra sync_commands inotify-tools
# Category   :  Administration
# Version	:  2.1.5
# Copyright  :  Simon Blandford, Onepoint Consulting Limited
# License	:  GPLv3 (see above)
##########################################################################
# Description
# Keep two Zimbra servers synchronised in near-realtime
##########################################################################
 
 
#******************************************************************************
#********************** Constants *********************************************
#******************************************************************************
 
LOG_LEVEL=5
REDO_LOG_HISTORY_DAYS=10
ERROR_CLEAR_MINUTES=10
LDAP_CHECK_MINUTES_INTERVAL=10
 
ZIMBRA_DIR="/opt/zimbra"
BASE_DIR="$ZIMBRA_DIR""/live_sync"
LOCKING_DIR="$BASE_DIR""/lock"
PID_DIR="$BASE_DIR""/pid"
LOG_DIR="$BASE_DIR""/log"
LOG_FILE="$LOG_DIR""/live_sync.log"
LDAP_TEMP_DIR="$BASE_DIR""/ldap"
LDAP_TEMP_LDIF="$BASE_DIR""/ldif.bak"
STATUS_DIR="$BASE_DIR""/status"
SSH_IDENTITY_FILE="$ZIMBRA_DIR""/.ssh/live_sync"
REDOLOG_DIR="$ZIMBRA_DIR""/redolog"
REDO_LOG_FILE="$REDOLOG_DIR""/redo.log"
ARCHIVE_DIR="$REDOLOG_DIR""/archive"
LIVE_SYNC_ARCHIVE_DIR="$REDOLOG_DIR""/live_sync_archives"
LDAP_DATA_DIR="$ZIMBRA_DIR""/data/ldap/"
BACKUP_DIR="$ZIMBRA_DIR""/backup"
SYNC_COMMANDS_SCRIPT="$BASE_DIR""/sync_commands"
SSH="ssh -i ""$SSH_IDENTITY_FILE"" -o StrictHostKeyChecking=no -o CheckHostIP=no"\
" -o PreferredAuthentications=hostbased,publickey"
LOCK_STATE_DIR="$LOCKING_DIR""/live_sync.lock"
STOP_FILE="$STATUS_DIR""/live_sync.stop"
LAST_GOOD_REDO_REPLAY="$STATUS_DIR""/last_good_redo_replay"
LAST_GOOD_REDO_SYNC="$STATUS_DIR""/last_good_redo_sync"
LAST_GOOD_REDO_STREAM="$STATUS_DIR""/last_good_redo_stream"
LAST_GOOD_LDAP_SYNC="$STATUS_DIR""/last_good_ldap_sync"
LAST_GOOD_LDAP_START="$STATUS_DIR""/last_good_ldap_start"
WATCHES_FILE="$STATUS_DIR""/watches"
PID_FILE_LDAP="$PID_DIR""/ldap_live_sync.pid"
PID_FILE_REDO="$PID_DIR""/redo_log_live_sync.pid"
CONF_FILE="$BASE_DIR""/live_sync.conf"
 
#******************************************************************************
#********************** Functions *********************************************
#******************************************************************************
 
#Format for log output with errors and warnings going to >&2
logit () {
  logit_1 () {
	echo -n "$( date ) :"
	case $ in
  	1)
    	echo -n "Error :"
    	;;
  	2)
    	echo -n "Warning :"
    	;;
  	3)
    	echo -n "Info :"
    	;;
  	4)
    	echo -n "Debug :"
    	;;
	esac
	echo $@
  }
  local msg_level output_chan
  if [ $1 -le $LOG_LEVEL ]; then
	msg_level=$1
	shift
	if [ $msg_level -le 2 ]; then
  	logit_1 $@ >&2
	else
  	logit_1 $@
	fi
  fi
}
 
#Detect HSM
detect_hsm () {
  local retval
  #LDAP must be running
  ldap status &>/dev/null || ldap start &>/dev/null
  #MySQL must be running
  mysql.server status &>/dev/null || mysql.server start &>/dev/null
  #Preserve mailbox running state
  zmmailboxdctl status &>/dev/null
  prev_zmmailbox_status=$?
  zmmailboxdctl start &>/dev/null
  zmvolume -l | grep "type: secondaryMessage" >/dev/null
  retval=$?
   if [ $prev_zmmailbox_status -ne 0 ]; then
	zmmailboxdctl stop &>/dev/null
  fi
  return $retval
}
 
#Ensure ldap, convertd and mysql servers are running and then replay redo logs
replay_redo_logs () {
  local server_failed
 
  ldap status &>/dev/null || ldap start &>/dev/null
  mysql.server status &>/dev/null || mysql.server start &>/dev/null
  server_failed=0
  if ! ldap status &>/dev/null; then
	logit 1 "Start of local ldap server failed"
	ldap status >&2
	#Return error to trigger a "break" in while loop
	server_failed=1
  fi
  if ! mysql.server status &>/dev/null; then
	logit 1 "Start of local mysql server failed"
	mysql.server status >&2
	#Return error to trigger a "break" in while loop
	server_failed=1
  fi
  if [ "x""$convertd_enabled" == "xtrue" ]; then
	#Make sure indexing works while replaying redo log
	zmconvertctl status &>/dev/null || zmconvertctl start &>/dev/null
	if ! zmconvertctl status &>/dev/null; then
  	logit 2 "Start of local convertd servers failed"
  	zmconvertctl status >&2
	fi
  fi
  [ $server_failed -eq 1 ] && return 1
  logit 3 "Replaying redologs..."
  if ! zmplayredo >/dev/null; then
	logit 2 "Replay of redolog failed"
	#No error returned here since "break" is not necessary
  else
	#If no errors then archive redo log files
	if ! mkdir -p "$LIVE_SYNC_ARCHIVE_DIR"; then
  	logit 1 "Unable to create directory $LIVE_SYNC_ARCHIVE_DIR"
  	exit 1
	fi
	mv -f "$ARCHIVE_DIR""/"* "$LIVE_SYNC_ARCHIVE_DIR""/" 2>/dev/null
	touch "$LAST_GOOD_REDO_REPLAY"
  fi
  logit 3 "Replaying redologs done"
  return 0
}
 
#The redo log sync daemon
redo_log_live_sync () {
  local stream_pid archived_file i archived_redo_log_file prev_zmmailbox_status secondary_storage
 
  logit 3 "Starting redo log live sync process"
 
  #Wait for lock directory to be successfully created
  while ! mkdir "$LOCK_STATE_DIR" &>/dev/null; do
	sleep 2
  done
  logit 3 "Detecting if HSM used"
  if detect_hsm; then
	logit 3 "HSM Detected"
	secondary_storage="yes"
  else
	logit 3 "No HSM Detected"
  fi
  rmdir "$LOCK_STATE_DIR"
 
  while [ ! -f "$STOP_FILE" ]; do
	while [ ! -f "$STOP_FILE" ]; do
  	#Wait for lock directory to be successfully created
  	while ! mkdir "$LOCK_STATE_DIR" &>/dev/null; do
    	sleep 2
  	done
  	[ -f "$STOP_FILE" ] && break
  	logit 3 "Syncing redologs..."
  	#If incremental backups are enabled then gather redo logs from backups and copy
  	#to local archive directory
  	redo_sync_fail="false"
  	for archived_redo_log_file in $( echo "gather""$REDO_LOG_HISTORY_DAYS" | \
      	$SSH "$remote_address" "$SYNC_COMMANDS_SCRIPT" ); do
    	if [ -f "$LIVE_SYNC_ARCHIVE_DIR""/""$( basename "$archived_redo_log_file" )" ]; then
      	logit 4 "Already processed so skipping: $archived_redo_log_file"
    	else
      	logit 4 "Syncing incremental backup file: $archived_redo_log_file"
      	if ! rsync -z -e "$SSH" --size-only "$remote_address"":""$archived_redo_log_file" \
          	"$ARCHIVE_DIR""/".; then
        	logit 2 "Rsync of a redolog, $archived_redo_log_file, failed"
        	redo_sync_fail="true"
      	fi
    	fi
  	done
 
  	#Suspend if HSM is running
  	if which zmhsm >/dev/null && zmhsm -u | grep "Currently running" >/dev/null; then
    	logit 3 "Replaying redologs is suspended while HSM process is active"
  	else
 	 
    	#Mailbox process must not be running now. Preserve state and stop.
    	zmmailboxdctl status &>/dev/null
    	prev_zmmailbox_status=$?
    	if [ $prev_zmmailbox_status -eq 0 ]; then
      	zmmailboxdctl stop &>/dev/null
    	fi
    	sleep 2
    	if zmmailboxdctl status &>/dev/null; then
      	logit 1 "Unable to stop local Zimbra mailbox service"
      	return 1
    	fi
 	 
    	logit 4 "Syncing $REDO_LOG_FILE"
    	if ! rsync -e "$SSH" -z \
      	"$remote_address"":$REDO_LOG_FILE" "$REDO_LOG_FILE"; then
      	logit 2 "Rsync of $REDO_LOG_FILE failed"
      	redo_sync_fail="true"
    	fi
    	logit 4 "Syncing $REDO_LOG_FILE done"
    	if [ "x""$redo_sync_fail" == "xfalse" ]; then
      	touch "$LAST_GOOD_REDO_SYNC"
    	else
      	break
    	fi
    	logit 4 "Syncing redologs done"
    	logit 4 "Purging redolog directory and archives"
    	#Purge local redolog directory
    	find $REDOLOG_DIR -mtime +$REDO_LOG_HISTORY_DAYS -type f -exec rm {} \;
    	#Purge any interrupted rsync files
    	find $REDOLOG_DIR -name '.redo*' -type f -exec rm {} \;
    	logit 4 "Purge redolog directory and archives done"
    	replay_redo_logs || break
 	 
    	#Restore mailboxd to previous running state or start if HSM is being used
    	if [ $prev_zmmailbox_status -eq 0 ] || \
        	[ "x""$secondary_storage" == "xyes" ] >/dev/null; then
      	logit 4 "Re-starting Zimbra mailbox service"
      	zmmailboxdctl start &>/dev/null
      	if ! zmmailboxdctl status &>/dev/null; then
        	logit 2 "Unable to re-start local Zimbra mailbox service"
      	fi
    	fi
  	fi
 	 
  	#If there are no incremental backups then remote archive directory will need purging
  	if [ "x""$incremental_backups" != "xtrue" ]; then
    	logit 4 "Purging remote redolog directory"
    	echo "purge""$REDO_LOG_HISTORY_DAYS" | \
      	$SSH "$remote_address" "$SYNC_COMMANDS_SCRIPT"
    	logit 4 "Purging remote redolog directory done"
  	fi
  	#Establish copy-and-live-stream of current redo.log file
  	logit 4 "Live streaming redolog"
  	echo stream | \
    	$SSH "$remote_address" \
    	"$SYNC_COMMANDS_SCRIPT" >"$REDO_LOG_FILE" &
  	stream_pid=$!
  	disown $stream_pid
  	#Delay as PID was sometimes not being found if checked immediately
  	sleep 5
  	#If successfully established stream then sit and wait for move to archive
  	if ps $stream_pid | grep "$SYNC_COMMANDS_SCRIPT" &>/dev/null; then
    	logit 4 "Live streaming redolog established"
    	touch "$LAST_GOOD_REDO_STREAM"
    	#Remove lock file, this is resting point
    	rmdir "$LOCK_STATE_DIR" &>/dev/null
    	#Wait for name to be passed of new archive file after redo.log is moved on remote server
    	#This is normal resting point of this process
    	archived_file=$( echo wait_redo | \
      	$SSH "$remote_address" "$SYNC_COMMANDS_SCRIPT" | \
      	tail -n 1 | grep -Eo "redo-.*log" )
    	#Kill stream
    	kill -KILL $( ps aux | grep "$SYNC_COMMANDS_SCRIPT" | \
      	grep -v grep | awk '{print $2}' ) &>/dev/null
    	#Mirror move operation on local server
    	if echo "$archived_file" | grep -E "redo-.*log" &>/dev/null; then
      	logit 4 "Moving redo.log to $archived_file"
      	mv -f "$REDO_LOG_FILE" "$ARCHIVE_DIR""/""$archived_file" 2>/dev/null
    	else
      	logit 2 "Archive file name not found"
    	fi
    	[ -f "$STOP_FILE" ] && break
  	else
    	logit 2 "Failed to start redolog streaming, PID=$stream_pid"
    	break
  	fi
	done
	rmdir "$LOCK_STATE_DIR" &>/dev/null
	#Wait $ERROR_CLEAR_MINUTES minutes for error to error to clear
	i=0
	while [ $(( i++ )) -lt 60 ] && [ ! -f "$STOP_FILE" ]; do
  	sleep $ERROR_CLEAR_MINUTES
	done
  done
  logit 3 "Ending redo log live sync process"
}
 
#The ldap sync daemon
ldap_live_sync () {
  local ldap_wait_pid i last_ldap_success_state
 
  last_ldap_success_state="false"
 
  logit 3  "Starting ldap live sync process"
  while [ ! -f "$STOP_FILE" ]; do
	while [ ! -f "$STOP_FILE" ]; do
  	#Wait for lock directory to be successfully created
  	while ! mkdir "$LOCK_STATE_DIR" &>/dev/null; do
    	sleep 3
  	done
  	if [ $zimbra_version -lt 8 ]; then
    	logit 3 "Syncing ldap using rsync"
    	#Use rsync for Zimbra older than verion 8
    	while [ 1 ]; do
      	#Check for changes during ldap sync operation
      	echo wait_ldap | \
        	$SSH "$remote_address" "$SYNC_COMMANDS_SCRIPT" &>"$WATCHES_FILE" &
      	ldap_wait_pid=$!
      	disown $ldap_wait_pid
      	if ! ps "$ldap_wait_pid" &>/dev/null; then
        	logit 2 "Unable to establish watch on remote LDAP directory, no ldap sync performed"
        	break
      	fi
      	#Wait for watches to be established
      	while ! grep "established" "$WATCHES_FILE" &>/dev/null && \
          	ps "$ldap_wait_pid" &>/dev/null; do
        	sleep 1
      	done
      	#Echo out status
      	cat "$WATCHES_FILE"
      	rm -f "$WATCHES_FILE"
     	 
     	 
      	#Rsync remote server to temporary local ldap directory
      	if ! rsync -e "$SSH" -aHz --sparse --force --delete \
        	"$remote_address"":$LDAP_DATA_DIR""/" "$LDAP_TEMP_DIR""/"; then
        	logit 2 "Rsync of ldap failed"
        	break
      	else
        	touch "$LAST_GOOD_LDAP_SYNC"
      	fi
      	ps $ldap_wait_pid &>/dev/null && break
      	logit 3 "Ldap changed during rsync. Re-syncing."
      	sleep 10
    	done
    	kill -KILL $ldap_wait_pid &>/dev/null
  	else
    	#Use ldif export for Zimbra 8 and over
    	logit 3 "Syncing ldap using ldif"
    	if ! echo dump_ldap | \
      	$SSH "$remote_address" "$SYNC_COMMANDS_SCRIPT" >"$LDAP_TEMP_LDIF"; then
      	logit 2 "Unable to fetch remote LDIF, no LDAP sync performed"
      	break
    	else
      	touch "$LAST_GOOD_LDAP_SYNC"
    	fi
  	fi
  	if which zmhsm >/dev/null && zmhsm -u | grep "Currently running" >/dev/null; then
    	logit 3 "LDAP update is suspended while HSM process is active"
  	else
    	#Stop ldap
    	ldap status &>/dev/null && ldap stop &>/dev/null
    	if ldap status &>/dev/null; then
      	logit 1 "Unable to stop local ldap server"
      	break
    	fi
    	if [ $zimbra_version -lt 8 ]; then
      	#Use rsync for Zimbra older than version 8
      	#rsync temporary local ldap directory to real local ldap directory
      	rsync -aH --sparse "$LDAP_TEMP_DIR""/" "$LDAP_DATA_DIR""/"
    	else
      	#Use LDIF import for Zimbra 8 and over
      	rm -rf "$LDAP_DATA_DIR""/mdb" && \
      	mkdir -p "$LDAP_DATA_DIR""/mdb/db" && \
      	mkdir -p "$LDAP_DATA_DIR""/mdb/log" && \
      	/opt/zimbra/libexec/zmslapadd "$LDAP_TEMP_LDIF"
      	if [ $? != 0 ]; then
        	logit 2 "Unable to import LDIF into local LDAP"
        	break
      	fi
    	fi
    	#Restart ldap
    	ldap status &>/dev/null || ldap start &>/dev/null
    	if ! ldap status &>/dev/null; then
      	logit 1 "Unable to restart local ldap server"
      	last_ldap_success_state="false"
    	else
      	last_ldap_success_state="true"
    	fi
    	logit 4 "Syncing LDAP done"
  	fi
  	rmdir "$LOCK_STATE_DIR" &>/dev/null
  	[ -f "$STOP_FILE" ] && break
  	#Wait for change in remote ldap over $LDAP_CHECK_MINUTES_INTERVAL intervals
  	echo wait_ldap | \
    	$SSH "$remote_address" "$SYNC_COMMANDS_SCRIPT" &
  	ldap_wait_pid=$!
  	disown $ldap_wait_pid
  	while [ ! -f "$STOP_FILE" ]; do
    	logit 4 "Start new LDAP monitor period"
    	#Repeat last ldap success so that no ldap change is not
    	#interpreted by Nagios as no ldap success.
    	if [ "x""$last_ldap_success_state" == "xtrue" ]; then
      	touch "$LAST_GOOD_LDAP_START"
    	fi
    	#Restart wait for ldap change if required
    	if ! ps $ldap_wait_pid &>/dev/null; then
      	echo wait_ldap | \
        	$SSH "$remote_address" "$SYNC_COMMANDS_SCRIPT" &
      	ldap_wait_pid=$!
      	disown $ldap_wait_pid
    	fi
    	#Wait $LDAP_CHECK_MINUTES_INTERVAL minutes
    	i=0
    	while [ $(( i++ )) -lt 60 ] && [ ! -f "$STOP_FILE" ]; do
      	sleep $LDAP_CHECK_MINUTES_INTERVAL
    	done
    	#If wait process is not still running then there was a change
    	ps $ldap_wait_pid &>/dev/null || break
  	done
	done
	rmdir "$LOCK_STATE_DIR" &>/dev/null
	#Wait $ERROR_CLEAR_MINUTES minutes for error to error to clear
	i=0
	while [ $(( i++ )) -lt 60 ] && [ ! -f "$STOP_FILE" ]; do
  	sleep $ERROR_CLEAR_MINUTES
	done
  done
  logit 3 "Ending ldap live sync process"
}
 
get_zimbra_config_globals () {
  #Query whether incremental backups are enabled
  incremental_backups=$( echo "query_incremental" | \
	$SSH "$remote_address" "$SYNC_COMMANDS_SCRIPT" )
    
  #Query whether convertd is installed and enabled
  ldap status &>/dev/null || ldap start &>/dev/null
  if ! ldap status &>/dev/null; then
	logit 1 "Unable to start local ldap server"
	exit 1
  fi
  if [ $( zmprov -l  gs `zmhostname` | \
      	grep -E "(zimbraServiceInstalled|zimbraServiceEnabled):[[:space:]]*convertd" | \
      	wc -l  ) -eq 2 ]; then
	convertd_enabled="true"
  else
	convertd_enabled="false"
  fi
}
 
kill_everything () {
  touch "$STOP_FILE"
  kill -KILL $( head -n 1 "$PID_FILE_LDAP" 2>/dev/null ) &>/dev/null
  kill -KILL $( head -n 1 "$PID_FILE_REDO" 2>/dev/null ) &>/dev/null
  kill -KILL $( ps aux | grep "live_syncd start" | grep -v grep | awk '{print $2}' ) &>/dev/null
  kill -KILL $( ps aux | grep "redo_log_live_sync" | grep -v grep | awk '{print $2}' ) &>/dev/null
  kill -KILL $( ps aux | grep "ldap_live_sync" | grep -v grep | awk '{print $2}' ) &>/dev/null
  kill -KILL $( ps aux | \
	grep "$SYNC_COMMANDS_SCRIPT" | grep -v grep | awk '{print $2}' ) &>/dev/null
  kill -KILL $( ps aux | grep "rsync" | grep -E "$REDOLOG_DIR""|""$LDAP_DATA_DIR""|""$BACKUP_DIR" | \
	awk '{print $2}' ) &>/dev/null
  #Kill redolog playback if running
  kill -KILL $( ps aux | grep -E "zimbra.*java.*PlaybackUtil" | grep -v grep | \
	awk '{print $2}' ) &>/dev/null
  rm -f "$STOP_FILE"
  rm -f "$PID_FILE_LDAP"
  rm -f "$PID_FILE_REDO"
  rmdir "$LOCK_STATE_DIR" &>/dev/null
}
 
quitting () {
  echo "Quitting"
  #Kill any hanging processes
  kill_everything
  trap - INT TERM SIGINT SIGTERM
  echo 'kill -KILL $( ps aux | grep live_syncd | grep -v grep | awk '"'"'{print $2}'"'"' ) &>/dev/null' | \
	at now && sleep 1 && rmdir "$LOCK_STATE_DIR" &>/dev/null
  exit
}
 
 
#******************************************************************************
#********************** Main Program ******************************************
#******************************************************************************
 
if [ "$( whoami )" != "zimbra" ]; then
  echo "Must run as zimbra user" >&2
  exit 1
fi
 
mkdir -p "$LOCKING_DIR"
mkdir -p "$PID_DIR"
mkdir -p "$LOG_DIR"
mkdir -p "$LDAP_TEMP_DIR"
mkdir -p "$STATUS_DIR"
chmod 755 "$STATUS_DIR"
 
if [ ! -f "$CONF_FILE" ]; then
  echo "Configuration file, $CONF_FILE, not found" >&2
  exit 1
fi
 
source "$CONF_FILE"
 
#Find all local addresses
server_addresses=$( /sbin/ifconfig |grep inet | grep -E '(([0-9]+\.){3}[0-9]+|[0-9a-f]+(:[0-9a-f]*){5})' |awk '//{print $2}' )
#server_addresses=$( /usr/sbin/ip a | grep -Po 'inet \K[\d.]+')
#Check configured server addresses are valid
if ! echo "$server1" | \
	grep -Ei "([0-9]+\.){3}[0-9]+|[0-9a-f]+(:[0-9a-f]*){5}" &>/dev/null; then
  echo "No valid IP address found for server1 in configuration file" >&2
  exit 1
fi
if ! echo "$server2" | \
	grep -Ei "([0-9]+\.){3}[0-9]+|[0-9a-f]+(:[0-9a-f]*){5}" &>/dev/null; then
  echo "No valid IP address found for server2 in configuration file" >&2
  exit 1
fi
 
#Deduce local address and assume other address is remote machine
if echo "$server_addresses" | grep "$server1" &>/dev/null; then
  local_address="$server1"
  remote_address="$server2"
else
  if echo "$server_addresses" | grep "$server2" &>/dev/null; then
	local_address="$server2"
	remote_address="$server1"
  else
	echo "Unable to identify local server address and assume remote address" >&2
	exit 1
  fi
fi
 
#Check remote server is OK
remote_server_status=$( echo "test" | \
  $SSH "$remote_address" "$SYNC_COMMANDS_SCRIPT" )
 
if [ "x""$remote_server_status" == "xbusy" ]; then
  echo "Remote server appears to have live_syncd process running" >&2
  echo "This can not run on both servers" >&2
  exit 1
fi
 
if [ "x""$remote_server_status" != "xOK" ]; then
  echo "Unable to run commands on remote server" >&2
  exit 1
fi
 
#Get major Zimbra version
zimbra_version=$( zmcontrol -v | grep -Eo "[0-9][^\.]*" | head -n 1 )
if [ ${#zimbra_version} -lt 1 ]; then
 zimbra_version=0
fi
 
case $1 in
  start)
	#Check for processes from this script and also redolog replay. Don't count PID files older than uptime.
	if [ -f "$PID_FILE_REDO" ] && \
    	[ $(( $( date +%s ) - $( stat -c '%Y' "$PID_FILE_REDO" ) )) -lt $( cat /proc/uptime | grep -Eo "[0-9]+" | head -n 1 ) ]; then
  	pid_found="yes"
	fi
	if [ -f "$PID_FILE_LDAP" ] && \
    	[ $(( $( date +%s ) - $( stat -c '%Y' "$PID_FILE_LDAP" ) )) -lt $( cat /proc/uptime | grep -Eo "[0-9]+" | head -n 1 ) ]; then
  	pid_found="yes"
	fi
	if [ $pid_found ] || \
    	ps aux | grep -E "zimbra.*java.*PlaybackUtil" | grep -v grep &>/dev/null; then
  	echo "Proccess already running"
	else
  	echo -n "Starting processes..."
  	get_zimbra_config_globals
  	echo "***************************************" >>"$LOG_FILE"
  	logit 3 "Starting $( basename "$0" )" >>"$LOG_FILE"
  	logit 3 "Incremental backups enabled : $incremental_backups" >>"$LOG_FILE"
  	logit 3 "Convertd enabled : $convertd_enabled" >>"$LOG_FILE"
 
  	ldap_live_sync >>"$LOG_FILE" 2>&1 &
  	echo $! >"$PID_FILE_LDAP"
  	redo_log_live_sync >>"$LOG_FILE" 2>&1 &
  	echo $! >"$PID_FILE_REDO"
  	echo "done"
	fi
	;;
  stop)
	touch "$STOP_FILE"
	[ -d "$LOCK_STATE_DIR" ] && echo "Waiting for sync operations to complete..."
	while [ -d "$LOCK_STATE_DIR" ]; do
  	sleep 5
	done
	rm -f "$STOP_FILE"
	replay_redo_logs
	kill_everything
	echo "done"
	;;
  status)
	if ps aux | grep -E "zimbra.*java.*PlaybackUtil" | grep -v grep &>/dev/null; then
  	echo "redolog is being replayed"
  	replay_stat=0
	else
  	replay_stat=3
	fi
	if [ -f  $PID_FILE_REDO ] && ps $( head -n 1 $PID_FILE_REDO 2>/dev/null ) &>/dev/null; then
  	echo "redo log sync process OK"
  	redo_stat=0
	else
  	echo "redolog sync process stopped"
  	redo_stat=3
	fi
	if [ -f  $PID_FILE_LDAP ] && ps $( head -n 1 $PID_FILE_LDAP 2>/dev/null ) &>/dev/null; then
  	echo "ldap sync process OK"
  	ldap_stat=0
	else
  	echo "ldap sync process stopped"
  	ldap_stat=3
	fi
	[ $ldap_stat == 3 ] && [ $redo_stat == 3 ] && [ $replay_stat == 3 ] && exit 3
	[ $ldap_stat == 0 ] && [ $redo_stat == 0 ] && exit 0
	exit 1
	;;
  kill)
	kill_everything
	;;
  *)
	trap quitting INT TERM SIGINT SIGTERM
	if ps aux | grep "redo_log_live_sync" | grep -v grep  &>/dev/null || \
    	ps aux | grep "ldap_live_sync" | grep -v grep  &>/dev/null || \
    	ps aux | grep -E "zimbra.*java.*PlaybackUtil" | grep -v grep &>/dev/null; then
  	echo "Proccess already running"
	else
  	echo "Starting processes in realtime"
  	get_zimbra_config_globals
  	logit 3 "Incremental backups enabled : $incremental_backups"
  	logit 3 "Convertd enabled : $convertd_enabled"
  	ldap_live_sync &
  	echo $! >"$PID_FILE_LDAP"
  	redo_log_live_sync &
  	echo $! >"$PID_FILE_REDO"
  	while [ 1 ]; do sleep 10; done
	fi
	;;
esac
sudo chown zimbra:zimbra /opt/zimbra/live_sync
sudo chmod +x  /opt/zimbra/live_sync
  • Remote command script

The following script should be saved as sync_commands in the /opt/zimbra/live_sync directory. This should be owned by user zimbra and made executable.

sudo vim /opt/zimbra/live_sync/sync_commands
#!/bin/bash
#    This program is free software: you can redistribute it and/or modify
#    it under the terms of the GNU General Public License as published by
#    the Free Software Foundation, either version 3 of the License, or
#    (at your option) any later version.
 
#    This program is distributed in the hope that it will be useful,
#    but WITHOUT ANY WARRANTY; without even the implied warranty of
#    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
#    GNU General Public License for more details.
 
#    You should have received a copy of the GNU General Public License
#    along with this program.  If not, see <http://www.gnu.org/licenses/>
 
##########################################################################
# Title      :  sync_commands
# Author     :  Simon Blandford <simon -at- onepointltd -dt- com>
# Date       :  2013-03-12
# Requires   :  zimbra sync_commands inotify-tools
# Category   :  Administration
# Version    :  2.1.3
# Copyright  :  Simon Blandford, Onepoint Consulting Limited
# License    :  GPLv3 (see above)
##########################################################################
# Description
# Keep two Zimbra servers synchronised in near-realtime, local agent
##########################################################################
 
 
#******************************************************************************
#********************** Main Program ******************************************
#******************************************************************************
 
 
if [ "$( whoami )" != "zimbra" ]; then
  echo "Must run as zimbra user" >&2
  exit 1
fi
 
#Check for rsync of redolog or ldap
if echo "$SSH_ORIGINAL_COMMAND" | \
  grep "rsync" | \
    grep -E "/opt/zimbra/redolog/|/opt/zimbra/data/ldap/|/opt/zimbra/backup/sessions/incr" &>/dev/null; then
  case "$SSH_ORIGINAL_COMMAND" in
  *\&*)
    echo "Rejected"
    ;;
  *\(*)
    echo "Rejected"
    ;;
  *\{*)
    echo "Rejected"
    ;;
  *\;*)
    echo "Rejected"
    ;;
  *\<*)
    echo "Rejected"
    ;;
  *\`*)
    echo "Rejected"
    ;;
  rsync\ --server*)
    $SSH_ORIGINAL_COMMAND
    ;;
  *)
    echo "Rejected"
    ;;
  esac
else
  #Not rsync
  case "$#" in
  0) read command
    ;;
  *) command=$1
    ;;
  esac
 
  check_inotify () {
    if ! which inotifywait &>/dev/null; then
      echo "inotifywait not found" >&2
      echo "Please install inotify-tools" >&2
      exit 1
    fi
  }
 
  #Extract numeric parameter from command name
  param=$( echo "$command" | grep -Eo "[0-9]+" )
  command=$( echo "$command" | grep -Eio "[a-z_]+" )
  
  case $command in
    test)
      if ps aux | grep "live_syncd" | grep -v grep &>/dev/null; then
        echo "busy"
      else
        echo "OK"
      fi
      ;;
    wait_redo)
      #Wait for redo log roll-over
      check_inotify
      kill -KILL $( ps aux | grep "inotifywait -r /opt/zimbra/redolog" | \
                    grep -v grep | awk '{print $2}' ) &>/dev/null
      inotifywait -r /opt/zimbra/redolog -e moved_to
      ;;
    wait_ldap)
      #Wait for ldap changes. Ignore log changes.
      check_inotify
      kill -KILL $( ps aux | grep "inotifywait -r /opt/zimbra/data/ldap" | \
                    grep -v grep | awk '{print $2}' ) &>/dev/null
      inotifywait -r /opt/zimbra/data/ldap -e modify \
        -e attrib -e close_write -e moved_to -e moved_from \
        --exclude "logs\/log\.|accesslog" \
        -e move -e delete -e delete_self
      ;;
    dump_ldap)
      #Extract the LDIF database and stream it
      /opt/zimbra/libexec/zmslapcat  "/tmp/zimbraldif"
      cat "/tmp/zimbraldif/ldap.bak"
      rm -rf "/tmp/zimbraldif"
      ;;
    stream)
      #Live-stream redolog
      #Kill any hanging previous tail commands
      kill -KILL $( ps aux | grep "tail -c +0 -f /opt/zimbra/redolog/redo.log" | \
                    grep -v grep | awk '{print $2}' ) &>/dev/null
      tail -c +0 -f /opt/zimbra/redolog/redo.log
      ;;
    gather)
      #Gather list of recent redologs from incremental backups and archive
      find '/opt/zimbra/backup/sessions/incr-'*'/redologs' \
           '/opt/zimbra/redolog/archive' \
            -name 'redo*.log' -type f -mtime -$param -print 2>/dev/null | \
            sort
      ;;
    purge)
      #Remove old archives
      find /opt/zimbra/redolog/archive -type f -mtime +$param -exec rm {} \;
      ;;
    query_incremental)
      #Query whether incremental backups are scheduled
      if which zmschedulebackup &>/dev/null && \
          zmschedulebackup -q | \
          grep -Eo "i([[:space:]]+[0-9\*\-]+){5}" &>/dev/null; then
        echo "true"
      else
        echo "false"
      fi
      ;;
    *)
      rsync
      ;;
  esac
fi
sudo chown zimbra:zimbra /opt/zimbra/live_sync/sync_commands
sudo chmod +x /opt/zimbra/live_sync/sync_commands
  • Configuration file

The configuration file simply contains the IP addresses of the live and mirror server. The order is not important since this is worked out by the script by seeing which IP address is assigned to the local machine. The configuration file name is saved as live_sync.conf and saved in the /opt/zimbra/live_sync directory and readable by user zimbra. The following is an example, you obviously should use the real IP addresses of your own live and mirror servers.

server1="192.168.108.10"
server2="192.168.108.11"
  • Enabling redo.log

For the Network edition, redo logs are already being created and are periodically moved to create incremental backups. For the open source version, redo logs archiving must be enabled.

To see the current redo log related settings, type the following as user zimbra:

sudo su - zimbra -c "zmprov gacf | grep "RedoLog"

To enable redo log rollover on the open source version, type…

sudo su - zimbra -c "zmprov mcf zimbraRedoLogDeleteOnRollover FALSE"
sudo su - zimbra -c "zmprov mcf zimbraRedoLogEnabled TRUE"

You may also want to make the redo log rotation more frequent to guarantee a file-system consistent redo log on the mirror server at least up to the last, say, thirty minutes. The live-streamed redo.log may not be consistent although it is unlikely this will ever be a problem except with the very last record in the log.

For example, to force rollover every half an hour, type…

sudo su - zimbra -c "zmprov mcf zimbraRedoLogRolloverFileSizeKB 1"
sudo su - zimbra -c "zmprov mcf zimbraRedoLogRolloverMinFileAge 30"

This will rollover if the size of the redo log is over 1KB after 30 mins, which is very likely unless the mail server is not sending or receiving any mail at all during this time.

You may want to reduce the zimbraRedoLogRolloverMinFileAge even further while setting and testing this script just so you don’t have to wait too long to see stuff happening between the servers.

Mirror Server

The mirror server should ideally have the same operating system as the live server and must have exactly the same version of Zimbra installed.

The hostname must also be exactly the same.

  • Install inotify-tools

inotify tools are required. For RHEL/CentOS, inotify-tools is provided by the epel-release repository.

As user root: (Centos)

yum install epel-release -y
yum install inotify-tools -y

while for Ubuntu, as user root:

 apt install inotify-tools -y
  • Create log rotation

The script will create a log file which can be handled by logrotate. As user root:

echo "/opt/zimbra/live_sync/log/live_sync.log {
    daily
    missingok
    copytruncate
    rotate 7
    notifempty
    compress
}">/etc/logrotate.d/zimbra_live_sync
  • First rsync between the servers

We now perform the first copy of the zimbra directory between the live and mirror server. On the mirror server, we must stop Zimbra. We leave Zimbra running on the live server for now to reduce downtime. This is what we call the hot sync.

We’re going to copy the whole /opt/zimbra so the script, its configuration and Zimbra itself is copied to the remote second node (mirror).

The following rsync command is run on the mirror server. Substitute live_server in the command below with the hostname or IP address of the live server.

Note: Copying the sparse files used by LDAP in Zimbra 8+ takes a very long time even though the file is small.

Generate a SSH key-pair without a passphrase on the mirror_server and transfer the id_rsa.pub key file to the live_server and add it to the /root/.ssh/authorized_keys file before running the rsync command below.

NB: Remote root login should be enabled on the live_server. As user root:

service zimbra stop
rsync -aHvz -e “ssh -p 22” --force --delete --sparse live_server:/opt/zimbra/ /opt/zimbra/
  • Second rsync between servers

This is where we need to stop Zimbra on the live server (cold sync) so that we can copy a consistent /opt/zimbra directory from the live to the mirror server. This is the only downtime required.

On the live server as user root:

service zimbra stop

On the mirror server as user root:

rsync -aHvz -e “ssh -p 22” --force --delete --sparse live_server:/opt/zimbra/ /opt/zimbra/

On the live server as user root:

service zimbra start

On the mirror server as user root (just to make sure we have a viable copy of Zimbra):

service zimbra start
service zimbra status
service zimbra stop
  • Running the script

Not only have we copied all the Zimbra data from the live to the mirror server, we have also copied the script and SSH keys. We should now be able to try running the script.

On the mirror server as user zimbra:

sudo su - zimbra -c  "cd /opt/zimbra/live_sync"
sudo su - zimbra -c  "./live_syncd start"

All being well the script has started without any complaints and we can now tail the log file to see that it is syncing as expected.

tail -f log/live_sync.log

(CTRL-C to exit tail command)

  • Init script

There is also an init script that may be useful for Ubuntu and Redhat/CentOS.

  1. CentOS/Redhat

Copy the below script as /etc/init.d/zimbra_live_sync.

sudo vim /etc/init.d/zimbra_live_sync
#!/bin/bash
# 
# ***** BEGIN LICENSE BLOCK *****
#    This program is free software: you can redistribute it and/or modify
#    it under the terms of the GNU General Public License as published by
#    the Free Software Foundation, either version 3 of the License, or
#    (at your option) any later version.
 
#    This program is distributed in the hope that it will be useful,
#    but WITHOUT ANY WARRANTY; without even the implied warranty of
#    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
#    GNU General Public License for more details.
 
#    You should have received a copy of the GNU General Public License
#    along with this program.  If not, see <http://www.gnu.org/licenses/>
# ***** END LICENSE BLOCK *****
# 
#
# Init file for zimbra live sync
#
# chkconfig: 345 99 01
# description: Zimbra live sync service
#
### BEGIN INIT INFO
# Provides:       zimbra_live_sync
# Required-Start: $network $remote_fs $syslog $time nscd cron
# Required-Stop:  $network $remote_fs $syslog $time
# Default-Start:  3 5
# Description:    Zimbra live sync service
### END INIT INFO
 
case "$1" in
        gentle-restart)
                su - zimbra -c "live_sync/live_syncd stop"
                su - zimbra -c "live_sync/live_syncd start"
                RETVAL=$?
                ;;
        restart)
                su - zimbra -c "live_sync/live_syncd kill"
                su - zimbra -c "live_sync/live_syncd start"
                RETVAL=$?
                ;;
        start)
                su - zimbra -c "live_sync/live_syncd start"
                RETVAL=$?
                ;;
        gentle-stop)
                su - zimbra -c "live_sync/live_syncd stop"
                RETVAL=$?
                ;;
        stop)
                su - zimbra -c "live_sync/live_syncd kill"
                RETVAL=$?
                ;;
        status)
                su - zimbra -c "live_sync/live_syncd status"
                RETVAL=$?
                ;;
        *)
                echo $"Usage: $0 {start|stop|restart|gentle-stop|gentle-restart|status}"
                RETVAL=1
                ;;
esac
exit $RETVAL

On both the live and the mirror server make the script executable and add the script to chkconfig.

CentOS/Redhat

chmod 755 /etc/init.d/zimbra_live_sync
chkconfig --add zimbra_live_sync

On the live server, make sure it doesn’t start on boot.

chkconfig zimbra_live_sync off

On the mirror server ensure that Zimbra doesn’t start on boot but the live sync script does.

chkconfig zimbra off
chkconfig zimbra_live_sync on

Ubuntu

chmod 755 /etc/init.d/zimbra_live_sync

On the mirror server, ensure Zimbra doesn’t start on boot but the live sync script does.

update-rc.d -f zimbra remove
update-rc.d zimbra_live_sync defaults
  • Failover

If the live server fails, then the procedure on the mirror server is simply to stop the live_sync script, and start Zimbra.

sudo su - zimbra
cd /opt/zimbra/live_sync
./live_syncd stop
zmcontrol start
  • Fallback

Simply run the script on the server to fail back to i.e. live and mirror are now reversed. As user zimbra (on ex-live server to be restored back to live):

cd /opt/zimbra/live_sync
./live_syncd start

Once the script has caught up and synced the two servers together. Stop Zimbra on the mirror server.

As user zimbra on mirror (failover server):

sudo su - zimbra -c "zmcontrol stop"

As user zimbra on live (restored server):

sudo su - zimbra -c "cd /opt/zimbra/live_sync"
sudo su - zimbra -c "./live_syncd stop"
sudo su - zimbra -c "zmcontrol zimbra start"

Nagios Integration

The script also generates useful status information for Nagios. The time since the last successful operation is measured and Nagios can raise an alert if any part of the script appears to have not been successful for a longer than expected time.

The Nagios script is to be run on the current Live server and can be run as any user as long as that user has read access to the status files created by live_syncd. These are in /opt/zimbra/live_sync/status. Save the script as check_zimbra_live_sync in the /opt/zimbra/live_sync directory.

sudo vim /opt/zimbra/live_sync/check_zimbra_live_sync
#!/bin/bash
#    This program is free software: you can redistribute it and/or modify
#    it under the terms of the GNU General Public License as published by
#    the Free Software Foundation, either version 3 of the License, or
#    (at your option) any later version.
 
#    This program is distributed in the hope that it will be useful,
#    but WITHOUT ANY WARRANTY; without even the implied warranty of
#    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
#    GNU General Public License for more details.
 
#    You should have received a copy of the GNU General Public License
#    along with this program.  If not, see <http://www.gnu.org/licenses/>
 
##########################################################################
# Title      :  check_zimbra_live_sync
# Author     :  Simon Blandford <simon -at- onepointltd -dt- com>
# Date       :  2012-08-25
# Requires   :  live_syncd
# Category   :  Administration
# Version    :  2.0.0
# Copyright  :  Simon Blandford, Onepoint Consulting Limited
# License    :  GPLv3 (see above)
##########################################################################
# Description
# Nagios plug-in for Zimbra live sync script
##########################################################################
 
#******************************************************************************
#********************** Constants *********************************************
#******************************************************************************
 
ZIMBRA_DIR="/opt/zimbra"
BASE_DIR="$ZIMBRA_DIR""/live_sync"
STATUS_DIR="$BASE_DIR""/status"
 
#Files that need age testing
LAST_GOOD_REDO_REPLAY="$STATUS_DIR""/last_good_redo_replay"
LAST_GOOD_REDO_SYNC="$STATUS_DIR""/last_good_redo_sync"
LAST_GOOD_REDO_STREAM="$STATUS_DIR""/last_good_redo_stream"
LAST_GOOD_LDAP_SYNC="$STATUS_DIR""/last_good_ldap_sync"
LAST_GOOD_LDAP_START="$STATUS_DIR""/last_good_ldap_start"
 
#******************************************************************************
#********************** Functions *********************************************
#******************************************************************************
 
usage () {
  echo "Usage: check_zimbra_live_sync -w <hours> -c <hours>" 
  exit 3
}
 
file_age () {
  echo $(( ($( date +%s) - $( stat -c "%Y" "$1" )) / 3600 ))
}
 
#Extract name of function being tested from file name
function_id () {
  echo "$1" | grep -Eo "[^_]*_[^_]*$"
}
 
file_report () {
  local file_under_test
  file_under_test="$1"
  age_of_file_under_test=$( file_age "$file_under_test" )
  
  #No file returns UNKOWN status
  if [ ! -f "$file_under_test" ]; then
    if [ ${#affected_function_list} -gt 1 ] && [ $status_code -eq 3 ]; then
      affected_function_list="$affected_function_list"",""$( function_id "$file_under_test" )"
    else
      affected_function_list="$( function_id "$file_under_test" )"
    fi
    status_code=3
    status_msg="Unknown"
    return
  fi
  
  #Test for files older than the critical number of seconds. Only test if no unknown status
  if [ $status_code -lt 3 ]; then
    if [ $age_of_file_under_test -ge $c ]; then
      if [ ${#affected_function_list} -gt 1 ] && [ $status_code -eq 2 ]; then
        affected_function_list="$affected_function_list"",""$( function_id "$file_under_test" )"
        affected_function_list="$affected_function_list""(""$age_of_file_under_test"" hours)"
      else
        affected_function_list="$( function_id "$file_under_test" )""(""$age_of_file_under_test"" hours)"
      fi
      status_code=2
      status_msg="Critical"
      return   
    fi
  fi
  
  #Test for files older than warning number of seconds. Only test if no unknown or critical status
  if [ $status_code -lt 2 ]; then
    if [ $age_of_file_under_test -ge $w ]; then
      if [ ${#affected_function_list} -gt 1 ] && [ $status_code -eq 1 ]; then
        affected_function_list="$affected_function_list"",""$( function_id "$file_under_test" )"
        affected_function_list="$affected_function_list""(""$age_of_file_under_test"" hours)"
      else
        affected_function_list="$( function_id "$file_under_test" )""(""$age_of_file_under_test"" hours)"
      fi
      status_code=1
      status_msg="Warning"
      return   
    fi
  fi
}
 
#******************************************************************************
#********************** Main Program ******************************************
#******************************************************************************
 
status_code=0
status_msg="OK"
affected_function_list="Live sync up to date"
 
for i in $@; do
  if [ $get_w ]; then
    w=$( echo "$i" | grep -Eo "[0-9]+" );
    unset get_w
  fi
  if [ $get_c ]; then
    c=$( echo "$i" | grep -Eo "[0-9]+" );
    unset get_c
  fi
  [ "x""$i" == "x-w" ] && get_w=1
  [ "x""$i" == "x-c" ] && get_c=1
done
 
if [ ! $w ] || [ ! $c ]; then
  usage
fi
 
file_report "$LAST_GOOD_REDO_REPLAY"
file_report "$LAST_GOOD_REDO_SYNC"
file_report "$LAST_GOOD_REDO_STREAM"
file_report "$LAST_GOOD_LDAP_SYNC"
file_report "$LAST_GOOD_LDAP_START"
 
echo "$status_msg"" : ""$affected_function_list"
exit $status_code

Use the following line in /etc/nagios/nrpe.cfg or in the /etc/nagios/nrpe.d directory to allow the script to be called by nrpe.

command[check_zimbra_live_sync]=/usr/lib/nagios/plugins/contrib/check_zimbra_live_sync -w $ARG1$ -c $ARG2$

You may need to adjust the path to the nagios plugin script depending on where the contributed scripts are stored.

Conclusion

To replay redo logs only requires that the mailbox process is stopped. This is done automatically by the script. The script will work whether Zimbra has been started on the mirror or not as it will enable or disable services as and when it needs them. Keeping the rest of Zimbra running will drastically reduce the time it takes to fail over. This is only an advantage when access to the server domain can be quickly flipped or has a failover mechanism.