ZFS sync script

Like all of my scripts this script is coming without any guaranties!!! You can use it on your own risk!

About the script

• It uses mbuffer. It is easy to compile.
• It uses gawk.
• The variable SECURE defines if you want to use ssh to encrypt your stream. Set it to yes or no.
• To mark the datasets to copy from the backup host use this on the source:

# /usr/sbin/zfs set de.timmann:auto-backup=<backup host> <dataset>

• Run the script on the destination/backup host.
• Make an ssh-key exchange to login without password.
• If you don't want to use root as backup-user on source host do this to create a zfssync user:

# useradd -m zfssync
# passwd -N zfssync
# usermod -K type=normal zfssync

Good luck!

zfs_sync.sh

#!/usr/bin/bash

# Written by Lars Timmann <L@rs.Timmann.de> 2013

# This script is a rotten bunch of code... rewrite it!

SRC_USER=zfssync
SRC_HOST=my_source_server
SRC_POOL=my_source_zpool
DST_POOL=my_local_destination_zpool
INITIAL_COPIES=3

# Default yes means use SSH for encryption over the net. Every other value means just mbuffer.
SECURE="no" 

MBUFFER_PORT=10001
MBUFFER_OPTS="-v 0 --md5 -s 128k -m 256M"

ZFS=/usr/sbin/zfs
SSH="/usr/bin/ssh -xc arcfour128"
#AWK=/usr/bin/gawk
AWK=/opt/sfw/bin/gawk
GREP=/usr/bin/grep
DATE=/usr/bin/date
MD5="/usr/bin/digest -a md5"
ROUTE=/usr/sbin/route
MBUFFER="/opt/mbuffer/bin/mbuffer"

# Guess the right IP for communication with source host
DST_HOST=$(${ROUTE} -vn get ${SRC_HOST} | ${AWK} '{ip=$2}END{print ip}')

MYNAME=$(/usr/bin/basename $0 .sh)
MYSELF=$(/usr/bin/hostname)

SRC_DATASETS=/tmp/${MYNAME}_src_ds.out
DST_DATASETS=/tmp/${MYNAME}_dst_ds.out
LOCK_FILE=/var/run/${MYNAME}.lck
TMP_FILE1=/tmp/${MYNAME}.tmp1
TMP_FILE2=/tmp/${MYNAME}.tmp2
BACKUP_PROPERTY="de.timmann:auto-backup"

START_TIME=$(${AWK} 'BEGIN{printf systime();}')
${AWK} -v time=${START_TIME} 'BEGIN{print "START:",strftime("%d.%m.%Y %H:%M.%S",time)}'


# Was tun bei Unterbrechung
# -------------------------
trap 'echo "\n--- Signal empfangen: Exiting ...\n"; \
      date ; \
      rm -f ${LOCK_FILE}; \
      sleep 3; kill -9 ${!} 2>/dev/null; exit 1' 1 2 3 13 14 15 18
###########################


if [ -f ${LOCK_FILE} ] ; then
  echo "$0 is allready running as PID $(/usr/bin/cat ${LOCK_FILE}) look in ${LOCK_FILE}"
  exit 1
else
  echo $$ > ${LOCK_FILE}
fi

${SSH} ${SRC_USER:+"${SRC_USER}@"}${SRC_HOST} "${ZFS} list -rH -t filesystem,snapshot,volume -o name,type,${BACKUP_PROPERTY} -s creation ${SRC_POOL}" > ${SRC_DATASETS} &
${ZFS} list -rH -t filesystem,snapshot,volume -o name,type -s creation ${DST_POOL} > ${DST_DATASETS} &
wait

function convert_to_poolname () {
  from_zfs=$1
  search=$2
  replace=$3
  echo ${from_zfs} | sed -e "s#^${search}#${replace}#g"
}

function is_available () {
  snapshot=$1
  list=$2
  ${AWK} -v snapshot=${snapshot} 'BEGIN{rc=1;}$1 == snapshot{print $1; rc=0;}END{exit rc;}' ${list}
  return $?
}

function expire_dst_pool_snapshots () {
  days_to_keep=$1
  min_to_keep=$2
  for expired_zfs in $(
  ${ZFS} list -o creation,name -S creation -t snapshot | \
  ${AWK} \
        -v days_to_keep=${days_to_keep} \
        -v min_to_keep=${min_to_keep} \
        -v DST_POOL="^${DST_POOL}" \
  '
  BEGIN{
    split("Jan:Feb:Mar:Apr:May:Jun:Jul:Aug:Sep:Oct:Nov:Dec",mon,":");
    for(m in mon){
      month[mon[m]]=m
    };
    expire_date=systime()-days_to_keep*60*60*24
  }
  $NF ~ DST_POOL {
    filesystem=$NF;
    gsub(/@.*$/,"",filesystem);
    split($4,time,":");
    filesystem_date=mktime(sprintf("%d %02d %02d %02d %02d 00", $5, month[$2], $3, time[1], time[2]));
    count[filesystem]++;
    if(filesystem_date < expire_date && count[filesystem] > min_to_keep )
    {
      print $NF;
    }
  }')
  do
    printf "$(${DATE}) Destroying snapshot ${expired_zfs}\n"
    ${ZFS} destroy ${expired_zfs}
  done
}

function get_src_list () {
  ${AWK} -v backup_server=${MYSELF} '
    ( $2=="filesystem" || $2=="volume" ) && $3==backup_server {
      path[$1]=1;
      for(name in path){
        # delete name from list, if name is substring of $1
        if( index($1,name)==1 && name != $1 && path[name]!=0 ){
          path[name]=0;
        }
      }
    }
    END{
      for(name in path){
        if(path[name]==1) print name
      }
    }
  ' ${SRC_DATASETS}
}

function first_snapshot () {
  ${AWK} -v zfs="${1}@" '
    $2=="snapshot" && $1 ~ zfs {
      first=$1;
      # und raus...
      nextfile;
    }
    END{
      print first;
    }
  ' $2
}

function last_snapshot () {
  ${AWK} -v zfs="${1}@" '
    $2=="snapshot" && $1 ~ zfs {
      last=$1;
    }
    END{
      print last;
    }
  ' $2
}

function get_recursive () {
  src_host=$1
  src_datasets=$2
  first=$3
  last=$4
  dst_pool=$5
  dst_datasets=$6

  if [ $# -lt 6 ] ; then
    echo "Called from line ${BASH_LINENO[$i]} with $# Arguments"
    end 1
  fi

  src_zfs=$(echo ${first} | ${AWK} -F'@' '{print $1}')
  first_snap=$(echo ${first} | ${AWK} -F'@' '{print FS""$2}')

  echo "Getting snapshot ${zfs}..."

  if [ "_${SECURE}_" == "_yes_" ]
  then
    # setup receiver
    ${MBUFFER} ${MBUFFER_OPTS} -l ${TMP_FILE1} -I 127.0.0.1:${MBUFFER_PORT} | \
      ${ZFS} recv -vFd ${dst_pool} 2>&1 &

    # start sender
    ${SSH} ${SRC_USER:+"${SRC_USER}@"}${SRC_HOST} \
      -R ${MBUFFER_PORT}:127.0.0.1:${MBUFFER_PORT} \
      "${ZFS} send -I ${first_snap} ${last} | ${MBUFFER} ${MBUFFER_OPTS} -O 127.0.0.1:${MBUFFER_PORT} 2>&1" >${TMP_FILE2} &
  else
    # setup receiver
    ${MBUFFER} ${MBUFFER_OPTS} -l ${TMP_FILE1} -I ${MBUFFER_PORT} | \
      ${ZFS} recv -vFd ${dst_pool} 2>&1 &

    # start sender
    ${SSH} ${SRC_USER:+"${SRC_USER}@"}${SRC_HOST} \
      "${ZFS} send -I ${first_snap} ${last} | ${MBUFFER} ${MBUFFER_OPTS} -O ${DST_HOST}:${MBUFFER_PORT} 2>&1" >${TMP_FILE2} &
  fi
  wait
  local_md5=$(grep md5 ${TMP_FILE1})
  remote_md5=$(grep md5 ${TMP_FILE2})
  local_summary=$(grep summary ${TMP_FILE1})
  remote_summary=$(grep summary ${TMP_FILE2})
  printf "remote %s\nlocal  %s\n" "${remote_md5}" "${local_md5}"
  printf "remote %s\nlocal  %s\n" "${remote_summary}" "${local_summary}"
  rm -f ${TMP_FILE1} ${TMP_FILE2}
}

function get_snapshot () {
  src_host=$1
  src_datasets=$2
  zfs=$3
  dst_pool=$4
  dst_datasets=$5
  if [ -z "$(is_available ${zfs} ${dst_datasets})" ] ; then
    echo "Getting snapshot ${zfs}..."
    if [ "_${SECURE}_" == "_yes_" ]
    then
      # setup receiver
      ${MBUFFER} ${MBUFFER_OPTS} -l ${TMP_FILE1} -I 127.0.0.1:${MBUFFER_PORT} | \
        ${ZFS} recv -vFd ${dst_pool} 2>&1 &
  
      # start sender
      ${SSH} ${SRC_USER:+"${SRC_USER}@"}${SRC_HOST} \
        -R ${MBUFFER_PORT}:127.0.0.1:${MBUFFER_PORT} \
        "${ZFS} send -R ${zfs} | ${MBUFFER} ${MBUFFER_OPTS} -O 127.0.0.1:${MBUFFER_PORT} 2>&1" >${TMP_FILE2} &
    else
      # setup receiver
      ${MBUFFER} ${MBUFFER_OPTS} -l ${TMP_FILE1} -I ${MBUFFER_PORT} | \
        ${ZFS} recv -vFd ${dst_pool} 2>&1 &
  
      # start sender
      ${SSH} ${SRC_USER:+"${SRC_USER}@"}${SRC_HOST} \
        "${ZFS} send -R ${zfs} | ${MBUFFER} ${MBUFFER_OPTS} -O ${DST_HOST}:${MBUFFER_PORT} 2>&1" >${TMP_FILE2} &
    fi
    wait
    local_md5=$(grep md5 ${TMP_FILE1})
    remote_md5=$(grep md5 ${TMP_FILE2})
    local_summary=$(grep summary ${TMP_FILE1})
    remote_summary=$(grep summary ${TMP_FILE2})
    printf "remote %s\nlocal  %s\n" "${remote_md5}" "${local_md5}"
    printf "remote %s\nlocal  %s\n" "${remote_summary}" "${local_summary}"
    rm -f ${TMP_FILE1} ${TMP_FILE2}
  fi
}

function timestamp () {
  echo $(${DATE} '+%Y%m%d-%H:%M')
}

function expire_backup_snapshots () {
  src_host=$1
  src_datasets=$2
  dst_datasets=$3
  src_last_to_keep=$4
  dst_pool=$5

  src_zfs=$(echo ${src_last_to_keep} | ${AWK} -F'@' '{print $1}')
  dst_zfs=$(convert_to_poolname ${src_zfs} ${SRC_POOL} ${dst_pool})
  dst_last_to_keep=$(convert_to_poolname ${src_last_to_keep} ${SRC_POOL} ${dst_pool})

  echo "Deleting old backup snapshots before ${dst_last_to_keep}"
  if ( ${ZFS} list -o name ${dst_last_to_keep} >/dev/null 2>&1 ) ; then
    for src_backup_snapshot in $(${AWK} -v src_backup="${src_zfs}@backup" -v src_last_to_keep="${src_last_to_keep}" '
        $1 == src_last_to_keep {
          exit 0;
        }
        $1 ~ src_backup {
          print $1;
        }
        ' ${src_datasets}) 
    do
      printf "\tDeleting on src ${src_backup_snapshot} ..."
      if ( ${SSH} ${SRC_USER:+"${SRC_USER}@"}${SRC_HOST} "${ZFS} destroy ${src_backup_snapshot}" ) ; then
        echo "done"
      else
        echo "failed"
      fi
    done
    for dst_backup_snapshot in $(${AWK} -v dst_backup="${dst_zfs}@backup" -v dst_last_to_keep=${dst_last_to_keep} '
        $1 == dst_last_to_keep {
          exit 0;
        }
        $1 ~ dst_backup {
          print $1;
        }
        ' ${dst_datasets}) 
    do
      printf "\tDeleting on destination ${dst_backup_snapshot} ..."
      if ( ${ZFS} destroy ${dst_backup_snapshot} ) ; then
        echo "done"
      else
        echo "failed"
      fi
    done
  else
    echo "Strange we do not have the copy of ${dst_last_to_keep} => STOP!"
  fi
}

function end () {
  /usr/bin/rm -f ${LOCK_FILE}
  exit $1
}

for src_zfs in $(get_src_list) ; do
  echo "Evaluating ${src_zfs}"
  dst_zfs=$(convert_to_poolname ${src_zfs} ${SRC_POOL} ${DST_POOL})
  last_src=$(last_snapshot ${src_zfs} ${SRC_DATASETS})
  last_dst=$(last_snapshot ${dst_zfs} ${DST_DATASETS})
  last_backup_src=$(${AWK} -v zfs="${src_zfs}@backup" '$1 ~ zfs{last=$1}END{print last}' ${SRC_DATASETS})
  last_backup_dst=$(${AWK} -v zfs="${dst_zfs}@backup" '$1 ~ zfs{last=$1}END{print last}' ${DST_DATASETS})
  last_dst_on_src=$(convert_to_poolname ${last_dst} ${DST_POOL} ${SRC_POOL})
  this_backup_src=${src_zfs}@backup_$(timestamp)

  # Create snapshot for incremental backups
  ${SSH} ${SRC_USER:+"${SRC_USER}@"}${SRC_HOST} "${ZFS} snapshot ${this_backup_src}"
  if [ -n "$(is_available ${dst_zfs} ${DST_DATASETS})" -a -z "${last_dst}" ] ; then
    echo "zfs is on dst, but no snapshots. Getting ${last_src}..."
    get_snapshot ${SRC_HOST} ${SRC_DATASETS} ${last_src} ${DST_POOL} ${DST_DATASETS}
  # Look for last backup snapshot on destination
  elif [ -n "${last_backup_dst}" ] ; then
    # Name of last backup snapshot on src
    last_dst_backup_on_src=$(convert_to_poolname ${last_backup_dst} ${DST_POOL} ${SRC_POOL})

    # If converted name is not empty and snapshot is in the list of src snapshots
    # then get all snapshots from last backup until now
    if [ -n "${last_dst_backup_on_src}" ] ; then
      if [ -n "$(is_available ${last_dst_backup_on_src} ${SRC_DATASETS})" ] ; then
        # Get the snapshot of this backup
        printf "%s\tsnapshot\n" ${this_backup_src} >> ${SRC_DATASETS}
        get_recursive ${SRC_HOST} ${SRC_DATASETS} ${last_dst_backup_on_src} ${this_backup_src} ${DST_POOL} ${DST_DATASETS} && \
          expire_backup_snapshots ${SRC_HOST} ${SRC_DATASETS} ${DST_DATASETS} ${this_backup_src} ${DST_POOL}
      fi
    fi
  elif [ -n "$(is_available ${dst_zfs} ${DST_DATASETS})" ] ; then
    # No last backup snapshot on dst but we have snapshots
    if [ -n "$(is_available ${last_dst_on_src} ${SRC_DATASETS})" ] ; then
      echo "Try to backup from ${last_dst_on_src} to ${this_backup_src}"
      first=${last_dst_on_src}
      last=${last_src}
      get_recursive ${SRC_HOST} ${SRC_DATASETS} ${first} ${last} ${DST_POOL} ${DST_DATASETS} && \
        expire_backup_snapshots ${SRC_HOST} ${SRC_DATASETS} ${DST_DATASETS} ${this_backup_src} ${DST_POOL}
      # Get the snapshot of this backup
      printf "%s\tsnapshot\n" ${this_backup_src} >> ${SRC_DATASETS}
      get_recursive ${SRC_HOST} ${SRC_DATASETS} ${last} ${this_backup_src} ${DST_POOL} ${DST_DATASETS} && \
        expire_backup_snapshots ${SRC_HOST} ${SRC_DATASETS} ${DST_DATASETS} ${this_backup_src} ${DST_POOL}
    else
      echo "OK I tried hard... now it is your job..."
    fi
  else
    # No existing copies for this zfs. Get the last <INITIAL_COPIES> copies
    first=$(${AWK} -v zfs=${src_zfs} -v intitial_copies=$((${INITIAL_COPIES}-1)) '
      $1 ~ zfs && $2=="snapshot" {
        last[++count]=$1;
      }
      END {
        if(count>intitial_copies){
          print last[count-intitial_copies]
        }else{
          print last[1]
        }
      }' ${SRC_DATASETS})
    last=$( ${AWK} -v zfs=${src_zfs} '$1 ~ zfs && $2=="snapshot"{last=$1}END{print last}' ${SRC_DATASETS} )
    get_snapshot ${SRC_HOST} ${SRC_DATASETS} ${first} ${DST_POOL} ${DST_DATASETS}
    get_recursive ${SRC_HOST} ${SRC_DATASETS} ${first} ${last} ${DST_POOL} ${DST_DATASETS} && \
      expire_backup_snapshots ${SRC_HOST} ${SRC_DATASETS} ${DST_DATASETS} ${this_backup_src} ${DST_POOL}
    # Get the snapshot of this backup
    printf "%s\tsnapshot\n" ${this_backup_src} >> ${SRC_DATASETS}
    get_recursive ${SRC_HOST} ${SRC_DATASETS} ${last} ${this_backup_src} ${DST_POOL} ${DST_DATASETS} && \
      expire_backup_snapshots ${SRC_HOST} ${SRC_DATASETS} ${DST_DATASETS} ${this_backup_src} ${DST_POOL}
  fi  

  echo
  echo --------------------------------------------------------------------------------
  date
  echo
done

# expire_dst_pool_snapshots days_to_keep min_to_keep
expire_dst_pool_snapshots 34 70

END_TIME=$(${AWK} 'BEGIN{printf systime();}')
${AWK} -v time=${END_TIME} 'BEGIN{print "END  :",strftime("%d.%m.%Y %H:%M.%S",time)}'
${AWK} -v start=${START_TIME} -v end=${END_TIME} 'BEGIN{print "DURATION:",strftime("%H:%M.%S",end-start-3600*strftime("%H",0))}'
end 0