-#/bin/bash
+#!/bin/bash
+# Author: Carl Baldwin & Alan Pippin
+# Description: This script replicates a remote zfs filesystem to a local zfs pool.
+# This script will keep all snapshots in sync, removing the ones
+# that have been deleted since the last replicate was performed.
+# This script will only send the new, or missing, snapshots since
+# the last replicate was performed.
# Usage: replicate <hostname> <zfs filesystem>
+
+# source our configuration
+config="${0%/*}/zfs-scripts.conf"
+[ -e "${config}.dist" ] && . ${config}.dist
+[ -e "${config}" ] && . ${config}
+
+# command line arg parsing
remote=$1
remote_fs=$2
remote_pool=${2%%/*}
-remote_lockdir="/tmp/zfs-admin-lock"
-hostname=`hostname`
-
-# Set the name of the local pool used to store the backup of the remote
-local_pool=backups
-
-# Set the email address to send notification to
-mailto=root@pippins.net
-mailx=/usr/bin/mailx
-# When this variable is set, local filesystems will be destroyed
-# before receiving a full streams into them from the remote source.
-destroy_local_filesystem_on_full_replicate=0
+# return non-zero exit code if any command in the pipe fails
+set -o pipefail
-# The ssh connection doesn't find zfs without this.
-zfs=/usr/sbin/zfs
+# get the backup pool from the command line or the config file if not specified
+if [[ -n $3 ]]; then
+ backup_pool=$3
+else
+ backup_pool=${backup_pool%% *} # use the first backup pool if none specified
+fi
# Setup our cleanup and exit trap
cleanup() {
if [[ -e "$remote_list" ]]; then
rm -f $remote_list
fi
- if [[ -n "$remote" ]]; then
- ssh $remote ls -d "$remote_lockdir" > /dev/null 2>&1
- if [[ $? == 0 ]]; then
- ssh $remote rm -rf "$remote_lockdir"
- fi
+ $ssh $remote ls -d "$lockdir" > /dev/null 2>&1
+ if [[ $? == 0 ]]; then
+ $ssh $remote rm -rf "$lockdir"
fi
}
fatal_and_exit() {
echo -e 2>&1 "$1"
+ # Destroy the current backup markers from the local backup_pool and remote_pool if they exist
+ if [[ -n "$current_backup_marker" ]]; then
+ # Local backup pool current backup marker
+ $zfs list -t snapshot ${backup_pool}/${current_backup_marker} > /dev/null 2>&1
+ if [ $? == 0 ]; then
+ $zfs destroy ${backup_pool}/${current_backup_marker}
+ fi
+ # Remote pool current backup marker
+ $ssh $remote zfs list -t snapshot ${current_backup_marker} > /dev/null 2>&1
+ if [ $? == 0 ]; then
+ $ssh $remote $zfs destroy ${current_backup_marker}
+ fi
+ fi
+ # send email notification
if [[ -n "$2" ]]; then
echo -e "$1" | $mailx -s "zfs replicate on $hostname failed" "$2"
fi
+ # exit
exit 1
}
trap fatal_and_exit INT
trap cleanup EXIT
+# Declare a function to handle the replicate operation
+replicate() {
+ zfs_send="$1"
+ zfs_recv="$zfs receive -vF -d ${backup_pool}/${remote_pool}"
+ glue="$throttle $throttle_opt"
+ if [[ $throttle_enable == 1 && -e $throttle ]]; then
+ # handle using the glue in the local and remote host case properly
+ if [[ -z "$ssh" ]]; then
+ # local host glue case
+ $zfs_send | $glue | $zfs_recv
+ else
+ # remote host glue case
+ $ssh $remote "$zfs_send | $glue" | $zfs_recv
+ fi
+ else
+ # no glue case - works for both the local and remote host cases
+ $ssh $remote $zfs_send | $zfs_recv
+ fi
+ # The return code of the zfs_send | zfs_recv operation will be returned to the caller
+}
+
# Make sure we have valid arguments
if [[ -z "$remote" ]] || [[ -z "$remote_fs" ]]; then
fatal_and_exit "Usage: $0 <hostname> <zfs filesystem>"
fi
-# Make sure the local pool and local receiving filesystem exist, or print some errors
-zpool list -H "$local_pool" >/dev/null 2>&1
+# check for localhost
+if [[ $remote = "localhost" ]]; then
+ remote=""
+ ssh=""
+fi
+
+# Make sure the local backup pool and local receiving filesystem exist, or print some errors
+zpool list -H "$backup_pool" >/dev/null 2>&1
if [ $? != 0 ]; then
- fatal_and_exit "-E- The local pool, '$local_pool' doesn't seem to exist." $mailto
+ fatal_and_exit "-E- The local backup pool on $hostname, '$backup_pool' doesn't seem to exist." $mailto
fi
-zfs list "$local_pool/$remote_pool" >/dev/null 2>&1
+zfs list "$backup_pool/$remote_pool" >/dev/null 2>&1
if [ $? != 0 ]; then
- echo >&2 "-I- The local filesystem for the remote pool, '$local_pool/$remote_pool' doesn't seem to exist."
- echo >&2 " Creating the local filesystem to receive the remote pool into: $local_pool/$remote_pool"
- $zfs create $local_pool/$remote_pool
+ echo >&2 "-I- The local filesystem for the remote pool, '$backup_pool/$remote_pool' doesn't seem to exist."
+ echo >&2 " Creating the local filesystem to receive the remote pool into: $backup_pool/$remote_pool"
+ $zfs create $backup_pool/$remote_pool
if [ $? != 0 ]; then
- fatal_and_exit "-E- remote $zfs create command failed" $mailto
+ fatal_and_exit "-E- remote $zfs on $hostname create $backup_pool/$remote_pool command failed" $mailto
fi
fi
-# Obtain the zpool guid for the local pool
-local_pool_guid=`zpool get guid $local_pool 2>&1 | grep $local_pool | awk '{ print $3 }'`
-zpool get guid $local_pool > /dev/null 2>&1
+# Obtain the zpool guid for the local backup pool
+backup_pool_guid=`zpool get guid $backup_pool 2>&1 | grep $backup_pool | awk '{ print $3 }'`
+zpool get guid $backup_pool > /dev/null 2>&1
if [ $? != 0 ]; then
- fatal_and_exit "-E- Unable to extract the guid for the local pool: $local_pool" $mailto
+ fatal_and_exit "-E- Unable to extract the guid for the local backup pool on $hostname: $backup_pool" $mailto
fi
# Turn on shell verbosity
# Create the remote lockdir before continuing with the replicate
# Spinlock on creating the lock
maxsleeptime=60
-maxattempts=100
+maxattempts=500
attempts=0
while true; do
- ssh $remote mkdir "$remote_lockdir" >/dev/null 2>&1
+ $ssh $remote mkdir "$lockdir" >/dev/null 2>&1
if [ $? != 0 ]; then
# Another zfs admin tool is running.
# Wait a random amount of time and try again
if [[ $attempts -gt $maxattempts ]]; then
# We've exceeded our maximum while loop count
echo "-E- The zfs filesystem has been locked down. Skipping replicate operation."
- fail_msg=`ssh $remote ls -ld $remote_lockdir 2>&1`
- fatal_and_exit "zfs-replicate-all unable to obtain zfs admin lock:\n$fail_msg" $mailto
+ fail_msg=`$ssh $remote ls -ld $lockdir 2>&1`
+ fatal_and_exit "zfs-replicate-all on $hostname unable to obtain zfs admin lock:\n$fail_msg" $mailto
fi
done
# Setup our backup marker names
-current_backup_marker=${remote_fs}@current-backup-${local_pool_guid}
-previous_backup_marker=${remote_fs}@previous-backup-${local_pool_guid}
+current_backup_marker=${remote_fs}@current-backup-${backup_pool_guid}
+previous_backup_marker=${remote_fs}@previous-backup-${backup_pool_guid}
# List the snapshots on the remote machine.
remote_list=$(mktemp /tmp/replicate.XXXXXX)
-ssh $remote \
+$ssh $remote \
$zfs list -H -t snapshot |
grep ^${remote_fs}@ |
awk '{print$1}' > $remote_list
-if [ $? != 0 ]; then
- fatal_and_exit "-E- remote $zfs list command failed" $mailto
+if [[ $? != 0 ]] && [[ $expect_empty_remote_list == 0 ]]; then
+ echo "-W- Unable to find $remote_fs on the remote host $hostname. Unable to proceed since the"
+ echo " expect_empty_remote_list option has not been set to allow this, skipping replicate operation."
+ fatal_and_exit "-E- remote $zfs list on $hostname for $remote_fs command failed" $mailto
fi
# List the snapshots on the local machine.
# If you do, it can mess up the common finding algorithm below.
local_list=$(mktemp /tmp/replicate.XXXXXX)
$zfs list -H -t snapshot |
- grep ^${local_pool}/${remote_fs}@ |
- grep -v ^${local_pool}/${current_backup_marker} |
- awk "{gsub(/^$local_pool./,\"\",\$1); print\$1}" > $local_list
-if [ $? != 0 ]; then
- fatal_and_exit "-E- local $zfs list command failed" $mailto
-fi
+ grep ^${backup_pool}/${remote_fs}@ |
+ grep -v ^${backup_pool}/${current_backup_marker} |
+ awk "{gsub(/^$backup_pool./,\"\",\$1); print\$1}" > $local_list
+# If no local snapshots exist, we may need to send the entire filesystem, which we'll do later
+# So, no error check here as a non-zero return code means the local filesystem or snaps are missing
+# We'll catch this later on as a case where we need to send the entire filesystem
# Destroy the current backup marker snapshot on the remote system if it exists
grep -q ${current_backup_marker} $remote_list
if [ $? == 0 ]; then
- ssh $remote $zfs destroy ${current_backup_marker}
+ $ssh $remote $zfs destroy ${current_backup_marker}
if [ $? != 0 ]; then
- fatal_and_exit "-E- remote $zfs destroy command failed" $mailto
+ fatal_and_exit "-E- remote $zfs destroy $current_backup_marker on $hostname command failed" $mailto
fi
fi
# Create the current backup marker snapshot on the remote system
-ssh $remote $zfs snapshot ${current_backup_marker}
+$ssh $remote $zfs snapshot ${current_backup_marker}
if [ $? != 0 ]; then
- fatal_and_exit "-E- remote $zfs snapshot command failed" $mailto
+ fatal_and_exit "-E- remote $zfs snapshot $current_backup_marker on $hostname command failed" $mailto
fi
# Check to see if the previous backup marker exists in the remote snapshot list.
if [ $no_markers == 0 ]; then
# We found backup markers, incrementally send the new snaps
- # First, rollback the local pool to the previous backup marker in case the previous
+ # First, rollback the local backup pool to the previous backup marker in case the previous
# backup was interrupted for some reason. If we don't do this, the zfs send -R command
# below may complain about snaps already existing as it tries to resend from the
# previous backup marker again from a previously interrupted replicate.
- $zfs rollback -r ${local_pool}/${previous_backup_marker}
+ $zfs rollback -rf ${backup_pool}/${previous_backup_marker}
if [ $? != 0 ]; then
- fatal_and_exit "-E- remote incremental $zfs rollback command failed" $mailto
+ sleep 120
+ $zfs rollback -rf ${backup_pool}/${previous_backup_marker}
+ if [ $? != 0 ]; then
+ fatal_and_exit "-E- remote incremental $zfs rollback $backup_pool/$previous_backup_marker command failed on $hostname" $mailto
+ fi
fi
# Now it should be safe to send the snaps
- ssh $remote $zfs send -R -I${previous_backup_marker} ${current_backup_marker} |
- $zfs receive -vF -d ${local_pool}/${remote_pool}
+ replicate "$zfs send -Rc -I${previous_backup_marker} ${current_backup_marker}"
if [ $? != 0 ]; then
- fatal_and_exit "-E- remote incremental $zfs send command failed" $mailto
+ fatal_and_exit "-E- remote incremental $zfs send $previous_backup_marker command failed on $hostname" $mailto
fi
else
# We didn't find any backup markers, next check to see if we have a common snapshot.
if [[ -n "$common" ]]; then
# We found a common snapshot, incrementally send the new snaps
- ssh $remote $zfs send -R -I${common/*@/@} ${current_backup_marker} |
- $zfs receive -vF -d ${local_pool}/${remote_pool}
+ replicate "$zfs send -Rc -I${common/*@/@} ${current_backup_marker}"
if [ $? != 0 ]; then
- fatal_and_exit "-E- remote incremental $zfs send command failed" $mailto
+ fatal_and_exit "-E- remote incremental $zfs send $(common/*@/@) command failed on $hostname" $mailto
fi
else
# We did not find any markers or a common snapshot
# At this point, we'll have to send the entire filesystem
# Destroy the local filesystem if it exists before receving the full replicate
- zfs list ${local_pool}/${remote_fs} > /dev/null 2>&1
+ zfs list ${backup_pool}/${remote_fs} > /dev/null 2>&1
if [ $? == 0 ]; then
if [[ $destroy_local_filesystem_on_full_replicate == 1 ]]; then
- $zfs destroy -r ${local_pool}/${remote_fs}
+ $zfs destroy -r ${backup_pool}/${remote_fs}
if [ $? != 0 ]; then
- fatal_and_exit "-E- remote full $zfs destroy command failed" $mailto
+ fatal_and_exit "-E- remote full $zfs destroy $backup_pool/$remote_fs command failed on $hostname" $mailto
fi
else
echo "-W- We need to destroy a local filesystem before receiving a full stream."
echo " However, since the option is set to prevent this, skipping replicate operation."
- fatal_and_exit "unable to destroy local filesystem:\n$zfs destroy -r ${local_pool}/${remote_fs} not able to run" $mailto
+ fatal_and_exit "unable to destroy local filesystem:\n$zfs destroy -r ${backup_pool}/${remote_fs} not able to run on $hostname" $mailto
fi
fi
# Send the full filesystem
- ssh $remote $zfs send -R ${current_backup_marker} |
- $zfs receive -vF -d ${local_pool}/${remote_pool}
+ replicate "$zfs send -Rc ${current_backup_marker}"
if [ $? != 0 ]; then
- fatal_and_exit "-E- remote full $zfs send command failed" $mailto
+ fatal_and_exit "-E- remote full $zfs send $current_backup_marker command failed on $hostname" $mailto
fi
fi
fi
# destroy the previous backup markers now that we've replicated past them
# don't check the return codes here because these may not exist, and that is ok
-$zfs destroy ${local_pool}/${previous_backup_marker} > /dev/null 2>&1
-ssh $remote $zfs destroy ${previous_backup_marker} > /dev/null 2>&1
+$zfs destroy ${backup_pool}/${previous_backup_marker} > /dev/null 2>&1
+$ssh $remote $zfs destroy ${previous_backup_marker} > /dev/null 2>&1
+sleep 1
# Rename the current backup marker to be the previous backup marker
-$zfs rename ${local_pool}/${current_backup_marker} ${local_pool}/${previous_backup_marker}
+$zfs rename ${backup_pool}/${current_backup_marker} ${backup_pool}/${previous_backup_marker}
if [ $? != 0 ]; then
- fatal_and_exit "-E- local $zfs rename command failed" $mailto
+ fatal_and_exit "-E- local $zfs rename $backup_pool/$current_backup_marker command failed on $hostname" $mailto
fi
-ssh $remote $zfs rename ${current_backup_marker} ${previous_backup_marker}
+$ssh $remote $zfs rename ${current_backup_marker} ${previous_backup_marker}
if [ $? != 0 ]; then
- fatal_and_exit "-E- remote $zfs rename command failed" $mailto
+ fatal_and_exit "-E- remote $zfs rename $current_backup_marker command failed on $hostname" $mailto
fi
-