remote=$1
remote_fs=$2
remote_pool=${2%%/*}
-hostname=`hostname`
+
+# return non-zero exit code if any command in the pipe fails
+set -o pipefail
+
+# 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() {
}
fatal_and_exit() {
echo -e 2>&1 "$1"
- # Destroy the backup markers on the local filesystem if they exist
+ # Destroy the current backup markers from the local backup_pool and remote_pool if they exist
if [[ -n "$current_backup_marker" ]]; then
- zfs list -t snapshot ${backup_pool}/${current_backup_marker} > /dev/null 2>&1
+ # 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
- fi
- if [[ -n "$previous_backup_marker" ]]; then
- zfs list -t snapshot ${backup_pool}/${previous_backup_marker} > /dev/null 2>&1
+ # Remote pool current backup marker
+ $ssh $remote zfs list -t snapshot ${current_backup_marker} > /dev/null 2>&1
if [ $? == 0 ]; then
- $zfs destroy ${backup_pool}/${previous_backup_marker}
+ $ssh $remote $zfs destroy ${current_backup_marker}
fi
fi
# send email notification
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: $0 <hostname> <zfs filesystem>"
# Make sure the local backup pool and local receiving filesystem exist, or print some errors
zpool list -H "$backup_pool" >/dev/null 2>&1
if [ $? != 0 ]; then
- fatal_and_exit "-E- The local backup pool, '$backup_pool' doesn't seem to exist." $mailto
+ fatal_and_exit "-E- The local backup pool on $hostname, '$backup_pool' doesn't seem to exist." $mailto
fi
zfs list "$backup_pool/$remote_pool" >/dev/null 2>&1
if [ $? != 0 ]; then
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
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 backup pool: $backup_pool" $mailto
+ fatal_and_exit "-E- Unable to extract the guid for the local backup pool on $hostname: $backup_pool" $mailto
fi
# Turn on shell verbosity
# We've exceeded our maximum while loop count
echo "-E- The zfs filesystem has been locked down. Skipping replicate operation."
fail_msg=`$ssh $remote ls -ld $lockdir 2>&1`
- fatal_and_exit "zfs-replicate-all unable to obtain zfs admin lock:\n$fail_msg" $mailto
+ fatal_and_exit "zfs-replicate-all on $hostname unable to obtain zfs admin lock:\n$fail_msg" $mailto
fi
done
$zfs list -H -t snapshot |
grep ^${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" $mailto
fi
# List the snapshots on the local machine.
grep ^${backup_pool}/${remote_fs}@ |
grep -v ^${backup_pool}/${current_backup_marker} |
awk "{gsub(/^$backup_pool./,\"\",\$1); print\$1}" > $local_list
-if [ $? != 0 ]; then
- fatal_and_exit "-E- local $zfs list command failed" $mailto
-fi
+# 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}
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}
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.
# 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 ${backup_pool}/${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
- if [[ $throttle_enable == 1 && -e $throttle ]]; then
- $ssh $remote $zfs send -R -I${previous_backup_marker} ${current_backup_marker} |
- $throttle $throttle_opt | $zfs receive -vF -d ${backup_pool}/${remote_pool}
- else
- $ssh $remote $zfs send -R -I${previous_backup_marker} ${current_backup_marker} |
- $zfs receive -vF -d ${backup_pool}/${remote_pool}
- fi
+ 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
- if [[ $throttle_enable == 1 && -e $throttle ]]; then
- $ssh $remote $zfs send -R -I${common/*@/@} ${current_backup_marker} |
- $throttle $throttle_opt | $zfs receive -vF -d ${backup_pool}/${remote_pool}
- else
- $ssh $remote $zfs send -R -I${common/*@/@} ${current_backup_marker} |
- $zfs receive -vF -d ${backup_pool}/${remote_pool}
- fi
+ 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
if [[ $destroy_local_filesystem_on_full_replicate == 1 ]]; then
$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 ${backup_pool}/${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
- if [[ $throttle_enable == 1 && -e $throttle ]]; then
- $ssh $remote $zfs send -R ${current_backup_marker} |
- $throttle $throttle_opt | $zfs receive -vF -d ${backup_pool}/${remote_pool}
- else
- $ssh $remote $zfs send -R ${current_backup_marker} |
- $zfs receive -vF -d ${backup_pool}/${remote_pool}
- fi
+ 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
# don't check the return codes here because these may not exist, and that is ok
$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 ${backup_pool}/${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}
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