Snapshot and backup a ZFS filesystem to a secondary pool

zfs-backup-readme.md

ZFS Backup Script

I use this script to backup a ZFS filesystem used for PostgreSQL and OpenStreetMap data. The script creates a new snapshot with the current date (YYYY-MM-DD), estimates the size between this snapshot and the previous snapshot, then transfers the snapshot to a secondary backup pool using ZFS send. The progress of the transfer is displayed with pv.

This script will not destroy snapshots. It will print out the destroy command if you want to remove the old snapshot from the primary pool to free up space.

This was created for ZFS on Linux; BSD users may have to replace some of the tools.

Requirements

• ZFS
• pv
• numfmt

License

MIT License

Authors

James Badger ([email protected])

zfs-backup.sh

#!/bin/bash
set -e

base="data/postgresql"
receive_pool="work"
now=$(date "+%Y-%m-%d")

zfs snapshot "$base@$now"

snapshots=$(zfs list -t snapshot -s name -H | grep "$base" | awk '{ print $1 }')

previous=""
current="$base@$now"
for snapshot in $snapshots; do
        if [[ "$snapshot" == "$current" ]]; then
                break
        else
                previous=$snapshot
        fi
done

if [[ "$previous" == "" ]]; then
        echo "Previous snapshot not found."
        exit 1
fi

size=$(zfs send -nvDLepP -i "$previous" "$current" | awk '/^size/{ print $2 }')
human_size=$(numfmt --to=iec-i --suffix=B $size)
echo "Backup is estimated to be $human_size"

zfs send -DLep -i "$previous" "$current" | pv -s ${size} | zfs receive -F -d $receive_pool

echo "Run this yourself:"
echo "zfs destroy $previous"