zfsbud: Simple ZFS snapshotting, replicating & backup rotating convenience bash script

Want to skip this post and go straight to the repository? Bye.

No filesystem for cavemen

The features of a ZFS filesystem including pooled storage, copy-on-write, snapshots, data integrity tools, compression and encryption truly elevate your backup solution of choice. Especially after years of using EXT4/NTFS/APFS like a caveman.

For the purists stepping out of that cave, software like TrueNAS will pose too big of an overhead, so they will hack together solutions specifically tailored to their needs.

zfsbud

(I realize it's corny, but you got to name that github repository somehow...)

As I recently went through that phase, let me save you some time by introducing a bash script I wrote that caters to all my ZFS needs. Its aim is to simplify common ZFS tasks by figuring out the boring stuff, while being verbose about the process and potential problems.

The features include:

• Creation of a snapshot of one or multiple datasets
• Rotation (selective deletion) of older snapshots of one or multiple datasets while keeping 8 daily, 5 weekly, 13 monthly and 6 yearly snapshots
• Replication of one or multiple datasets to a local or remote location through ZFS send
• Smart/automatic handling of initial & consecutive send operations
• Works with encrypted datasets
• Dry run mode for output testing without actual changes being made
• Optional logging

A practical example of usage

If you know your way around ZFS, go ahead and look at the code and code examples in this repo: github.com/gbytedev/zfsbud. Otherwise stick around and check out this simple example:

Example setup

Local machine

• local_pool
  • dataset1
  • dataset2

Remote machine

• remote_pool

Initial sending of two datasets to the remote

zfsbud.sh --send remote_pool --initial --rsh "ssh user@server -p22" local_pool/dataset1 local_pool/dataset2

Now the remote pool received both datasets along with all snapshots.

• If the destination is local, remove --rsh.
• If you wish to create a fresh snapshot before sending, add --create|-c.

Consecutive sending of a dataset to the remote

zfsbud.sh --send remote_pool --rsh "ssh user@server -p22" local_pool/dataset1

This command is basically the previous command without the --initial flag. The script will figure out the most recent common snapshot between the source and destination and will send all changes up to the newest source snapshot.

• If you wish to create a fresh snapshot before sending, add --create|-c.
• If you wish to rotate (selectively remove) old snapshots created with zfsbud, add --remove-old|-r.
• If the destination is local, remove --rsh.

Create a snapshot of two datasets, rotate old snapshots and backup to remote

zfsbud.sh -c -r -s remote_pool -e "ssh user@server -p22" local_pool/dataset1 local_pool/dataset2

This rotates snapshots by creating a new snapshot for each dataset and selectively removing old snapshots; the changes are then sent to the remote. This wold be a good command to put into a cron task running once a day.

• In addition to the newest snapshot, the most recent common snapshot plus 8 daily, 5 weekly, 13 monthly and 6 yearly snapshots are kept.
• Only snapshots made by this script are deleted; the snapshot prefix defining what is a zfsbud snapshot can be overridden with --snapshot-prefix|-p <cute_name_>.

Dry run & logging

--dry-run|-d provides calculated/assumed output information without making actual changes. This is highly recommended before a new utilization of this script.

--log|-l tells the script to log to the home directory, while --log-path|-L <path/to/file> allows defining the file path.

You can find the script in this repo: github.com/gbytedev/zfsbud. Let me know if you find it useful and how you would improve it.