ZFS Snapshot Management

With my new fileserver/NAS, I am using a ZFS raid array for my file storage. One aspect of the ZFS file system that I didn’t really investigate all that closely before I installed it was the snapshot feature. ZFS snapshots basically store a “picture” or “image” of all the files on the system at a specific point in time. The logic behind snapshots is that you can then roll the filesystem back to that point in time, in case there was a problem.

This is, of course, a great feature for a variety of reasons. If you accidentally deleted a file, you could recover it this way. If someone hacked your server and encrypted your files, you could recover them this way (assuming the snapshots aren’t also encrypted).

However, there is a small issue here. In order for the snapshots to work, they basically have to store all of the files that were in existence when the snapshot was made. In other words, so long as you retain a snapshot, you will also have to retain all of the files associated with that snapshot, which will take up all the corresponding space as well. Technically, once you delete a file, it won’t show up in your file directory, but the file system has to retain a copy of it as part of the snapshot.

It was this issue that finally got me to look more closely at snapshot management. My fileserver has four 4TB hard drives in it, which gives me roughly 8TB of storage. I can always upgrade the drives to bigger drives if needed, but I don’t have 8TB of files at this point. I have around 4TB, but my fileserver was reporting less and less available space, even after I deleted some files I didn’t want/need. I was confused until I realized that I had set up some automatic snapshots when I first set up the fileserver. Those snapshots included many of the files I had deleted, so no space was freed up when I deleted those files. I had to delete the snapshots in order to free up the space. It took some doing, but I finally figured out how to manage snapshots a bit better.

First, here’s how to list your existing snapshots:

zfs list -t snapshot -r [ZFSpoolnamehere]

If you have any snapshots, you should see a list like this:

To create a snapshot, you can use this code:

sudo zfs snapshot -r [ZFSpoolnamehere]@[nameofsnapshot]

The above command puts you in root (sudo) and calls zfs, telling it to create a snapshot. The “-r” part of the command tells zfs to make a recursive snapshot, which means it will create snapshots of each of your top directories in your ZFS pool. You then have to invoke the ZFS pool followed by the “@” symbol and then the name of your snapshot. Here’s my actual command:

sudo zfs snapshot -r ZFSNAS@2021_01_02

This command creates a snapshot of my entire ZFS pool and names it with the date it was created – January 2nd, 2021.

Once I’ve created my snapshot, I can see it in the list of snapshots I have created with the command above.

Of course, there is also the issue of deleting snapshots. This was the issue I needed to address that got me started on ZFS snapshots – I had some very old snapshots that were taking up terabytes of space on my fileserver. To delete a snapshot, you can use the following code:

sudo zfs destroy [ZFSpoolname]/[directoryname]@[nameofsnapshot]

Here’s an example of my actual code used to delete an old snapshot:

sudo zfs destroy ZFSNAS/files@2020_12_25

This command tells zfs to delete my snapshot in the ZFSNAS pool that was made of the “files” directory on December 25, 2020 (@2020_12_25). When I check the list of snapshots after deleting that one, it’s no longer there and the space that was reserved for that snapshot is freed up.

Luckily, I haven’t needed to roll back to a snapshot at this point. However, it is nice knowing that my filesystem on my NAS has that capability.

For more information on working with ZFS snapshots, see here.

Finally, there is the issue of how many snapshots to keep and when to make them. Given my use case – a fileserver/NAS that stores my music, my video collection, my old photos (not the current year’s photos), and my old files, I actually don’t need particularly frequent snapshots. I decided that one snapshot per month would be sufficient. You can automate this by putting it into a crontab job. However, I simply added this to my calendar to create a snapshot once a month. That way, I can also delete the corresponding snapshot from 12 months prior. Basically, I am keeping one year’s worth of snapshots with one snapshot per month. This gives me a year’s worth of backups to which I can roll back should I need to but also regularly frees up any space that might be taken up by particularly old backups.

I can imagine that other use scenarios would require different frequencies of snapshots for a zfs system (e.g., a production server or someone who uses a zfs raid as their primary file storage). Since my server is primarily used for storing old files and relatively unchanging collections, once a month snapshots are sufficient.