G’day guys,
I made another video explaining how I setup automatic Tiered Snapshots and how to use them for data recovery on Windows, macOS, the shell and in your VMs
Enjoy
7 Likes
Don’t forget to update your YouTube page. (It’s this thread 
)
The TrueNAS Forum Thread for this video:
TBA
You just did it in time it took me to tease you, Gosh Darn It.
1 Like
Ye Olde Chicken and Egg 
I rate it a 9/10 on the video.
I cannot give it a perfect 10, since he referred to “zee vols” as “zed vols”. Simple mistake, but accuracy is key.
2 Likes
But Zed Eff Ess is okay
1 Like
@Stux Thanks for a great tutorial!
I’m struggling to work out how I can leverage this to do off-site backups with history. I’ve Googled backing up TrueNAS to a share and everyone says “just use rsync.” In my understanding rsync only gets you one copy of the data not backup history.
To use the snapshots feature and be able to restore a complete local system failure, I would need a starting point from where all of the snapshots are based. I could just do an rsync of the full data that I want to backup then rsync the .zfs folders. However, this falls into the ever growing snapshot pitfall.
I then thought about having a job that re-rsynced the full data set every so often and killed old snapshots. However there is a possibility of data changing during the full rsync and then losing that change.
Are there any other proven ways to do this?
I was hopeful that TrueCloud would be capable to backing up to a remote share, but that is not an option as far as I could tell. (I don’t want to pay for cloud storage OR have my data hosted somewhere.)
The TrueNAS documentation talks about using replication for backups but I have not been able to find a lot of information on this.
Any help is appreciated!
The trick is to replicate to another TrueNAS.
Perhaps offsite.
In a personal scenario perhaps at a friend or relatives.
1 Like
Does it have to be a TrueNAS system?
I currently have a Synology that I want to move to be my offsite backup and use TrueNAS as my primary. (I am currently lacking offsite backup but I don’t waht to have to purchase 2 systems.)
To support and leverage ZFS atomic snapshots (and incremental replication streams), both sides need to be ZFS.
Thanks for the video, @Stux.
I’m about to set the snapshots schedule you proposed; however, I have one concern. This schedule definitely fits general-purpose file share. But what about a workstation’s backup share?
Let’s assume that a backup of a workstation takes about an hour. During that hour several 10-minutes snapshots would be made (because of dataset changes). But recovering from these snapshots doesn’t make much sense. Because they will contain incomplete network-disconnect-like backups.
I’m not sure about windows, but macOS incomplete backups can screw up the entire backup history (I’m looking at you, <my-hostname>.purgeable). So, shouldn’t snapshots of backup shares have been taken less often? Like just hourly or even daily?
TrueNAS creates separate snapshots for MacOS timemachine backups when they complete.
These are maintained automatically, and allow you to restore the time machine backup itself if it becomes corrupted.
I normally don’t backup my backups…
Meanwhile, Timemachine offers its own “snapshot” feature, which is why I exclude Time Machine from my snapshot schedule.
Wow, sounds like a nice feature. However:
- I now see only your tiered snapshots of my TimeMachine share. And there were definitely successful backups before I set up the schedule. Perhaps it should be manually set up in the current versions (I’m using the “Basic time machine share” preset).
- I found users were complaining of this feature not preserving the retention policy. Also, they complained about not being able to see this feature in the GUI and the lack of documentation for the feature and share aux params. That was back in 2021. I didn’t find a GUI option or documentation now.
- I found a bug report that prevents this feature from taking snapshots from new version of macOS backups. That was back in 2023. While the bug was fixed long ago, I consider that these auto-snapshots can be unreliable.
Summing up, I’ll stick to your tiered snapshots for ̶g̶o̶o̶d̶ now.
1 Like
Hi Stux,
Thank you for this great video. I’m actually re-doing my snapshots because of this video.
I had one question. In your video, you first started off by creating a manual snapshot of your root pool (i think?). But I have a root dataset that has all my child datasets (and nested child datasets).
Should I take a manual snapshot of my root dataset instead (where you did the root pool/I guess in your case it’s kind of both the root pool and the root dataset?). Is there advantage (or disadvantage) to taking manual snapshot of the storage pool instead of my root dataset?
Or is that first manual snapshot in your video not required for the automated snapshots? It was more of just an example?
Thank you!
It was an example
Thank you very much, Stux!
I have another question that’s kind of off topic:
Do you know if shadow copies are buggy and unreliable when using nested datasets? (in windows 11 environment for clients)
I’m getting a lot of weird issues where sometimes a shadow copy just doesn’t get created on it’s schedule, sometimes there’s shadow copies then later they disappear, lots of times the shadow copy doesn’t show all the contents of the folder and then sometimes it will (most of time it’s missing items - all folders have the exact same permissions), most of time the shadow copy will get created with only what was added but it won’t show anything else that exists in the folder. Sometimes it will show a shadow copy and you try to open it and you get the message "The previous version of “XXX” from Today, June 29 2025, 3:00 PM no longer exists. Select a different version and try again (even though I just opened it a few minutes ago - restarting smb service might be the cause of this for some reason). Restarting smb service may or may not help in some situations. I’ve been testing it for many hours with no luck…
I started searching it online and see other people complain about shadow copies and nested datasets but they seem to be from a year+ ago.
I’m testing just using a single dataset and shadow copies seems to work perfectly so far. Even when nesting folders.
If that’s the case (and it’s an on-going issue), I may have to rebuild my truenas structure to just use single datasets (no nested datasets) and then create the directory structure within windows.
Thanks! Just really want to know if you know it to be glitchy for nested datasets.
Well, using nested datasets is not recommended inside SMB shares, as such, I haven’t tested it, and I would not be surprised if it were unreliable.
Roger that. Thanks again, you saved me from a lot of pain in the future! Be well please!