Truenas Snapshots being automatically deleted unintentionally

TrueNAS General
,
1

Hello,
I seem to be having an issue where my snapshots are only maintained for an arbitrary amount of time (I think 1 week) even with a 1 month retention period. my only remaining guess as of now is that replication is somehow deleting them.

When the snapshots are deleted in the HDD mirror, it causes the next replication to fail due to not having an incremental base on dataset.

All of my daily snapshots on the mirror pool are retained for the correct amount of time, it’s just the weekly snapshots that are deleted, any snapshots transferred to the archive pool are maintained.

I have tried quite a few different changes/tweaks to the setup but it seems to be consistently deleting ever since I moved to my new HDD mirror method, I never had any issues when directly replicating from the SSD to the Archive drive, I do it this way because the Archive drives are loud and I want them to idle as much as possible.

My current setup is listed below (hopefully enough info, if not I’m happy to post more) hopefully I’m doing something stupid :slight_smile:, any help is greatly appreciated.

Thanks,
Josh

Truenas version:
Community 25.04.1 (but this has been an issue for 2 months, so multiple versions)

My current setup is as follows:
SSD Daily snapshot → (Weekly replication) → HDD mirror:

image
image326×350 9.19 KB

→ (Monthly replication) → HDD RaidZ1 archive
image
image278×346 8.67 KB

I’m going to use Veeam as the example, however, proxmox uses the same exact method just with a different final retention.

First step - Snapshot SSD:

image
image774×451 14.5 KB

Second step - Replicate to specific HDD mirror dataset:

image
image752×1573 53.5 KB

Third step - Recursive snapshot HDD mirror:

image
image755×469 15.3 KB

Fourth step - Replicate to specific Archive dataset:

image
image752×1619 50.8 KB

2

Having a play around with snapshots/replication, it appears to be my weekly replication of the daily snapshots to the HDD mirror that is erasing the weekly snapshots.

It erases the snapshot during the following:
Snapshot SSD 1 → Replicate to HDD → Snapshot HDD 2 → Snapshot SSD 3 → Replicate to HDD → HDD now no longer has the snapshot 2, but contains snapshot 1 and 3.

It’s almost as if it’s overwriting snapshot 2 because its not in the path of the the SSD’s snapshots, which seems weird because they’re both the same set of data, you’d expect snapshot 2 to still exist and be in between 1/3

3

I have reverted to doing the weekly snapshots on the SSD and replicated them over to the mirror and eventually the archive, this adds an extra snapshot task per dataset that I didn’t want to do, but it works.

I was having a look around (admittedly mostly fighting with ChatGPT to give me even a slightly sane answer) and it suggested that I do zfs send of the snapshots, would doing a ZFS send still align with TN’s pruning (I think that’s built in to ZFS rather than TN specific?),

I.E does it sound like a reasonable thing to do:
send all the daily snapshots to the correct location through a ZFS send script, recursive snapshot the mirror HDD weekly through TN and then replicate that to the final archive dataset.