Naming scheme peculiarities

[rrrnld]

I have zfs snapshots on multiple datasets that I want sanoid to consider as its own. My understanding of #800 (comment) is that it should be enough to rename those to match sanoid's scheme. So I rename snapshots with the following naming scheme:

data1/sftp/staging@post-backup--2023-05-15T00:00:12Z

to sanoid's scheme like so:

data1/sftp/staging@autosnap_2023-05-15_00:00:12_daily

My sanoid config looks like this:

[template_staging]
        frequently = 0
        hourly = 0
        daily = 30
        monthly = 12
        yearly = 5
        autosnap = yes
        autoprune = yes

[data1]
	recursive = zfs
	use_template = staging

I rename the snapshot first, then run sanoid. It still creates a new daily snapshot. This is what it looks like for the last two days:

data1/sftp/staging@autosnap_2023-05-13_00:00:13_daily
data1/sftp/staging@autosnap_2023-05-13_21:15:10_daily
data1/sftp/staging@autosnap_2023-05-14_00:00:11_daily
data1/sftp/staging@autosnap_2023-05-14_21:15:23_daily

Is there a misunderstanding on my side? I'd like to create monthly and yearly snapshots if necessary, but consider the daily snapshots that were taken before and prune them according to the rules given in the config file. Is this possible? Thanks!

[ovizii]

Did you ever figure this out? oh, and btw. how did you rename the snapshots? Is there a way to mass-rename somehow?

[rrrnld]

I wrote a custom script that lists snapshots via zfs list -r -t snapshot $DATASET | grep pattern and renames them one by one

[phreaker0]

@heyarne the problem is that you used "recursive = zfs" instead of "recursive = yes".

With "recursive = zfs" zfs handle the recursion (which is atomic), so sanoid will only look for snaphosts on the root dataset of the recursion (which is data1 in your case) for deciding if a snapshot needs to be taken.

[ovizii]

Is that documented somewhere? I can't seem to find any info about this fact.

[phreaker0]

only in the configuration file

sanoid/sanoid.conf

Line 31 in a7e6c2d

# you can also handle datasets recursively in an atomic way without the possibility to override settings for child datasets.

[rrrnld]

Am I at risk of losing snapshots if I change the setting to recursive = yes?

[phreaker0]

No, pruning is done the same way with recursive = zfs and recursrive = yes, according to your configured prune policy.