Windows allows to access previous versions of files (snapshots) in the property dialog of the mounted drive. If find this a very useful and intuitive way to access snapshots.
Samba and BTRFS supports this as well. You can find some short guide here:
1 Like
Thanks for bringing this to light. Samba’s Jeremy Allison was kind enough to alert us about this himself a while ago. We just din’t get to it yet(excuses, i know).
Here’s the issue I just created. Hope to get to it soon.
Perfect. Thanks. I’ll test it as soon as it is ready.
Ping - this is something that would be very welcome by me as well.
Since the samba integration is based on snapper, another nice feature made possible by this would be snapper’s “time based” snapshots (I do not know if there’s a better name).
For example, snapshot is taken every hour, and last ten hourly, daily, monthly and yearly snapshots are saved (e.g. they are automatically pruned so that the hourly schedule does not create 365*24 backups a year).
1 Like
the vfs line(s) to add to support snapshots under smb are easy to configure, but indeed a timeline feature like snapper offers would be great
another thing i currently want to test is manually building a timeline of snapshots using the prefix tag as %S for the naming format for shadow_copy2. Unfortunately the snapshot location has changed since i last checked with that vfs module. Now they are not inside the share but in the layer above (in my case the pool) and shadow_copy2 has a bug regarding shadow:snapdir which prevents it from being absolute (dont know if fixed now, but returns “realpath does not return an absolute path” or something similar when set to an absolute path) and using ../.snapshots/<share> doesnt work either.
in theory creating such sheduled tasks for snapshots every 30min, 1h, 1day etc with your specifc time to hold and the prefix 00-59 should work. I will test this when the vfs module is working again in some way.
@suman, maybe you can look into the samba vfs shadow_copy2 issue? it was definitely working when snapshots where put inside the datasetand not the pool/level above
i can provide further error logs and such if you need them and ofc i can test it on a dedicated machine