VolumeSnapshot support?

[rbren]

Is it possible to support VolumeSnapshots, e.g. by making a copy of the directory?

[nvtkaszpir]

VolumeSnapshots are bit more complex than just simple file copying.

If you just look for some backup process then local-path-provisioner is just using a directory/path in container - so you can copy fie but it is advised to stop container before file copying to avoid file corruptions, especially if the app stores extra data in memory (such as databases).
Something like this:

• stop app container which uses that pvc
• spawn new container and mount that pvc as read only, copy files elswhere, stop the container
• start app container

Second step could be also done from host side.

What do you do in the backup process depends on the use case, this can be something like rsync, copy, rclone....

[rbren]

Yeah I understand that implementing the VolumeSnapshot API would involve a lot of work.

I'm specifically interested in being able to run tests that use the VolumeSnapshot API. Ideally I'd be able to do this with a KIND cluster and local-path-storage.

[nvtkaszpir]

You could try to run different provisioner one provisoner on top of another, if supported :D

[haslersn]

Could this be supported if the backing filesystem on the host is btrfs?

[rbren]

Have learned a bit more since opening the issue.

In particular, stopping/starting the underlying pods is not part of the VolumeSnapshot API. The user is expected to handle all of that, esp if e.g. they're running a database. The VolumeSnapshot API simply tells the storage provider to make a backup of the volume in its current state.

So I think adding VolumeSnapshot support may indeed be as easy as creating a tarball or copy of the directory (and then unpacking that to create a PVC from a VS).

[haslersn]

adding VolumeSnapshot support may indeed be as easy as creating a tarball or copy of the directory

Creating a tarball or copy is not atomic and therefore is not a "snapshot". Instead, a file system with snapshot support needs to be used.

[github-actions]

This issue is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 5 days.

[mlazowik]

Not stale

[github-actions]

This issue is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 5 days.

[github-actions]

This issue was closed because it has been stalled for 5 days with no activity.

[mlazowik]

Not stale

EDIT: looks like closed by bot are not reopened by bot