Track CSI VolumeGroupSnapshot implementation

[blackpiglet]

Describe the problem/challenge you have

VolumeGroupSnapshot is already added to CSI v1.8 spec, which will be released with k8s v1.27.
It will be an Alpha feature in v1.27, but because it's an out-of-tree implementation, it's will not be shipped with k8s v1.27 GA together.
It's still under development in the external-snapshotter repository:
https://github.com/kubernetes-csi/external-snapshotter/pulls?q=is%3Apr+is%3Aopen+%22Group+Snapshot+Controller%22

Describe the solution you'd like

Let's track its progress, and have a try in Velero when it's ready.

Anything else you would like to add:

Environment:

• Velero version (use velero version):
• Kubernetes version (use kubectl version):
• Kubernetes installer & version:
• Cloud provider or hardware configuration:
• OS (e.g. from /etc/os-release):

Vote on this issue!

This is an invitation to the Velero community to vote on issues, you can see the project's top voted issues listed here.
Use the "reaction smiley face" up to the right of this comment to vote.

• 👍 for "The project would be better with this feature added"
• 👎 for "This feature will not enhance the project in a meaningful way"

[reasonerjt]

Let's hold this till this is more mature in upstream.

[reasonerjt]

Given this is till an alpha in upstream, I don't think we have to deliver any concrete output, we'll just keep an eye on it and start to implement it when it reaches a more mature level.

[kaovilai]

Notes from @sseago

Basically the volumegroupsnapshot work is built on top of the phase 1 output (phase 2/parallel action isn't necessary/relevant). Once the ability to identify ItemBlocks is in place, on the backup side you'll need to modify the pod plugin to identify the items needed for the VolumeGroupSnapshot -- i.e. for the pod, list the PVCs, for each PVC, if it belongs to a volume group, list the other PVCs, for each of these other PVCs, identify the mounting pod (if any), all of these items belong to the ItemBlock and should be returned by the new GetAdditionalItems API call (as well as the normal Execute additional items list). That will make sure that hooks are handled properly for the associated pods. The rest of the backup work is probably on the pvc/csi plugin side. I'm guessing we'll want to modify the code which creates VolumeSnapshots and instead create a VolumeGroupSnapshot for volumes in a group (if it doesn't already exist, as created by the pvc plugin for another pvc in the group).

"ability to identify ItemBlocks" I believe is part of #7474

I would love to drive this feature for 1.15 release.

[sseago]

@kaovilai Note that my comments above track what we'll need to do on the backup side. On restore, I haven't really thought through it yet. We'll definitely need CSI plugin/datamover changes. In terms of restore workflow, we'll need to dig in and investigate. Since we'll need to do something similar with restore pod hooks, we may well need to modify RIA in the same way and then make similar ItemGroup workflow modifications on restore (and, as above, ignoring the question of multithreading, we need item blocks for making sure pod hooks are handled appropriately, not for parallelism).