[Epic] CSI snapshot integration

[skriss]

No description provided.

[skriss]

we've got 3 issues under this epic, but we need to break this down further. To do as a team once we finalize design.

[dns2utf8]

What is the status of this? Is it possible to backup the contents of a pvc with version 1.2.0?

[skriss]

@dns2utf8 in v1.2, you can back up non-CSI volumes (e.g. EBS, Google PD, Azure Disk) using snapshots, or any other type of volume using restic (https://velero.io/docs/v1.2.0/restic/).

This issue covers integrating with the new CSI snapshot API, to be able to take snapshots of CSI-provisioned volumes.

[dns2utf8]

Do I understand correctly that this means CSI volumes are not snapshoted as of today?
I would like to backup the PVs of a MariaDB and a PostgreSQL.

[skriss]

@dns2utf8 that is correct. As of today, Velero can snapshot non-CSI volumes, or use restic to back up CSI volumes.

[dns2utf8]

Thank you for your replies. I had some initial success with restic.
However, now the contents of the PV don't get into the s3 bucket anymore and the pods are stuck at the "Terminating" state.
I searched the web for a solution, is the MountPropagation still needed with the securityContext: privileged: true?
My test-setup is in this repo: https://gitlab.com/dns2utf8/multi_file_writer

[nrb]

Currently waiting on upstream providers to add v2.0.0 controller/v1beta1 CRD support to their drivers. I've logged an issue on the Amazon driver to start.

While we could proceed with the v1alpha1 code, I'd prefer to keep moving forward and help update drivers rather than move backward and end up updating a lot of code later.

[ashish-amarnath]

This is the issue with the GCP driver: kubernetes-sigs/gcp-compute-persistent-disk-csi-driver#457

[ashish-amarnath]

There are 3 more issues pending in this epic. https://github.com/vmware-tanzu/velero/labels/CSI
Which are not going to make it for 1.5.
Punting the remaining issues to 1.6.