Epic: PVC (Persistent Volume Claims) on Self-Hosted

[atduarte]

Summary

To ensure workspace data durability we are moving from backups to object storage to using persistent volume claims (PVCs) and snapshots. We have started with SaaS (#7901) and will then focus on self-hosted with this epic.

Value

Bringing the reduced chance of data loss, and associated system trust, to self-hosted customers.

Acceptance Criteria

• Self-hosted customers can benefit from the same durability guarantees offered on SaaS.
• There's a safe upgradability path.
• Installer/KOTS
  • installer: allow to specify storageClass in gitpod.yaml #10613
  • installer: add a test to ensure that CSI is working as expected prior to installing Gitpod #10614
• Tests
  • Manually test PVC CSI snapshot backup/restore in AWS #10211
  • Manually test PVC CSI snapshot backup/restore in Azure #10212

Persona(s)

Self-hosted admins and users.

Complexities

• Support of major clouds and on-prem setups will require extensive research and testing

[mrzarquon]

Cross Zone support will need to be tested. GKE has support for region wide persistence, but EKS's GP2/ebs type doesn't. EBS Snapshots are supported by the EBS-CSI driver, only standard in eks 1.23.

-snipped out of date info-

Updating with info on snapshot usage, they are region specific stored in a local S3 bucket.

Ensuring we can set variables such as encryption keys and data lifecycle periods for the snapshots would be essential so there's not long running PVCs that are tying stopped workspaces to a specific availability zone.

[jenting]

I think we don't need to work on this issue anymore. @atduarte

[kylos101]

Closing as it's for self-hosted.