Skip to main content

Backup and Recovery

The point of backup is not to say “we have backups.” The point is to restore a tenant, workspace, pack, service, or platform environment without guessing.

Worka has several kinds of state, and not all of them matter equally in a restore.

What you need to protect​

Protect at least:

  • canonical platform state
  • workspace and workflow state
  • pack and release metadata
  • service metadata
  • secret records and the ability to recover them safely
  • persistent service data
  • artifact metadata and, where required, artifact storage itself

Do not assume you can rebuild everything from source. Some of what users care about is execution history, approvals, attachments, and workspace state.

Plan recovery by outcome​

Think in terms of the recovery outcomes you need:

  • restore the whole platform
  • restore one tenant
  • restore one workspace
  • restore one service and its data
  • restore after accidental revocation or deletion

Those scenarios do not all use the same runbook.

Test what you back up​

The only useful backup is one you have restored successfully.

Practise:

  • restoring into a test environment
  • bringing a workspace back to an operable state
  • recovering a service and verifying its route and health
  • restoring secrets or secret references safely
  • checking that approvals, activity, and release lineage still make sense after restore

A restore that technically succeeds but loses the ability to explain what happened is still a poor restore.