How-to: Roll back artifacts safely¶
This guide shows how to how-to: Roll back artifacts safely in a reliable, repeatable way.
Who this is for¶
- SRE / on-call responding to a bad publish or stale/incorrect artifacts
- Data engineers who need a reversible release process
Goal¶
Roll back the “current” manifest pointer to a known-good version with a clear verification step.
Quick paths¶
- Roll back the manifest: How-to: Roll back to a previous artifact version
- Artifacts and versioning (concepts): Artifacts and versioning
- Output layout (where pointers live): Output layout (local filesystem)
- Stale artifacts runbook: Runbook: Stale artifacts
Safety checklist¶
-
Confirm you are rolling back the right (tenant, surface, segment)
-
Make the scope explicit in your command and verification steps.
-
Roll back the manifest pointer
-
Use the canonical flow: How-to: Roll back to a previous artifact version
-
Verify the rollback
-
Confirm “current” points to the expected version: Output layout (local filesystem)
Read next¶
- Artifacts and manifest lifecycle: Artifacts and manifest lifecycle
- Runbook: stale manifest: Runbook: Stale manifest
- Pipelines operations index: Operations (recsys-pipelines)