Guides ยท Business
Data Pipeline Rollback Basics
Undo bad data runs
Rolling back data pipelines involves checkpoints and idempotent/reversible writes, tagging bad outputs, stopping downstream jobs, and running backfills with validated inputs to restore correct data.
- data pipeline
- rollback
- backfill
- checkpoints
- reversible
Detect and Stop
Tag bad runs; halt downstream jobs to limit spread.
Revert
Use checkpoints/versions to delete or overwrite bad outputs.
Backfill
Re-run with correct inputs; validate before unfreezing downstream.