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.

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.

Related Terms