Legacy ETL to Azure Modernization Playbook

Tell us what you're building

We use these answers to surface the prompts, skills, and MCP configs that fit your stack — and to substitute stack values like {{database}} into the prompts you copy. Content (PRDs, code, etc.) stays in your repo. Everything you enter here is stored in your browser — nothing is sent to a server.

Steps you'll go through

1. 01

   Pipeline Inventory and Audit

   6-10 weeks

   Inventory every legacy mapping / workflow / package — active vs dead, who consumes it, complexity, criticality, owner. Pipeline counts always come in higher than expected.

2. 02

   Strategy Decision (per Pipeline Category)

   2-3 weeks

   Not every pipeline needs the same approach. Categorize each into rewrite, tool-assisted translation, redesign, retire, or replace-with-platform-feature.

3. 03

   Target Architecture and Framework

   4-6 weeks

   Stand up the Azure-side architecture and the metadata-driven pipeline framework before migrating any specific pipeline. Reuses Phases 2-5 of the Azure Data Platform Modernization playbook.

4. 04

   Behavior Parity Test Framework

   4-6 weeks

   Lock down legacy pipeline behavior as ground truth — including the quirks (rounding, null handling, date / sort / aggregation behavior). The new pipelines must match these to keep trust.

5. 05

   Pilot Migration (1-3 Representative Pipelines)

   6-10 weeks

   Validate the framework end-to-end on real pipelines: one simple, one medium, one complex. The pilot is where you discover what the framework didn't anticipate.

6. 06

   Mass Migration

   1-3 quarters

   Migrate the bulk of pipelines in batches using the established patterns. Common batching: by source system, by consumer, by complexity, or by owner.

7. 07

   Cutover and Decommission Legacy ETL

   4-8 weeks

   Disable legacy pipelines, verify zero traffic for 30+ days, decommission infrastructure, cancel licenses, archive legacy code for compliance and reference.

8. 08

   Optimize and Stabilize

   Ongoing — especially the first 6 months post-migration

   Tune performance, rationalize cost, build operational maturity, refine patterns. The migration is over — the platform owns the work now.