Episode Details
Back to Episodes
GitHub, Azure DevOps, or Fabric – Who’s Actually In Charge? Medallion Architecture, CI/CD & GitOps for Fabric Warehouse
Season 1
Published 7 months ago
Description
Here’s the uncomfortable truth: without CI/CD, your beautiful Medallion Architecture is just a very expensive CSV swamp wearing a Gold badge. In this episode, we start right where most Fabric teams are stuck—Bronze ingestion scripts living in notebooks, Silver transformations hacked in prod, and Gold dashboards patched at 3 a.m.—and show how GitHub or Azure DevOps becomes the actual control plane. You’ll see how treating notebooks, SQL scripts, and pipeline configs as code (versioned, reviewed, and promoted) turns Fabric Warehouse from “please don’t break” into something you can roll back, test, and move between dev, test, and prod without midnight firefights.
BRONZE WITHOUT ROLLBACK – YOUR CSV GRAVEYARD
Bronze is where the first goblins spawn: corrupted feeds, schema drift, duplicate loads—quietly poisoning everything upstream while pipelines still show a comforting green checkmark. We walk through how bad timestamps, header changes, and extra columns in “raw” zones become permanent damage when ingestion logic isn’t in Git and deployments go straight to production. You’ll learn three non‑negotiables for Bronze: keep every ingestion notebook/script in source control, parameterize connections and schemas instead of hard‑coding prod, and run pre‑deploy schema/dry‑run checks in CI so bad changes never hit your landing zone. With those guardrails, an ingestion failure becomes a quick rollback and redeploy—not a weeks‑long data‑rebuild panic.
SILVER: WHERE GOVERNANCE DIES QUIETLY
Silver looks clean on the surface—standardized types, deduped rows, pretty column names—but this is where governance often dies in silence. When fixes live in ad‑hoc notebooks and “just this once” patches against production, dev, test, and prod stop matching, and you only notice when numbers don’t reconcile in front of leadership. We show how to force discipline into Silver: every transformation change goes through a pull request, automated data‑quality checks (nulls, uniqueness, schema compatibility) run in CI, and promotions happen only via pipelines—not manual edits. That way, Silver becomes the first layer where “what happens in dev is exactly what happens in prod,” instead of three different realities with the same table names.
GOLD AT 3 A.M. – ANALYTICS UNDER PRESSURE
Gold is the only layer executives actually see—dashboards, KPIs, quarter‑end numbers—and it’s where shortcuts cost the most. One untested schema tweak in Silver or a hotfix in Gold can break trust instantly when financials or key metrics suddenly don’t match past reports. We talk about building mirrored environments (dev/test/prod) for Gold models and reports, wiring deployment pipelines from Git so only tested changes ship, and banning “panic SQL” edits in production warehouses. When GitHub or Azure DevOps becomes the single source of truth for Gold, 3 a.m. calls turn from “what changed?” into “which commit broke this?”—and that’s a problem you can actually solve.
WHAT YOU’LL LEARN
BRONZE WITHOUT ROLLBACK – YOUR CSV GRAVEYARD
Bronze is where the first goblins spawn: corrupted feeds, schema drift, duplicate loads—quietly poisoning everything upstream while pipelines still show a comforting green checkmark. We walk through how bad timestamps, header changes, and extra columns in “raw” zones become permanent damage when ingestion logic isn’t in Git and deployments go straight to production. You’ll learn three non‑negotiables for Bronze: keep every ingestion notebook/script in source control, parameterize connections and schemas instead of hard‑coding prod, and run pre‑deploy schema/dry‑run checks in CI so bad changes never hit your landing zone. With those guardrails, an ingestion failure becomes a quick rollback and redeploy—not a weeks‑long data‑rebuild panic.
SILVER: WHERE GOVERNANCE DIES QUIETLY
Silver looks clean on the surface—standardized types, deduped rows, pretty column names—but this is where governance often dies in silence. When fixes live in ad‑hoc notebooks and “just this once” patches against production, dev, test, and prod stop matching, and you only notice when numbers don’t reconcile in front of leadership. We show how to force discipline into Silver: every transformation change goes through a pull request, automated data‑quality checks (nulls, uniqueness, schema compatibility) run in CI, and promotions happen only via pipelines—not manual edits. That way, Silver becomes the first layer where “what happens in dev is exactly what happens in prod,” instead of three different realities with the same table names.
GOLD AT 3 A.M. – ANALYTICS UNDER PRESSURE
Gold is the only layer executives actually see—dashboards, KPIs, quarter‑end numbers—and it’s where shortcuts cost the most. One untested schema tweak in Silver or a hotfix in Gold can break trust instantly when financials or key metrics suddenly don’t match past reports. We talk about building mirrored environments (dev/test/prod) for Gold models and reports, wiring deployment pipelines from Git so only tested changes ship, and banning “panic SQL” edits in production warehouses. When GitHub or Azure DevOps becomes the single source of truth for Gold, 3 a.m. calls turn from “what changed?” into “which commit broke this?”—and that’s a problem you can actually solve.
WHAT YOU’LL LEARN
- Why a Medallion Architecture without CI/CD is just structured chaos in Fabric Warehouse.
- How Bronze ingestion turns into a CSV graveyard without Git, schema checks, and rollback paths.
Listen Now
Love PodBriefly?
If you like Podbriefly.com, please consider donating to support the ongoing development.
Support Us