← All modules

E5 — Implementing CI/CD

Why this module matters

10% of the Associate exam, and the section most likely to be underestimated. It's four or five questions on a 45-question paper — but they're cheap points, because unlike ingestion or transformation there's no judgment call. The syntax is small, the workflow is fixed, and the terminology changed recently enough that a lot of study material is now wrong.

That last part matters more than the exam. Every organization eventually discovers that the job running production was configured by someone who left, in a UI, with no record of what they clicked. CI/CD on Databricks is the answer to that, and Databricks has been unusually direct about what the answer is.

The one idea to hold onto

"Declarative Automation Bundles are the recommended approach to CI/CD on Databricks." — CI/CD on Databricks (updated 2026-07-24)

A bundle is a directory with a databricks.yml in it. That one file describes your jobs, your pipelines, your dashboards, your clusters, and the environments you deploy them to. The Databricks CLI reads it and makes the workspace match.

   YOUR REPO                          CI RUNNER                    WORKSPACES
  ┌────────────────┐              ┌────────────────┐            ┌──────────┐
  │ databricks.yml │              │ bundle validate│  ──dev──▶  │   dev    │
  │ resources/*.yml│  ──git──▶    │ bundle deploy  │            ├──────────┤
  │ src/           │              │ bundle run     │  ──prod─▶  │   prod   │
  │ tests/         │              └────────────────┘            └──────────┘
  └────────────────┘               one config, many targets

One codebase. Many targets. No clicking. Everything else in this module is detail.

⚠️ Two renames you must get right

Databricks renamed two things in this section, and both renames are on the exam guide itself:

The bundle rename was explicitly non-breaking:

"This name change is non-breaking. The bundle CLI command and all of your existing configuration does not need to be modified." — Bundles FAQs (2026-07-10)

So databricks bundle deploy is still databricks bundle deploy. Your databricks.yml is untouched. Only the words changed. The official May 2026 exam guide writes it as "Declarative Automation Bundles (formerly Databricks Asset Bundles)" and "Databricks Git Folders (formerly Databricks Repos)" — so the exam itself is doing the mapping for you.

Old material referring to "DABs" and "Repos" is still technically accurate about the mechanics. Use it, but translate as you read.

What the exam actually asks

The four objectives for Section 5, verbatim from the May 2026 exam guide:

1. Manage your code development workflow within the Databricks workspace UI, including creating and switching between branches in Databricks Git Folders (formerly Databricks Repos), committing and pushing changes, and creating pull requests using Databricks Git integration.
2. Understand environment-specific configuration using Automation Bundle (formerly Databricks Asset Bundles) variables and overrides while promoting the same codebase across dev, test, and prod targets.
3. Deploy Declarative Automation Bundles (formerly Databricks Asset Bundles) to package, configure, and promote Lakeflow Jobs, Lakeflow Spark Declarative Pipelines, and other workspace assets across dev, test, and prod environments.
4. Understand the Databricks CLI to validate, deploy, and manage Declarative Automation Bundles (formerly Databricks Asset Bundles) and other workspace assets in automated CI/CD workflows.

Notice what's not there: no Terraform, no Jenkins, no Airflow. Objectives 2 and 3 are both about variables, overrides, and targets — that's where to spend your study time.

What you'll be able to do

1. Explain what a bundle is, what it contains, and why "just click it in the UI" fails at scale.
2. Read and write a databricks.yml — every top-level key, resources, targets, variables, substitutions, presets.
3. Run the full CLI workflow: init, validate, plan, deploy, run, destroy, and the commands that migrate existing workspaces into bundles.
4. Explain exactly what mode: development and mode: production change, without looking it up.
5. Drive Git folders from the workspace UI, and say where they sit relative to bundles.
6. Write a GitHub Actions workflow that authenticates with OIDC — no stored secret — validates, deploys, and runs a bundle.

Lessons

Then: Cheat sheet · Lab · Quiz

A note on the docs

This module hit more live documentation inconsistencies than any other module in the course — seven of them, all flagged where they arise. Two are worth knowing before you start:

• The release notes call the engine-migration command databricks bundle migrate; the CLI reference and the migration guide both call it databricks bundle deployment migrate.
• Databricks' own GitHub Actions examples authenticate with a personal access token, while the best-practices page one click away says workload identity federation is "the most secure way to authenticate."

Neither is a disaster. But you will copy a doc example into a pipeline one day and watch it fail, and it's worth knowing in advance that the failure might not be yours.

There's also a live deadline in this section: the Terraform deployment engine "will be disabled in new releases of the Databricks CLI in September 2026" (What's coming, 2026-07-30). That is next month. Lesson 3 covers what to do about it.

Facts verified 2026-07-31 against the pages cited in each lesson.

▶ Play this module · 51 min

Keeps playing into the following modules — 304 min from here to the end of the course.