Databricks-ML-Pro ML Ops Practice Question
A nightly Databricks job trains a model, registers a new version in Unity Catalog, and then updates a Model Serving endpoint that serves the 'champion' alias. The endpoint must switch to the new version only after the job's validation step passes. Which approach correctly enforces this?
⚠ Common exam trap
The trap here is thinking that registering a new model version automatically causes a serving endpoint to switch to it, when endpoints only change when their referenced alias or version changes.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Assign the 'champion' alias to the new version only inside the validation-passed branch of the job, and leave the endpoint configured to serve the alias.
Decoupling promotion from deployment is the key: the endpoint always serves the 'champion' alias, and the job reassigns that alias only after validation succeeds. This makes the alias the single source of truth for what is in production and keeps rollback to a one-step alias change, while the validation branch ensures unvalidated versions never reach the endpoint.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Configure the endpoint with a scale-out policy that adds replicas whenever a new model version is registered.
Why it's wrong here
Scale-out policies govern how many replicas serve traffic and respond to load, not which model version is served. Registering a new version does not trigger a version change on the endpoint, so this approach leaves the endpoint serving the old version regardless of validation outcome.
- ✗
Delete the previous model version after training so the endpoint is forced to pick up the newest registered version.
Why it's wrong here
Serving endpoints resolve versions through aliases or explicit version references, not by elimination. Deleting the prior version does not make the endpoint select the new one, and it destroys the rollback target, which is the opposite of what a safe promotion workflow requires.
- ✗
Have the job update the endpoint configuration to reference the new model version number directly after training completes.
Why it's wrong here
Updating the endpoint to a fixed version number bypasses the alias-based promotion flow and makes the endpoint insensitive to future alias changes. It also couples the endpoint to a specific version, so rollback requires editing the endpoint again rather than reassigning the alias.
- ✓
Assign the 'champion' alias to the new version only inside the validation-passed branch of the job, and leave the endpoint configured to serve the alias.
Why this is correct
Because the endpoint serves the alias, moving the alias to the validated version is the promotion step, and doing it only after validation passes guarantees the endpoint never serves an unvalidated version. Rollback is a single alias reassignment back to the prior version.
About these practice questions
One of 300 original Databricks-ML-Pro practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official Databricks exam blueprint
This Databricks-ML-Pro practice question is part of Courseiva's free Databricks certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the Databricks-ML-Pro exam.