Databricks ML Professional Practice Question 29: Git Sourced Job
Databricks Certified Machine Learning Professional · Test 3 · question 9 of 10
A production retraining job runs a notebook stored in an engineer's personal workspace folder. Last week someone edited the notebook while a run was in progress, and the team cannot now determine which version of the code produced the model that run registered.
What should they change?
Show answer and explanation
Answer:
Databricks recommends that staging and production jobs reference a remote Git repository rather than workspace files. When a run begins, the job takes a snapshot of the specified branch, tag or commit, and every task in that run uses that one version, so an edit made mid-run cannot affect a run already underway. The run history then records the commit SHA for each run, which answers the team's second problem directly: any model can be traced to the exact code that produced it.
Option A narrows who can cause the problem without removing it. Two people can still edit during a run, a shared folder has no versioning, and there is still no commit recorded against the run.
Option C swaps one unversioned location for another. Reading a notebook from storage gives no branch or commit reference and no snapshot guarantee, and it drops the code out of code review entirely.
Option D depends on a manual step that is precisely what gets forgotten under pressure, and a hand-written version string is not evidence of what actually ran.
Reference: Use Git with Lakeflow Jobs