Free Databricks DEA practice questions

The Databricks Data Intelligence Platform is built on the lakehouse architecture.

A lakehouse combines the low-cost, open storage of a data lake with the reliability and performance of a warehouse. Data lives once, in open Delta Lake tables on cloud object storage; Delta adds ACID transactions and row-level updates; Unity Catalog governs access to it for every workload. The warehouse-plus-lake design is exactly the two-copy, two-governance setup the team wants to leave, and plain Parquet in a lake lacks transactional updates.

1. Note the two problems: duplicated copies and two governance points.
2. Look for the option that keeps one copy of data for BI, SQL and ML.
3. Check it also provides reliable (transactional) updates: Delta Lake does, raw Parquet does not.
4. Pick the lakehouse with Unity Catalog as the single governance layer.

Exam tip: Lakehouse = one copy in open Delta tables on object storage + warehouse-grade reliability + one governance layer (Unity Catalog).

Lakeflow Connect managed connectors are the most automated way to ingest popular SaaS applications into Unity Catalog.

Salesforce has a Lakeflow Connect managed connector. It is built on Lakeflow Spark Declarative Pipelines and adds source-specific authentication, incremental reads, automated retries and schema evolution, so there is no ingestion code to own. A REST notebook meets the schedule but puts API maintenance on the team. An export-plus-Auto Loader design needs a separate export job. A partner tool works but brings a second vendor, which the stem excludes.

1. Identify the source type: a popular SaaS application.
2. Check whether a managed connector exists for it; Salesforce has one.
3. Apply the constraints: no code to maintain and no new vendor.
4. Choose the managed connector with a daily schedule.

Exam tip: Popular SaaS source + no code to maintain = Lakeflow Connect managed connector.

dropna and fillna both accept a subset of columns, and the order of cleaning steps changes the result.

The requirement treats each column differently, so each step must be scoped. dropna(subset=["customer_id"]) removes only rows missing the identifier, and fillna({"country": "UNKNOWN"}) touches only country. Including phone in the dropna subset, with or without thresh, loses rows that are only missing a phone; filling every string column first hides the null identifiers from the later dropna.

1. List what should happen to each column: drop, keep or fill.
2. Scope dropna to the column whose null makes a row unusable.
3. Scope fillna to the column that gets a default, using a dictionary.
4. Check that no step runs before another and hides the nulls it needs.

Exam tip: Scope dropna and fillna with subset or a dictionary; unscoped calls hit every column.

Each task in a Lakeflow job can have its own retry policy: how many times to retry and how long to wait between attempts.

A transient failure that a rerun fixes is exactly what task retries are for. The policy is set on the task, needs no code change, and reruns only that task after the configured wait. Run if settings control whether downstream tasks start, duration thresholds only notify, and concurrency settings govern overlapping runs rather than recovery.

1. Spot the symptom: a transient error that a later rerun fixes.
2. Look for a setting that reruns the failed task automatically.
3. Recall that retries and the wait between them are configured per task.
4. Reject settings that only notify, change downstream behaviour or allow overlap.

Exam tip: Transient task failure: add a task retry policy with a wait between attempts.

Git folders let you create a branch from a chosen base branch and switch to it from the Git dialog.

Isolating work until review is the job of a feature branch. In the Git dialog you choose Create Branch, name it and pick main as the base, and the folder switches to the new branch. Committing to main, even from a second clone, bypasses review, and copies outside Git cannot be pushed or reviewed at all.

1. Identify the requirement: keep work off main until review.
2. Recall that a branch created from main isolates the changes.
3. Create it from the Git dialog in the Git folder.
4. Reject options that commit to main or leave version control.

Exam tip: New work in a Git folder starts on a new branch created from the base branch.