Data-Engineer-Associate - Data Ingestion and Loading (21% of the exam) - Section 2.6

Prioritise between Auto Loader, Lakeflow Connect, partner connectors, and other ingestion methods based on volume, frequency, data types, and governance needs.

Given a scenario's data volume, arrival frequency, source system and governance requirement, choose between Auto Loader, COPY INTO, Lakeflow Connect standard or managed connectors, Partner Connect tools and custom notebook ingestion. Justify why the alternatives are weaker for that scenario.

Auto Loader versus COPY INTOLakeflow ConnectPartner Connectingestion frequencygovernance requirements

Practice question for this objective

Free sampleData Ingestion and Loadinghard

A data team must bring records from a niche marketing SaaS application into Unity Catalog tables on a daily schedule. Lakeflow Connect offers no managed connector for that application, the vendor exposes only a paginated REST API, and the team has no capacity to build and maintain authentication, pagination and incremental change handling itself. The landed tables must still be governed and audited in Unity Catalog. Which approach fits the requirement?

  • ARun an Auto Loader stream configured with a schema location that reads directly from the vendor REST API endpoint and appends the responses to a Unity Catalog managed table.
  • BWrite a COPY INTO statement whose source location is the vendor REST API endpoint, and schedule it hourly as a Lakeflow Jobs task against a Unity Catalog managed table.
  • CBuild a bespoke Python task in Lakeflow Jobs that calls the vendor API, handles paging and watermarks in its own state table, and writes the output with a Delta append.
  • DUse Partner Connect to set up a validated ingestion partner tool that pulls the application through its own maintained connector and lands the results in Unity Catalog tables. Correct
Choose Partner Connect when a governed source has no managed Lakeflow Connect connector and the team cannot maintain bespoke API extraction. Auto Loader and COPY INTO both ingest from cloud object storage paths, so neither can reach an application that is only available over a REST API. Lakeflow Connect covers the SaaS and database sources it ships connectors for. When a source falls outside that set, Partner Connect provides validated ingestion partners whose connectors are maintained by the vendor, while the resulting tables are still registered, governed and audited in Unity Catalog.

Why A is wrong: Tempting because Auto Loader is the default incremental ingestion tool on Databricks, but it reads files from cloud object storage only, so it cannot call a REST API and has no way to page through one.

Why B is wrong: Tempting because COPY INTO is idempotent and simple to schedule, but its source must be a cloud object storage path or an external location, and it has no mechanism for calling a web service.

Why C is wrong: Tempting because it would technically work, but it puts the exact authentication, pagination and incremental logic the team said it cannot maintain back into code they own.

Why D is correct: Correct, because Partner Connect exists for exactly the sources Databricks does not ship a managed connector for, and the partner maintains the API handling while the landed tables remain governed by Unity Catalog.

See more Data-Engineer-Associate practice questions, answers explained.

Exam traps in Data Ingestion and Loading

Answers that look right on this material and are not. Each one is a distractor from a different question in the Data-Engineer-Associate bank for this domain.

  • Auto Loader in directory listing mode for the clickstream prefix, and a second Auto Loader stream reading hourly Salesforce extracts that a bespoke task writes to storage first.

    Why it is wrong: Tempting because one tool for both sources looks consistent, but directory listing degrades as the prefix accumulates files, and the extract task is the Salesforce API code the team refuses to own.

  • It deletes each source file once the corresponding rows have been committed, so the directory never grows large enough to slow discovery down.

    Why it is wrong: Tempting because a shrinking directory would indeed keep discovery cheap, but Auto Loader never deletes source files, and cleanup remains the responsibility of the storage owner.

  • Load the SaaS objects with COPY INTO run hourly, pointing the statement at the application endpoint so the team keeps a single SQL based ingestion pattern.

    Why it is wrong: Tempting because one pattern for everything is attractive, but COPY INTO reads files from a storage path and cannot call a software as a service application endpoint at all.

Examworthy is not affiliated with or endorsed by Databricks. Original, blueprint-aligned practice material only.