DP-750 Certification Practice Question #60

Question

Your team loads several Delta tables in Unity Catalog. For each load scenario below, you must choose the single most appropriate write strategy. The available tiles are Delta/Spark write operations. DRAG-AND-DROP — Match each load scenario (slot) to the correct operation (tile). Each tile is used exactly once. ```mermaid flowchart LR subgraph Tiles["Operation Tiles"] T1["MERGE INTO ... WHEN MATCHED UPDATE / WHEN NOT MATCHED INSERT"] T2["MERGE INTO ... WHEN NOT MATCHED THEN INSERT (insert-only)"] T3["INSERT INTO (append)"] T4["INSERT OVERWRITE"] end subgraph Slots["Load Scenarios"] S1["S1: CDC feed of customer changes —
update existing rows by key, insert new ones (upsert)"] S2["S2: Append-only event logs that may
contain duplicate uniqueId values to skip"] S3["S3: Bronze raw ingest — always add
new immutable records, never update"] S4["S4: Fully replace a small static
lookup table on every run"] end S1 -.-> T1 S2 -.-> T2 S3 -.-> T3 S4 -.-> T4 ```

Accepted Answer

A CDC feed that updates matched keys and inserts new keys is the canonical `MERGE INTO ... WHEN MATCHED THEN UPDATE ... WHEN NOT MATCHED THEN INSERT` upsert (S1->T1). Append-only logs that may contain duplicates are deduplicated at write time with an **insert-only** merge keyed on `uniqueId` using only `WHEN NOT MATCHED THEN INSERT` (S2->T2). Bronze raw ingest that never updates simply appends new immutable rows with `INSERT INTO` (S3->T3). A small static lookup that must be fully replaced each run uses `INSERT OVERWRITE`, which atomically replaces the table contents (S4->T4).

More DP-750 practice questions