Databricks Certified Data Engineer Associate Practice Questions: Delta Lake and Data Management Domain

A) Run OPTIMIZE on the table and use ZORDER BY device_id to compact small files and colocate data commonly used in filters

B) Run VACUUM with a very low retention period to delete as many old files as possible and speed up queries

C) Repartition the table by device_id to create one partition per device for maximum pruning

D) Disable ACID transactions on the table to reduce overhead from the Delta transaction log

A) Increase the VACUUM retention period to at least 13 months so that historical files remain available for time travel over the full compliance window.

B) Keep VACUUM retention at 7 days but rely on change data feed (CDF) to reconstruct older table versions beyond that period.

C) Disable VACUUM entirely so that no files are ever removed, guaranteeing time travel for all historical versions indefinitely.

D) Run VACUUM more frequently with the same 7-day retention, assuming that more frequent runs will preserve more history.

A) Bronze layer, because it stores the raw, unfiltered data exactly as received

B) Silver layer, because it is responsible for cleaned and validated data

C) Gold layer, because it is closest to business users and BI tools

D) No layer, because Delta Lake enforces all constraints automatically from the schema

A) Enable Change Data Feed (CDF) on the Delta table so downstream consumers can read row-level inserts, updates, and deletes incrementally without always relying on a full MERGE

B) Increase the frequency of the MERGE INTO job so that each run processes fewer rows and completes faster overall

C) Partition the table by transaction_id so that each MERGE INTO operation only touches a single partition

D) Disable ACID transactions on the table so that MERGE INTO operations can bypass the transaction log and run faster

A) CDF only captures changes from the point in time when it was enabled and is also subject to retention settings, so older changes may not be available.

B) CDF automatically backfills all historical changes for the table, but the job must use time travel to access them.

C) CDF is enabled by default for all Delta tables, so the failure must be due to a schema mismatch in the downstream job.

D) CDF stores all historical changes indefinitely, so the failure is likely caused by an unrelated networking issue.

A) Periodically run OPTIMIZE on the Delta tables to compact small files and, where beneficial, use ZORDER BY on commonly filtered columns.

B) Increase the cluster size for all queries so that more executors can read the many small files in parallel.

C) Partition the `logs_bronze` table by a high-cardinality field such as `session_id` to spread data across more partitions.

D) Disable streaming ingestion and switch to a single large daily batch write to avoid small files.

A) Keep the existing event_date partitions and Z-ORDER, but increase the partition granularity to event_timestamp to better match user_id filters.

B) Remove Z-ORDER and rely only on the existing event_date partitions, since partition pruning will handle user_id and session_id filters efficiently.

C) Convert the table to use Liquid Clustering on user_id and session_id and remove the static event_date partitions and Z-ORDER.

D) Drop all partitions and Z-ORDER settings and leave the table unpartitioned and uncluttered so the optimizer can fully control data layout.

A) Use Delta Sharing to grant the partner secure, governed access to the provider’s live Delta table so the partner can query it without copying data into their own storage.

B) Export the Delta table as Parquet files into a shared object storage bucket that both organizations can mount, ensuring the partner does not need to copy data again.

C) Schedule a daily job to copy the Delta table into the partner’s cloud account and rely on access controls there, since Delta Sharing requires data replication to the consumer.

D) Convert the Delta table to a UniForm table and let the partner read it as Iceberg from their environment, because UniForm automatically enforces centralized access control across organizations.

A) Ingest raw JSON directly into a single Gold Delta table and let analysts filter out bad records in their queries.

B) Use a multi-hop Bronze/Silver/Gold Delta architecture, enforcing constraints and transformations as data moves from raw to curated tables.

C) Store all data as Parquet files in a single folder and rely on an external catalog to track lineage.

D) Write streaming data directly into a Silver Delta table with schema enforcement disabled to avoid failures from malformed records.

A) The write will succeed because schema definitions do not enforce constraints at write time

B) The write will fail because the NOT NULL constraint is enforced at write time, unlike a simple nullable schema

C) The write will succeed but the null values will be automatically filled with defaults

D) The write will be partially committed, with only valid rows written and invalid rows silently dropped