Pass Databricks Certified Data Engineer Professional Exam in First Attempt Guaranteed!

Keep filter predicates aligned with the stored partition column. This design action addresses a different case (partition pruning), where a query wraps business_date in a transformation and reads many partitions. Direct predicates on partition keys can enable partition pruning. In the stated problem, file-level statistics are missing for the filter column; this choice would not establish the required data skipping statistics decision.

Choose supported liquid clustering on commonly filtered device fields. This design action addresses a different case (liquid clustering cardinality), where partitioning by device would produce many tiny partitions. Liquid clustering organizes data by useful filter keys without high-cardinality partition directories. In the stated problem, file-level statistics are missing for the filter column; this choice would not establish the required data skipping statistics decision.

Use predictive optimization on supported managed Delta tables. This design action addresses a different case (predictive optimization), where operators run blanket manual OPTIMIZE jobs regardless of table use. Predictive optimization can automate supported managed-table maintenance based on workload signals. In the stated problem, file-level statistics are missing for the filter column; this choice would not establish the required data skipping statistics decision.

Confirm authorized query succeeds while excluded tables remain inaccessible. This validate action addresses a different case (foreign catalog permission), where catalog discovery was mistaken for full table access. Unity Catalog usage and object privileges are still required for foreign data objects. In the stated problem, engineers export the full source database into cloud object storage each hour; this choice would not establish the required federated connection decision.

Confirm the partner reads only the approved shared dataset. This validate action addresses a different case (open protocol recipient), where the partner cannot use an internal Databricks-to-Databricks credential flow. Open Delta Sharing can support non-Databricks recipients using the sharing protocol. In the stated problem, engineers export the full source database into cloud object storage each hour; this choice would not establish the required federated connection decision.

Compare remote rows scanned and local query results before and after. This validate action addresses a different case (federated pushdown), where an analyst pulls millions of rows before a selective filter. Supported predicate pushdown can reduce data transferred from a federated source. In the stated problem, engineers export the full source database into cloud object storage each hour; this choice would not establish the required federated connection decision.

Inspect MERGE condition and target files read. This diagnose action addresses a different case (predicate-limited merge), where the merge scans unrelated historical partitions. Constraining merge matches to relevant partitions can reduce unnecessary target scanning. In the stated problem, the consumer reads only newly appended rows and cannot see updates or deletes; this choice would not establish the required change data feed consumers decision.

Inspect per-task shuffle read bytes and spilled data. This diagnose action addresses a different case (shuffle partition count), where the current partition count is too low for data volume. Shuffle partition sizing affects memory pressure and task parallelism. In the stated problem, the consumer reads only newly appended rows and cannot see updates or deletes; this choice would not establish the required change data feed consumers decision.

Inspect filters and partition scan counts in the physical plan. This diagnose action addresses a different case (partition pruning), where a query wraps business_date in a transformation and reads many partitions. Direct predicates on partition keys can enable partition pruning. In the stated problem, the consumer reads only newly appended rows and cannot see updates or deletes; this choice would not establish the required change data feed consumers decision.

Create a test table and confirm approved analysts can query it. This validate action addresses a different case (schema-level grant inheritance), where analysts must repeatedly request identical SELECT grants for each table. Unity Catalog schema-level privileges can apply to existing and future child objects. In the stated problem, new data stewards cannot maintain its grants; this choice would not establish the required ownership reassignment decision.

Confirm managed table storage resolves to the approved location. This validate action addresses a different case (managed storage ownership), where the data platform cannot apply consistent lifecycle governance. Managed locations centralize storage ownership for Unity Catalog managed assets. In the stated problem, new data stewards cannot maintain its grants; this choice would not establish the required ownership reassignment decision.

Read the intended dataset and verify access is denied outside the scope. This validate action addresses a different case (external location governance), where raw cloud permissions bypass central catalog governance. External locations map cloud paths to governed Unity Catalog storage credentials. In the stated problem, new data stewards cannot maintain its grants; this choice would not establish the required ownership reassignment decision.

Inspect expectation metric counts in the event log. This diagnose action addresses a different case (expectation monitoring), where analysts confuse successful pipeline status with zero rejected records. Quality metric events reveal input and dropped records separately from update status. In the stated problem, the Jobs status only reports a high-level failure; this choice would not establish the required pipeline event log decision.

Inspect current notification settings and run failure statuses. This diagnose action addresses a different case (job email notification), where operators manually check the run page every morning. Lakeflow Jobs support status-triggered notifications for run outcomes. In the stated problem, the Jobs status only reports a high-level failure; this choice would not establish the required pipeline event log decision.

Inspect run status, task IDs and time windows in supported jobs system tables. This diagnose action addresses a different case (job operational history), where operators only see the most recent run in the Jobs UI. Jobs system records support historical analysis of runs and tasks. In the stated problem, the Jobs status only reports a high-level failure; this choice would not establish the required pipeline event log decision.

Select append writes and stable event IDs for immutable source events. This design action addresses a different case (append-only delta bronze), where a nightly overwrite accidentally removes historical audit events. Append-only ingestion keeps previous raw records while adding new batches. In the stated problem, restarting with a fresh checkpoint replays previously ingested messages; this choice would not establish the required kafka offset durability decision.

Use Auto Loader with managed file discovery and checkpoint/schema locations. This design action addresses a different case (auto loader file discovery), where listing the full historical directory every run exceeds the ingestion SLA. Auto Loader is designed for incremental new-file discovery with persisted state. In the stated problem, restarting with a fresh checkpoint replays previously ingested messages; this choice would not establish the required kafka offset durability decision.

Use one durable checkpoint and clear idempotency keys for replay. This design action addresses a different case (retry-safe file loads), where a partially processed input file is loaded twice into Delta. Checkpointed incremental ingestion or stable event-level deduplication prevents repeated effects. In the stated problem, restarting with a fresh checkpoint replays previously ingested messages; this choice would not establish the required kafka offset durability decision.

Inspect row-level expectation metrics alongside pipeline event types. This diagnose action addresses a different case (quality metric interpretation), where analysts misread a quality count as the number of pipeline task failures. Expectation metrics describe data rows and must be interpreted separately from orchestration status. In the stated problem, two events have the same event time and yield different surviving rows; this choice would not establish the required latest-per-key with ties decision.

Inspect existing expectation rules and failed-update event details. This diagnose action addresses a different case (expectation fail action), where the pipeline continues despite critical null records. Fail-on-violation expectations stop updates rather than silently filtering rows. In the stated problem, two events have the same event time and yield different surviving rows; this choice would not establish the required latest-per-key with ties decision.

Inspect original timezone offsets and parsed timestamp semantics. This diagnose action addresses a different case (timestamp timezone normalization), where events from different regions are ordered incorrectly in a UTC dashboard. Time-zone normalization is required before comparing timestamps from different sources. In the stated problem, two events have the same event time and yield different surviving rows; this choice would not establish the required latest-per-key with ties decision.

Query supported billing system tables and join approved usage metadata. This implement action addresses a different case (billing usage monitoring), where cost reports omit some job and workspace labels. Databricks system billing tables can expose usage records for cost attribution. In the stated problem, no baseline is available for the warehouse query execution; this choice would not establish the required sql query history decision.

Record stage timestamps and expose a freshness query. This implement action addresses a different case (end-to-end sla), where operators monitor compute uptime rather than data freshness. SLA monitoring should measure source event time to published table availability. In the stated problem, no baseline is available for the warehouse query execution; this choice would not establish the required sql query history decision.

Set the supported job notification destination for failed runs. This implement action addresses a different case (job email notification), where operators manually check the run page every morning. Lakeflow Jobs support status-triggered notifications for run outcomes. In the stated problem, no baseline is available for the warehouse query execution; this choice would not establish the required sql query history decision.

Inspect run intervals, executor utilization and billed usage. This diagnose action addresses a different case (compute utilization), where most provisioned resources sit idle between job runs. Compute metrics and usage history distinguish idle capacity from active work. In the stated problem, consumer lag grows while microbatch intervals appear stable; this choice would not establish the required streaming backlog decision.

Inspect stage summary, shuffle read sizes and longest task duration. This diagnose action addresses a different case (spark ui shuffle metrics), where operators only monitor cluster-wide CPU and miss skewed task durations. Spark UI stage metrics reveal shuffle, spill and task skew. In the stated problem, consumer lag grows while microbatch intervals appear stable; this choice would not establish the required streaming backlog decision.

Inspect query history for SQL text identity, duration and warehouse. This diagnose action addresses a different case (sql query history), where no baseline is available for the warehouse query execution. Query history/system query records support comparison of execution time and workload. In the stated problem, consumer lag grows while microbatch intervals appear stable; this choice would not establish the required streaming backlog decision.

Cause a safe test failure and confirm exactly one incident is created. This validate action addresses a different case (job webhook destination), where only personal email notifications are configured. Supported job notification destinations can integrate workflow status with an incident system. In the stated problem, workspace notebook logs do not show the full audited account activity; this choice would not establish the required audit log review decision.

Insert a new source event and measure time until it appears downstream. This validate action addresses a different case (pipeline update cadence), where scheduled updates are spaced several hours apart. Pipeline update frequency determines available end-to-end freshness. In the stated problem, workspace notebook logs do not show the full audited account activity; this choice would not establish the required audit log review decision.

Confirm the profile scans fewer files with correct result rows. This validate action addresses a different case (query profile scans), where operators assume the query is always CPU-bound. Query Profile identifies scan, pruning and shuffle behavior for SQL workloads. In the stated problem, workspace notebook logs do not show the full audited account activity; this choice would not establish the required audit log review decision.

Monitor entrypoint failures after runtime or package upgrades. This monitor action addresses a different case (python task entrypoints), where the task exits before setting up its argument parser. A standard Python entrypoint and explicit arguments avoid notebook runtime assumptions. In the stated problem, engineers assume an intermediate filter executed before any action; this choice would not establish the required spark lazy evaluation decision.

Monitor job run arguments for unexpected target or region values. This monitor action addresses a different case (job parameters), where code contains hard-coded region IDs and catalog names. Task parameters allow runtime-specific values without changing notebook source. In the stated problem, engineers assume an intermediate filter executed before any action; this choice would not establish the required spark lazy evaluation decision.

Monitor refresh duration and staleness against dashboard SLA. This monitor action addresses a different case (materialized view choice), where each dashboard query recomputes the same expensive aggregation. A materialized view maintains a precomputed query result through refresh. In the stated problem, engineers assume an intermediate filter executed before any action; this choice would not establish the required spark lazy evaluation decision.