Microsoft DP-700 Practice Test 35

Question 340

A dataflow has a valid saved Power Query definition, but its last scheduled refresh failed and downstream rows stayed old. What explains the discrepancy?

A saved dataflow definition is not evidence that the most recent refresh and destination write completed successfully.
The Dataflow's saved query definition automatically refreshes all published destinations.
The dashboard must be in a Fabric domain with the same name as the Dataflow.
The dataflow's preview used a small sample, which deletes all production destination rows.

Explanation

Correct Answer

Refresh execution has its own status and per-activity diagnostics; design-time validity alone does not update the output.

Incorrect Answers

Persisting Power Query steps doesn't mean a scheduled refresh succeeded, so source rows may stay unchanged.

Domain organization does not execute Power Query refreshes or guarantee updated table contents.

A small preview does not ordinarily delete destination tables; inspect the failed refresh error chain.

Question 341

After repairing a failed Dataflow Gen2 refresh, which evidence demonstrates recovery?

Confirm that the Power Query editor preview contains recent records without checking the refresh.
Inspect the next successful refresh record and specific destination activity metrics, then confirm that representative target rows changed as expected.
Read only the dashboard's last-visited timestamp, not the Dataflow's run history.
Validate the source connection using a different privileged identity from the scheduled refresh account.

Explanation

Correct Answer

A completed refresh paired with target-state verification distinguishes operational recovery from a misleading green editor save.

Incorrect Answers

Editor previews can show current source data while a scheduled destination refresh remains failed.

A viewer timestamp cannot establish when Power Query transformations and destination writes last succeeded.

Permission differences can hide the failed production refresh identity or gateway connection issue.

Question 342

A Fabric notebook that previously completed quickly now consumes much more capacity. Which observability sources should be compared?

Compare only notebook elapsed time and ignore Spark task-level skew and resource metrics.
Raise the Spark driver memory limit without inspecting the slow executor-stage tasks.
Inspect Spark application monitoring, job duration, executor/driver utilization and the corresponding Fabric capacity metrics for the run.
Check Dataflow refresh history instead of the notebook's Spark application and Capacity Metrics.

Explanation

Correct Answer

Per-stage Spark telemetry and capacity usage distinguish a code/data skew issue from workspace-wide pressure.

Incorrect Answers

Runtime alone cannot distinguish more data, skewed shuffle, slower library/runtime or capacity contention.

Driver memory may not address a few skewed partitions responsible for long executor task tails.

An unrelated Dataflow history cannot attribute the notebook's regression to Spark tasks or shared capacity.

Question 343

A notebook’s wall-clock duration increases, and the last Spark stage shows severe task skew while other tasks finish quickly. What explains the slow run?

The long Spark stage must be a delayed Dataflow Gen2 visual refresh.
The notebook needs a new Fabric domain rather than changes to partitioning or joins.
The job should disable Spark monitoring to avoid overhead during the slow stage.
One or a few skewed partitions dominate the critical path even after most parallel Spark tasks have finished.

Explanation

Correct Answer

Skew concentrates work in straggler tasks; aggregate CPU or total row count alone can conceal this bottleneck.

Incorrect Answers

A notebook's Spark task skew concerns Spark partition distribution, not the unrelated Dataflow user interface.

Domain membership cannot rebalance dominant keys that make a handful of Spark tasks much slower.

Suppressing diagnostics would hide the skew evidence and cannot correct its execution imbalance.

Question 344

A Spark job optimization is proposed after an unexplained cost spike. What is a sound validation?

Benchmark equivalent inputs before and after the change, comparing stage duration, task skew, executor utilization and total capacity consumption.
Compare notebook runtime before and after without keeping input sizes or query plans comparable.
Check only aggregate capacity-unit consumption rather than long-tail Spark task metrics.
Assert optimization success when a new result has fewer rows but no reconciliation.

Explanation

Correct Answer

Controlled per-stage and capacity comparisons test whether the optimization helps the same workload instead of only shifting the timing window.

Incorrect Answers

Timing alone can misattribute a larger dataset or changed query plan to a proposed optimization.

Capacity totals cannot show whether the specifically skewed stage and stragglers improved.

Fewer output rows can reflect data loss, not performance improvement from safe Spark tuning.

Question 345

An Eventstream source reports 1,000 accepted messages per minute but the intended destination receives only 700. What should be compared?

Treat all 300 missing messages as delivery loss before reviewing the configured Filter branch.
Inspect source throughput, transform/filter behavior, destination ingress metrics and error/drop indicators in Eventstream monitoring.
Disable source event instrumentation so input counts more closely resemble output.
Increase KQL storage retention to force previously filtered records into Eventhouse.

Explanation

Correct Answer

Input-to-output flow measurements locate lost or intentionally filtered events between the source and each routed destination.

Incorrect Answers

The source/destination count difference may be intentional if debug severities are excluded.

Suppressing telemetry conceals the discrepancy and cannot establish correct routing.

Retention applies to accepted records; it does not change upstream Eventstream routing/filtering.

Question 346

A live Eventstream sends fewer events to Eventhouse than its input counter. A Filter deliberately removes debug messages. Why can the counters differ?

Every Eventstream source message must appear at every destination, even when an intentional Filter is attached.
The Eventhouse destination is guaranteed to receive the same count because it uses KQL.
The filter intentionally excludes part of the source stream, so destination output counts need not match unfiltered input totals.
The missing 300 messages prove OneLake deleted a source partition after Eventstream publication.

Explanation

Correct Answer

Monitoring must distinguish expected filtered records from failed or dropped messages rather than treating every count difference as data loss.

Incorrect Answers

The filtered branch legitimately excludes debug events; the counters measure different points in the stream.

Query language doesn't override upstream transformation rules that drop unwanted source categories.

A documented filter is a more direct explanation than a speculative physical storage deletion.

Question 347

An Eventstream throughput discrepancy is investigated. What controlled test establishes the explanation?

Compare total source input and destination counts without segmenting records by severity.
Run tests using only critical events and conclude the debug-exclusion rule works.
Inspect dashboard refresh age instead of destination event IDs and filtered routing.
Send a known mix of permitted and filtered events, then reconcile source accepted, filter output and target ingested counts plus errors.

Explanation

Correct Answer

A labelled event fixture proves whether the count delta follows the intended routing/filter policy or indicates a delivery failure.

Incorrect Answers

Aggregate counts alone cannot distinguish intentional filter removal from unwanted delivery loss.

Critical-only test data cannot prove that debug messages are excluded as configured.

Dashboard refresh timing doesn't show where individual source events are filtered or dropped.

Question 348

An Eventhouse operator needs evidence of query throttling, ingestion errors and capacity pressure. Which first-party monitoring feature should be enabled or accessed?

Use Eventhouse workspace monitoring and its monitoring KQL database, alongside relevant Eventhouse overview and capacity metrics.
Inspect only the latest user's KQL text, without capacity or ingestion telemetry.
Use Spark Monitoring Hub executor counters as the sole health record for Eventhouse.
Rely on the Eventhouse item listing without enabling or querying detailed monitoring data.

Explanation

Correct Answer

Eventhouse monitoring exposes query, ingestion and throttling records that are necessary for resource and failure analysis.

Incorrect Answers

Slow Eventhouse queries may reflect concurrent capacity pressure, throttling or ingestion faults beyond text shape.

Spark executor metrics don't cover native Eventhouse KQL engine ingestion and query throttling.

The item existing doesn't provide actual error, resource or performance time-series evidence.

Question 349

An Eventhouse appears slow but an engineer checks only one user’s KQL query text, not the service’s capacity telemetry. What diagnostic boundary is missing?

A query running slowly can always be diagnosed from its KQL text alone.
The investigation omits Eventhouse-wide query/capacity/throttling observations that can explain shared resource contention.
The Eventhouse monitoring database is unnecessary if the query returns some rows.
Spark executor metrics in an unrelated workspace fully describe Eventhouse ingestion health.

Explanation

Correct Answer

A well-formed single query can still wait when overall service concurrency or capacity is constrained.

Incorrect Answers

The cause may be capacity, concurrent workload, throttling or ingestion; text inspection omits service telemetry.

Successful queries can still run under pressure or be throttled intermittently; operational metrics are required.

Spark job resource stats are not equivalent to Eventhouse query and ingestion engine telemetry.

Question 350

An Eventhouse monitoring database is enabled. Which validation confirms useful operational signals are being collected?

Verify that a monitoring database object exists but never run its diagnostic queries.
Inspect only the last successful query and ignore failed or throttled requests.
Query the relevant monitoring tables for recent ingestion and query activity, including known throttling events and corresponding timestamps.
Use a static screenshot of a quiet-hour capacity gauge as peak-period evidence.

Explanation

Correct Answer

Actual records correlated to workload times demonstrate that the monitoring source is populated and relevant.

Incorrect Answers

The item existing does not prove telemetry categories are populated or available to operators.

Successful requests don't establish absence of intermittent failures or ingestion errors under load.

One low-load screenshot cannot establish query contention at the reported incident time.

Implementing Data Engineering Solutions Using Microsoft Fabric (DP-700) Practice Tests

Microsoft DP-700 Practice Test 33 - 10 Questions (Last Reviewed 02-Oct-2026)
Microsoft DP-700 Practice Test 34 - 10 Questions (Last Reviewed 01-Oct-2026)
Microsoft DP-700 Practice Test 35 - 10 Questions (Last Reviewed 30-Sep-2026)
Microsoft DP-700 Practice Test 36 - 10 Questions (Last Reviewed 29-Sep-2026)
Microsoft DP-700 Practice Test 37 - 10 Questions (Last Reviewed 28-Sep-2026)

Pass Your Microsoft DP-700 Exam Easy!

Question 340

Explanation

Correct Answer

Incorrect Answers

Question 341

Explanation

Correct Answer

Incorrect Answers

Question 342

Explanation

Correct Answer

Incorrect Answers

Question 343

Explanation

Correct Answer

Incorrect Answers

Question 344

Explanation

Correct Answer

Incorrect Answers

Question 345

Explanation

Correct Answer

Incorrect Answers

Question 346

Explanation

Correct Answer

Incorrect Answers

Question 347

Explanation

Correct Answer

Incorrect Answers

Question 348

Explanation

Correct Answer

Incorrect Answers

Question 349

Explanation

Correct Answer

Incorrect Answers

Question 350

Explanation

Correct Answer

Incorrect Answers

Implementing Data Engineering Solutions Using Microsoft Fabric (DP-700) Practice Tests