cert
cert-1
cert-2

Pass Microsoft DP-700 Exam in First Attempt Guaranteed!

Get 100% Latest Exam Questions, Accurate & Verified Answers to Pass the Actual Exam!
30 Days Free Updates, Instant Download!

cert-5
cert-6
DP-700 Exam - Verified By Experts
DP-700 Premium Bundle
$14.99

DP-700 Premium Bundle

$74.99
$89.98
  • Premium File 160 Questions & Answers. Last update: Nov 10, 2026
  • Training Course 160 Video Lectures
 
$89.98
$74.99
accept 159 downloads in last 7 days
block-screenshots
Purchase Microsoft DP-700 Exam Training Products Individually
Microsoft DP-700 Premium File
Premium File 160 Q&A
$76.99$69.99
Microsoft DP-700 Video Training Course
Training Course 160 Lectures
$21.99 $19.99
Microsoft DP-700 Practice Test 27

Question 260

A source is connected and a default stream appears automatically. What should the operator understand?

  1. Use a supported Manage Fields transformation to select and rename output fields.
  2. Publish the modified Eventstream so destination changes take effect.
  3. Apply a Filter transformation before the critical-alarm destination.
  4. Use the generated default stream as the raw event flow before any derived transformations.

Explanation

Correct Answer

D:

Every Eventstream source flow produces the default stream; a derived stream appears after supported operations are defined. Compare raw and transformed stream records.

Incorrect Answers

A:

Field management shapes streaming events before delivery without changing the source producer.

B:

Editing the stream canvas alone does not activate changes in the running eventstream.

C:

A filter restricts events using explicit conditions while preserving the upstream raw stream for other outputs.

Question 261

A KQL analyst wants event counts in disjoint five-minute buckets. Which expression pattern fits?

  1. Use a sliding window of ten minutes with a two-minute slide.
  2. Use summarize count() by bin(event_time, 5m) in KQL.
  3. Make the foreachBatch write idempotent using batch identifiers or stable business keys.
  4. Increase the watermark delay to a justified arrival-lateness tolerance with a state-cost review.

Explanation

Correct Answer

B:

Binning creates nonoverlapping time buckets for grouped event counts. Test records around a five-minute boundary.

Incorrect Answers

A:

Sliding windows overlap, unlike tumbling windows, and allow each event to contribute to multiple windows.

C:

Custom sinks can re-execute batches after failure; checkpointing alone does not make arbitrary external side effects exactly once.

D:

An overly short watermark risks dropping valid late data from stateful results, while a longer delay retains more state.

Question 262

A stream processor needs overlapping ten-minute windows every two minutes. What window type is required?

  1. Filter invalid readings and summarize avg(temperature) by plant_id, bin(event_time, 1h).
  2. Use a sliding window of ten minutes with a two-minute slide.
  3. Define an event-time watermark based on acceptable source lateness.
  4. Increase the watermark delay to a justified arrival-lateness tolerance with a state-cost review.

Explanation

Correct Answer

B:

Sliding windows overlap, unlike tumbling windows, and allow each event to contribute to multiple windows. Check an event's membership in consecutive windows.

Incorrect Answers

A:

Filtering invalid values before grouped averaging yields interpretable per-plant, per-hour measures.

C:

Watermarks bound old aggregation state while retaining events within the planned lateness tolerance.

D:

An overly short watermark risks dropping valid late data from stateful results, while a longer delay retains more state.

Question 263

A Spark streaming aggregation retains huge state because late data is unbounded. What should be configured?

  1. Define an event-time watermark based on acceptable source lateness.
  2. Filter invalid readings and summarize avg(temperature) by plant_id, bin(event_time, 1h).
  3. Use a sliding window of ten minutes with a two-minute slide.
  4. Make the foreachBatch write idempotent using batch identifiers or stable business keys.

Explanation

Correct Answer

A:

Watermarks bound old aggregation state while retaining events within the planned lateness tolerance. Measure state size and test delayed events.

Incorrect Answers

B:

Filtering invalid values before grouped averaging yields interpretable per-plant, per-hour measures.

C:

Sliding windows overlap, unlike tumbling windows, and allow each event to contribute to multiple windows.

D:

Custom sinks can re-execute batches after failure; checkpointing alone does not make arbitrary external side effects exactly once.

Question 264

A monitoring query must identify the latest event for each device while retaining the associated payload. Which KQL pattern is suitable?

  1. Increase the watermark delay to a justified arrival-lateness tolerance with a state-cost review.
  2. Define an event-time watermark based on acceptable source lateness.
  3. Give each distinct streaming query its own durable checkpoint location.
  4. Use summarize arg_max(event_time, *) by device_id.

Explanation

Correct Answer

D:

Arg_max selects a row corresponding to each group's maximum timestamp, unlike max() alone, which loses other row fields. Compare returned payloads with known latest timestamps.

Incorrect Answers

A:

An overly short watermark risks dropping valid late data from stateful results, while a longer delay retains more state.

B:

Watermarks bound old aggregation state while retaining events within the planned lateness tolerance.

C:

Checkpoint metadata belongs to one query's offsets and state; sharing directories can corrupt recovery and duplicate results.

Question 265

A producer sometimes sends events 15 minutes late, but the Spark watermark is five minutes. What should the engineer reconsider?

  1. Filter invalid readings and summarize avg(temperature) by plant_id, bin(event_time, 1h).
  2. Use summarize count() by bin(event_time, 5m) in KQL.
  3. Use summarize arg_max(event_time, *) by device_id.
  4. Increase the watermark delay to a justified arrival-lateness tolerance with a state-cost review.

Explanation

Correct Answer

D:

An overly short watermark risks dropping valid late data from stateful results, while a longer delay retains more state. Replay 15-minute-late events and monitor state size.

Incorrect Answers

A:

Filtering invalid values before grouped averaging yields interpretable per-plant, per-hour measures.

B:

Binning creates nonoverlapping time buckets for grouped event counts.

C:

Arg_max selects a row corresponding to each group's maximum timestamp, unlike max() alone, which loses other row fields.

Question 266

A structured streaming job restarts and unexpectedly repeats output after using a shared checkpoint directory. What design is required?

  1. Give each distinct streaming query its own durable checkpoint location.
  2. Use summarize count() by bin(event_time, 5m) in KQL.
  3. Define an event-time watermark based on acceptable source lateness.
  4. Make the foreachBatch write idempotent using batch identifiers or stable business keys.

Explanation

Correct Answer

A:

Checkpoint metadata belongs to one query's offsets and state; sharing directories can corrupt recovery and duplicate results. Restart two queries independently and verify offsets.

Incorrect Answers

B:

Binning creates nonoverlapping time buckets for grouped event counts.

C:

Watermarks bound old aggregation state while retaining events within the planned lateness tolerance.

D:

Custom sinks can re-execute batches after failure; checkpointing alone does not make arbitrary external side effects exactly once.

Question 267

A KQL query must return hourly average temperature by plant while excluding null sensor values. What aggregation fits?

  1. Increase the watermark delay to a justified arrival-lateness tolerance with a state-cost review.
  2. Filter invalid readings and summarize avg(temperature) by plant_id, bin(event_time, 1h).
  3. Make the foreachBatch write idempotent using batch identifiers or stable business keys.
  4. Use a sliding window of ten minutes with a two-minute slide.

Explanation

Correct Answer

B:

Filtering invalid values before grouped averaging yields interpretable per-plant, per-hour measures. Verify one known hourly sample by hand.

Incorrect Answers

A:

An overly short watermark risks dropping valid late data from stateful results, while a longer delay retains more state.

C:

Custom sinks can re-execute batches after failure; checkpointing alone does not make arbitrary external side effects exactly once.

D:

Sliding windows overlap, unlike tumbling windows, and allow each event to contribute to multiple windows.

Question 268

A foreachBatch Spark sink writes to a nontransactional service and duplicates records after recovery. Which safeguard is needed?

  1. Make the foreachBatch write idempotent using batch identifiers or stable business keys.
  2. Increase the watermark delay to a justified arrival-lateness tolerance with a state-cost review.
  3. Use summarize count() by bin(event_time, 5m) in KQL.
  4. Define an event-time watermark based on acceptable source lateness.

Explanation

Correct Answer

A:

Custom sinks can re-execute batches after failure; checkpointing alone does not make arbitrary external side effects exactly once. Replay a batch and compare target records.

Incorrect Answers

B:

An overly short watermark risks dropping valid late data from stateful results, while a longer delay retains more state.

C:

Binning creates nonoverlapping time buckets for grouped event counts.

D:

Watermarks bound old aggregation state while retaining events within the planned lateness tolerance.

Question 269

A Copy activity repeats three times and ultimately succeeds, but the SLA misses its deadline. What run information helps explain delay?

  1. Use the unique pipeline run ID in the monitoring view.
  2. Inspect per-attempt retry counts, error reasons, and the duration contributed by retries.
  3. Review pipeline activity start/end times and success dependency configuration.
  4. Track concurrent run counts, queueing or activity durations, and capacity pressure over time.

Explanation

Correct Answer

B:

A final success can hide repeated backoff and lost time; activity attempt history reveals why the total run was slow. Compare the retry timeline with a normal successful run.

Incorrect Answers

A:

Run IDs distinguish executions of the same pipeline, including reruns using identical parameters.

C:

Execution ordering is visible in the activity timeline and should match the intended dependency graph.

D:

Peak load diagnosis needs a time series of execution and resource behavior, not one successful sample.

Question 270

A pipeline has been intermittently failing for weeks. Where can an engineer review status across multiple executions?

  1. Use the Fabric monitoring hub run history filtered for the pipeline.
  2. Review pipeline activity start/end times and success dependency configuration.
  3. Track concurrent run counts, queueing or activity durations, and capacity pressure over time.
  4. Follow the upstream/downstream run relationship to the specific child run.

Explanation

Correct Answer

A:

Run history exposes success, failure, and duration patterns that a single editor execution cannot show. Plot failure timestamps by run ID.

Incorrect Answers

B:

Execution ordering is visible in the activity timeline and should match the intended dependency graph.

C:

Peak load diagnosis needs a time series of execution and resource behavior, not one successful sample.

D:

Run hierarchy connects orchestration stages so the relevant child error is not lost in parent status.

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