Microsoft DP-700 Practice Test 30

Question 290

After introducing an enrichment join, events from devices absent in the lookup table have missing type values. What explains the gap?

Every unregistered device must automatically be assigned the alphabetically first registered type.
The streaming engine lacks enough nodes to normalize the device ID column.
The join reference does not have a matching device key or the join policy excludes unmatched events.
Sorting incoming events by timestamp should produce their static device type.

Explanation

Correct Answer

A lookup can attach values only for matching source identifiers; the team must define handling for unknown devices and reference refresh timing.

Incorrect Answers

A reference-key miss should follow an explicit unmatched-event policy, not an arbitrary default attribute.

Insufficient compute would affect throughput; no matching lookup row explains missing enrichment for specific IDs.

Temporal ordering doesn't retrieve device attributes without a matching reference record.

Question 291

A stream enrichment process is implemented. Select TWO checks that demonstrate correct behavior.

Test only an event from a device with a missing reference row and require a fabricated type.
Compare counts before and after the join but never inspect the enriched type column.
Send a known device ID and confirm its destination record contains the expected reference attributes.
Send an unknown device ID and confirm the specified unmatched-event policy, such as a null type or quarantine, is honored.
Validate enrichment using a production administrator who has broader reference-data access.

Explanation

Correct Answers

A matched reference fixture verifies that the enrichment path actually joins the correct key.

Negative-key testing detects silent loss or accidental mislabeling of unrecognized devices.

Incorrect Answers

A missing key should exercise the documented null or quarantine policy, not invent a type.

Equal counts can coexist with incorrect device-type values or failed lookup mappings.

Admin success may hide the intended processor identity's access limitation to the device lookup table.

Question 292

An Eventhouse table holds one event per sensor and the analyst must count events in five-minute calendar bins. Which KQL pattern fits?

Use `summarize count() by bin(EventTime, 5m)` with any required grouping dimensions and an event-time filter.
Run `summarize count()` without a `by bin(EventTime,5m)` group.
Use `take 5` after the KQL query and interpret the five rows as five-minute bins.
Group by `EventTime` raw timestamp without binning.

Explanation

Correct Answer

KQL binning assigns timestamps to fixed-width buckets and summarize calculates per-bucket counts.

Incorrect Answers

A bare summarize collapses all selected events into one total, not five-minute time buckets.

`take` limits rows, not window duration or grouping semantics.

Individual high-resolution timestamps create many tiny groups rather than required five-minute intervals.

Question 293

A KQL time series returns one global total instead of counts per five-minute window. Its query uses `summarize count()` without `bin()`. Why?

The summarize statement omitted `take 5` before counting records.
The summarize grouping clause does not include the time bin, so all selected records collapse into one aggregate.
The EventTime column is named incorrectly for time-series aggregation.
The Eventhouse database lacks SQL dynamic data masking on sensor IDs.

Explanation

Correct Answer

KQL computes one result per group; omitting the grouping time expression removes temporal partitions.

Incorrect Answers

Taking five rows limits records, not the number or size of event-time groups.

A column name only matters if referenced correctly; a missing `by bin` collapses all event times.

Masking does not create five-minute groups for a KQL count.

Question 294

A query was revised to count events per five-minute bin. What fixture most directly verifies the boundaries?

Verify the grand total equals the number of input events without checking bin assignments.
Use five sample rows from exactly the same minute and assume adjacent bins are correct.
Insert sample events just before, on and after five-minute boundaries and check their `bin(EventTime,5m)` grouping and counts.
Compare only the output column's name to the specified five-minute label.

Explanation

Correct Answer

Boundary events test the grouping semantics and whether the chosen event-time column, not ingest time, drives bins.

Incorrect Answers

A global count can match while all records were incorrectly combined into one time bucket.

A same-minute fixture cannot reveal five-minute boundary handling.

Column names don't prove `bin(EventTime,5m)` grouping and counts at the timestamp boundary.

Question 295

A KPI must represent the last 15 minutes of events, recalculated every minute, not a disjoint set of 15-minute buckets. Which window semantics are appropriate?

Use 15-minute tumbling windows and refresh the dashboard every minute.
Use a one-minute tumbling window and report it as the last 15 minutes.
Sort all events by event ID and compute a single all-time aggregate.
Use a supported sliding/hopping window with the required duration and advance interval, subject to engine-specific event-time support.

Explanation

Correct Answer

An overlapping 15-minute window updated each minute measures a rolling period, unlike nonoverlapping tumbling windows.

Incorrect Answers

UI refresh cannot create overlapping windows; the underlying aggregate still advances at 15-minute boundaries.

One-minute nonoverlapping totals omit the preceding fourteen minutes from each rolling KPI.

An all-time count ignores the rolling time horizon and required one-minute recalculation.

Question 296

A streaming KPI changes only every 15 minutes although product requirements specify updates every minute. Its aggregation uses a nonoverlapping 15-minute tumbling window. Why?

The chosen tumbling window advances at full-window boundaries rather than computing overlapping snapshots each minute.
The sliding KPI failed because Eventhouse ingestion requires one warehouse table per minute.
The chart refresh interval was configured to 15 seconds while the query uses 15-minute tumbling bins.
Sorting every event by ID would permit the tumbling calculation to slide.

Explanation

Correct Answer

Disjoint windows cannot represent a continually updated 15-minute sliding measure by themselves.

Incorrect Answers

An ingestion table count does not determine the temporal advance interval of a windowed aggregation.

Faster display refresh cannot create overlapping one-minute windows from a fixed 15-minute tumbling query.

ID sorting is independent of event-time window advancement semantics.

Question 297

A sliding-window event KPI replaces a tumbling aggregate. What validation proves the intended behavior?

Test only at the end of a full fifteen-minute block.
Use events with known timestamps across a 15-minute span and check consecutive one-minute results share the correct overlapping input set.
Check the chart title for the phrase Rolling 15m rather than query execution windows.
Compare only one output count before and after changing the window strategy.

Explanation

Correct Answer

Observing overlapping membership and expected changes at each minute distinguishes sliding from tumbling output.

Incorrect Answers

One boundary can pass even when intermediate one-minute overlapping updates never occur.

A visual label cannot establish the duration, hop interval or eligible event-time membership.

A one-point comparison cannot establish correct overlap across successive one-minute observations.

Question 298

A high-volume Delta-compatible event feed needs stateful PySpark transformations and a Delta sink. Which runtime should implement it?

Store the Structured Streaming checkpoint in an ephemeral notebook-local directory.
Use stateless scheduled Spark batches with blind appends to the Delta sink.
Use Spark Structured Streaming with supported source and sink formats, an explicit checkpoint and compatible output mode in a Fabric notebook/job.
Choose a non-Delta sink while claiming Delta transactional output guarantees.

Explanation

Correct Answer

Structured Streaming supports distributed incremental processing and stateful recovery for Spark code rather than relying on periodic full table scans.

Incorrect Answers

A restarted job can lose offsets/state and reprocess previously committed events.

The requirement calls for stateful PySpark streaming and replay-safe writes, not independent uncontrolled appends.

A sink without the required transactional semantics doesn't meet the requested Delta-compatible storage behavior.

Question 299

A Spark stream reprocesses old source data after restart when the checkpoint path changes between runs. What is the root issue?

The Fabric workspace default Spark pool was resized after the restart.
The event producer registered new rows with different Fabric sensitivity labels.
The output Delta table lacks V-Order optimization for its stored Parquet files.
The restarted query has lost its prior offset/state checkpoint and therefore starts with fresh-stream semantics.

Explanation

Correct Answer

The durable checkpoint associates progress with the query; changing it can replay previously processed inputs.

Incorrect Answers

Pool sizing does not restore lost structured-streaming offsets or replay protection if the checkpoint URI changes.

Labels don't substitute for the checkpoint state that tracks which offsets the streaming query committed.

Storage layout cannot recover a streaming query's previous state/offset checkpoint after a path change.

Question 300

A Structured Streaming job has been configured for restart safety. What acceptance test is meaningful?

Stop and restart after a controlled batch using the same checkpoint, then reconcile new and previously committed output keys without unintended duplicates.
Delete the checkpoint and rerun from the beginning, then accept duplicated sink records.
Check the notebook executes without exceptions, but do not reconcile output event IDs.
Run two separate queries with different checkpoint paths and compare only their input counts.

Explanation

Correct Answer

A two-session test demonstrates checkpoint recovery and sink consistency rather than just initial streaming success.

Incorrect Answers

Removing checkpoint state invites reprocessing instead of verifying stable recovery and idempotent output.

A successful execution can duplicate previously committed records after restart.

Different paths mean different query state; similar counts don't establish exactly-once business results.

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

Microsoft DP-700 Practice Test 28 - 10 Questions (Last Reviewed 07-Oct-2026)
Microsoft DP-700 Practice Test 29 - 10 Questions (Last Reviewed 06-Oct-2026)
Microsoft DP-700 Practice Test 30 - 10 Questions (Last Reviewed 05-Oct-2026)
Microsoft DP-700 Practice Test 31 - 10 Questions (Last Reviewed 04-Oct-2026)
Microsoft DP-700 Practice Test 32 - 10 Questions (Last Reviewed 03-Oct-2026)

Pass Your Microsoft DP-700 Exam Easy!

Question 290

Explanation

Correct Answer

Incorrect Answers

Question 291

Explanation

Correct Answers

Incorrect Answers

Question 292

Explanation

Correct Answer

Incorrect Answers

Question 293

Explanation

Correct Answer

Incorrect Answers

Question 294

Explanation

Correct Answer

Incorrect Answers

Question 295

Explanation

Correct Answer

Incorrect Answers

Question 296

Explanation

Correct Answer

Incorrect Answers

Question 297

Explanation

Correct Answer

Incorrect Answers

Question 298

Explanation

Correct Answer

Incorrect Answers

Question 299

Explanation

Correct Answer

Incorrect Answers

Question 300

Explanation

Correct Answer

Incorrect Answers

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