Q01 - Question
A Fabric administrator reviews the Microsoft Fabric Capacity Metrics app and sees that throttling occurs repeatedly during warehouse workloads. The administrator needs to interpret this result and decide what to do next. Which conclusion and action are supported?
Domain: Monitor and optimize an analytics solution (30–35%) Type: Single choice
- A. The warehouse queries are not parameterized. Rewrite the queries so that they use predicates.
- B. The warehouse sessions are not authenticated. Review the connections in sys.dm_exec_connections.
- C. The file operations to OneLake storage do not consume CUs. Stop monitoring storage activity.
- D. The processes require more capacity than the purchased capacity license provides. Use the utilization trends to optimize the capacity license.
D is correct.
Explanation: Throttling indicates that your processes require more capacity than is available within the constraints of your purchased capacity license. You can use the utilization trends in the Microsoft Fabric Capacity Metrics app to optimize the capacity license for your needs.
A is incorrect: Query parameterization is used so that query insights views can aggregate metrics. It is not a cause of throttling.
B is incorrect: Throttling is not an authentication condition. The DMVs describe connections, sessions, and requests, not capacity limits.
C is incorrect: In data warehouse workloads, CUs are consumed by data read and write activities. This includes the underlying file operations to OneLake storage, so storage activity still matters for cost.
Q02 - Question
You want to tune query performance in a Fabric data warehouse. You need to find SQL commands that have run more than once and that have the longest median time to complete. Which approach should you use?
Domain: Monitor and optimize an analytics solution (30–35%) Type: Single choice
- A. Query sys.dm_exec_requests and order the results by total_elapsed_time descending.
- B. Query queryinsights.frequently_run_queries and order the results by number_of_runs descending.
- C. Query queryinsights.long_running_queries, filter on number_of_runs greater than 1, and order by median_total_elapsed_time_ms descending.
- D. Query queryinsights.exec_requests_history and filter on start_time for the previous hour.
C is correct.
Explanation: The queryinsights.long_running_queries view returns details of query execution time. It includes number_of_runs and median_total_elapsed_time_ms. You can filter on number_of_runs greater than 1 and sort by the median time to find commands that are repeatedly slow.
A is incorrect: sys.dm_exec_requests returns information about active requests only. It does not provide historical median completion times.
B is incorrect: queryinsights.frequently_run_queries ranks commands by how often they run and shows successful and failed runs. It does not rank by median completion time.
D is incorrect: queryinsights.exec_requests_history returns details of each completed query. A time filter on this view does not give aggregated median execution times for repeated commands.
Q03 - Question
A pipeline load sometimes fails and leaves partial data in a lakehouse. You want a cleanup notebook to run automatically when the job fails, without waiting for someone to read an alert. Which solution meets this requirement?
Domain: Monitor and optimize an analytics solution (30–35%) Type: Single choice
- A. Configure the Schedule failures page to email the data engineering team when the pipeline fails.
- B. Add the Submitted by and Refresh type columns to the monitoring hub view.
- C. Create an Activator alert from Real-Time hub for the job failed event of the pipeline, and set the action to run a Fabric activity that starts the cleanup notebook.
- D. Create an Activator alert for the job failed event of the pipeline, and set the action to send an email.
C is correct.
Explanation: Activator connects to Fabric job events through Real-Time hub. When the job failed condition is met, the Run a Fabric activity action can trigger a notebook, such as a cleanup notebook, without human action.
A is incorrect: Schedule failures sends email alerts. It does not run a follow-up job, so someone must read the email and respond.
B is incorrect: Column options change what the monitoring hub displays. They do not trigger any action when a job fails.
D is incorrect: The email action notifies people but does not run the cleanup notebook. A person must still start the cleanup.
Q04 - Question
Users report that reports built on a Fabric data warehouse are slow right now. You need a list of the requests that are currently running, with the login name and client network address of each user, ordered by how long each request has been executing. What should you do?
Domain: Monitor and optimize an analytics solution (30–35%) Type: Single choice
- A. Join sys.dm_exec_connections, sys.dm_exec_sessions, and sys.dm_exec_requests, filter on requests that have a status of running, and order by total_elapsed_time descending.
- B. Query queryinsights.exec_requests_history and filter on start_time for the previous hour.
- C. Query queryinsights.frequently_run_queries and order by number_of_runs descending.
- D. Open the Microsoft Fabric Capacity Metrics app and filter the interface to warehouse activity.
A is correct.
Explanation: The dm_exec DMVs return information about the current state of the data warehouse. Joining connections, sessions, and requests gives you the login name, client address, command, and elapsed time. Filtering on running requests and ordering by total_elapsed_time descending lists the longest-running active requests first.
B is incorrect: queryinsights.exec_requests_history returns details of completed SQL queries. Queries can take up to 15 minutes to appear, so it is not a view of current activity.
C is incorrect: queryinsights.frequently_run_queries returns historical, aggregated details about frequently run commands. It does not list requests that are running now.
D is incorrect: The Microsoft Fabric Capacity Metrics app shows capacity utilization trends and throttling. It does not return login names, client addresses, or elapsed time for active requests.
Q05 - Question
Your organization uses Microsoft Fabric. A Fabric administrator must determine which warehouse processes are consuming capacity units (CUs) and whether the purchased capacity is being exceeded. Which approach should the administrator use?
Domain: Monitor and optimize an analytics solution (30–35%) Type: Single choice
- A. Query sys.dm_exec_sessions in the data warehouse and review the authenticated sessions.
- B. Install the Microsoft Fabric Capacity Metrics app and filter the interface to show only warehouse activity.
- C. Query queryinsights.long_running_queries and sort the results by median_total_elapsed_time_ms.
- D. Query queryinsights.frequently_run_queries and sort the results by number_of_runs.
B is correct.
Explanation: The Microsoft Fabric Capacity Metrics app is an app that an administrator can install in a Fabric environment and use to monitor capacity utilization. You can filter it to show only warehouse activity and observe utilization trends to determine what processes are consuming CUs and whether any throttling is occurring.
A is incorrect: sys.dm_exec_sessions returns information about authenticated sessions. It does not show CU consumption or throttling.
C is incorrect: queryinsights.long_running_queries returns details of query execution time. It does not show capacity utilization.
D is incorrect: queryinsights.frequently_run_queries returns details of frequently run queries. It does not show capacity utilization.
Q06 - Question
A data analyst uses query insights views to review how often the following statements run in a Fabric data warehouse: SELECT * FROM sales WHERE orderdate > '01/01/2023' and SELECT * FROM sales WHERE orderdate > '12/31/2021'. The analyst also runs a query and then immediately checks the views for it. Which two statements should the analyst use to interpret the results? (Select two.)
Domain: Monitor and optimize an analytics solution (30–35%) Type: Multi-select
- A. The two statements are considered the same query because queries with predicates are parameterized and the parameterized statements match exactly.
- B. The views show each query as soon as it completes, regardless of concurrent workloads.
- C. The two statements are always counted as separate queries because the date values are different.
- D. Depending on concurrent workloads, queries can take up to 15 minutes to appear in the query insights views.
A and D are correct.
Explanation: Query insights views provide aggregated metrics by parameterizing queries that have predicates. Two statements are treated as the same query when the parameterized statements are an exact match, so the two sales queries are counted together. Queries can also take up to 15 minutes to appear in the views, so a query that was just run might not be listed yet.
B is incorrect: Depending on concurrent workloads, queries can take up to 15 minutes to be reflected in the views. Do not expect a query to appear immediately.
C is incorrect: The different date values do not make the queries separate. After parameterization, the two statements match exactly and are counted as the same query.
Q07 - Question
A pipeline fails each morning because a schema change at the source breaks a table load. The on-call team works in Microsoft Teams. Which two actions should you perform to investigate the failure and notify the team?
Domain: Monitor and optimize an analytics solution (30–35%) Type: Multiple choice
- A. Configure Activator to run the same pipeline again on every failure.
- B. Configure an Activator rule on the job failed event that posts a message to the on-call Teams channel.
- C. Wait for analysts to report missing data before checking the monitoring hub.
- D. Open the activity details in the monitoring hub to review the error message.
B and D are correct.
Explanation: An Activator rule can post a message to a Teams channel when a job fails, which gives the on-call team immediate visibility. The activity details show the error message and error codes, which help identify the root cause, such as a schema change.
A is incorrect: If a job failed because of a source schema change, running a pipeline again just fails again. Activator is not a substitute for investigating and fixing the root cause.
C is incorrect: Relying on manual checks lets failures go undetected until users report missing data. You should configure notifications so the right people know about problems immediately.
Q08 - Question
You are a workspace Contributor. You want to configure email alerts for scheduled item failures across the whole workspace from one place. You do not want to open each item individually. Which approach should you use?
Domain: Monitor and optimize an analytics solution (30–35%) Type: Single choice
- A. Create an Activator rule for each scheduled item in Real-Time hub.
- B. Enable workspace monitoring and query the Eventhouse database for failures each morning.
- C. Open Historical runs for each item and review the last 30 days of runs.
- D. Use the Schedule failures page in the monitoring hub, select Configure notifications, choose an item, and add recipients.
D is correct.
Explanation: The Schedule failures page in the monitoring hub lets you configure email alerts for scheduled item failures across the workspace from one place. You need at least the Contributor role in the workspace.
A is incorrect: Activator can send alerts, but it requires a separate rule for each source. It is not the centrally managed email option for scheduled failures.
B is incorrect: Workspace monitoring stores diagnostic logs for querying with KQL or SQL. It does not configure failure notifications, and a manual daily query does not send alerts.
C is incorrect: Historical runs show up to 30 days of run history for one item. They do not send notifications and require you to open each item.
Q09 - Question
A data pipeline copies data, runs a Dataflow Gen2, and refreshes a semantic model. One activity in the pipeline is optional. The pipeline shows a status of Succeeded, but analysts report that some expected data is missing from a report. What should you do first?
Domain: Monitor and optimize an analytics solution (30–35%) Type: Single choice
- A. Treat the pipeline as healthy because the overall status is Succeeded.
- B. Review the status of each individual activity in the pipeline run.
- C. Check only the semantic model refresh, because it is the last step in the pipeline.
- D. Wait for the next scheduled run and compare the overall pipeline status again.
B is correct.
Explanation: A pipeline can show as succeeded even if an optional activity failed. Review the status of each activity in the pipeline to see the full picture and find the activity that caused the missing data.
A is incorrect: The overall pipeline status does not show whether an optional activity failed, so it can hide the problem.
C is incorrect: The pipeline contains several activities with dependencies. Checking only the last step can miss a failed earlier activity.
D is incorrect: Waiting delays the investigation. Analysts already report missing data, and the overall status will not show activity-level failures.
Q10 - Question
A data engineering team wants to query raw diagnostic logs and metrics from Fabric items by using KQL. They also want to correlate failures across pipelines and notebooks and build custom Real-Time dashboards. Which action meets these requirements?
Domain: Monitor and optimize an analytics solution (30–35%) Type: Single choice
- A. Enable workspace monitoring in the workspace settings on the Monitoring tab.
- B. Filter the monitoring hub by item type and add the Duration column.
- C. Select Historical runs for a single pipeline.
- D. Configure notifications on the Schedule failures page.
A is correct.
Explanation: When you enable workspace monitoring, Fabric creates an Eventhouse database in the workspace that collects diagnostic logs and metrics. You can query the raw log data with KQL or SQL, correlate failures across items, and build custom Real-Time dashboards. Workspace users need at least the Contributor role to access the data, and the data is retained for 30 days.
B is incorrect: Filters and column options change the view of recent activity in the monitoring hub. They do not store raw logs that you can query with KQL.
C is incorrect: Historical runs show up to 30 days of run history for one item. They do not support correlation across pipelines and notebooks.
D is incorrect: The Schedule failures page configures email alerts for scheduled item failures. It does not provide queryable log data.