Monitor and optimize an analytics solution

DP-700 · 66 questions

• A regional archives supervisor wants a single place each morning to check the status of last night's digitization pipeline runs, Dataflow Gen2 refreshes, and notebook jobs across every branch workspace before opening the reading room. Which Fabric monitoring surface should they check first?
• The overnight digitization pipeline for a museum branch has started finishing later each week, and gallery staff need it done before the doors open. Which Fabric monitoring view lets an engineer see how long each activity inside that pipeline run took, so the slow stage can be identified?
• A conservation team wants to be notified automatically if the humidity-sensor stream from a rare-manuscripts vault stops sending readings for more than 15 minutes, without anyone having to watch a dashboard. Which Fabric capability is purpose-built for defining this kind of condition-and-action rule against live data?
• An engineer suspects that a gallery's vibration-sensor Eventstream has stopped receiving events, but no alert has fired yet. Which built-in Eventstream capability lets them confirm, visually, whether data is currently flowing through the stream?
• A data engineer's PySpark notebook that cleans OCR output from digitized oral-history transcripts has been running far longer than usual overnight. Which Fabric monitoring surface lets them inspect the notebook's Spark application, including its stages and executors, to see where time is being spent?
• A Dataflow Gen2 that denormalizes and cleans loan-tracking records from branch collection-management systems runs nightly, and the team wants to see whether recent runs succeeded and how long each one took, without opening the Dataflow's editor. Where should they look?
• A branch's ticketing semantic model refreshes nightly so morning attendance dashboards are current, but stakeholders report that yesterday's numbers looked stale. Which monitoring step should the team take first to determine whether the scheduled refresh actually ran and succeeded?
• Leadership wants to be notified by email the moment the nightly ticketing semantic model refresh fails, rather than discovering it the next morning when the dashboard looks wrong. Which approach lets them configure that kind of automatic, condition-based notification?
• With dozens of branch workspaces each running their own nightly pipelines, dataflows, and notebooks, an operations lead wants to quickly narrow the Monitoring hub down to only the runs that failed in the last 24 hours across every workspace they can access. Which capability of the Monitoring hub supports this?
• A conservation team ingests vault humidity and light readings into an Eventhouse continuously, and they want to check whether any recent batches of telemetry failed to ingest into the KQL database, rather than waiting for a downstream query to come up short. Where does this kind of ingestion status information live?
• The overnight digitization pipeline across several branches has recently started running slower, and an engineer suspects the shared Fabric capacity is being throttled by another heavy workload rather than the pipeline itself having gotten worse. Which monitoring tool should they check to confirm whether capacity throttling is occurring?
• A workspace admin is setting up automated monitoring so that if a branch's visitor-count Eventstream reports a value above a defined threshold, a Teams message is sent to the operations channel automatically. Conceptually, what two things must be defined together for this to work in Data Activator?
• During a busy exhibit-opening week, an engineer checking the Monitoring hub notices that a branch's nightly ingestion pipeline has been showing "In Progress" for far longer than its usual runtime. Which monitoring action lets them confirm which specific activity inside that run is still executing, before deciding whether to investigate further?
• An engineer reviewing a Spark application for last night's OCR-cleanup notebook run notices one stage took ten times longer than every other stage. Which monitoring detail inside that Spark application view would help them see whether the slow stage was caused by uneven data distribution across tasks?
• A team wants an automatic alert if a branch's gallery-temperature Eventstream reports a reading outside the safe conservation range for more than five minutes continuously, not just a single spike. Which Data Activator concept lets them require the condition to hold for a sustained period before triggering?
• Reviewing several weeks of a branch's ticketing semantic model refresh history, an analyst notices the refresh duration has been climbing steadily even though the schedule hasn't changed. What does monitoring this trend, rather than only the most recent run, make visible that a single run's status would not?
• An engineer needs to decide whether to check the Monitoring hub or a specific Dataflow Gen2's own refresh history to answer the question "did every scheduled dataflow refresh across all our branch workspaces succeed last night?" Which is the better starting point, and why?
• A branch's Dataflow Gen2 that ingests loan-tracking updates is scheduled to run once nightly, but the refresh history shows two runs starting within the same hour on several recent nights. Which monitoring habit would have surfaced this pattern earliest?
• A team wants an automatic notification whenever the daily visitor count ingested from ticketing scanners into an Eventhouse exceeds a set number, so operations can plan extra staffing. Which piece of information must be defined for a Data Activator rule to act on this specific numeric value?
• An analyst monitoring a branch's semantic model refresh history notices refreshes have grown long enough that they sometimes are still running when the museum opens and dashboards are first viewed. What does this observation, drawn purely from monitoring the refresh history, indicate the team should evaluate?
• A workspace admin has defined a Data Activator condition for a failed nightly pipeline, but wants the resulting alert to reach the operations team's Teams channel rather than a single person's inbox. What must be configured on the alert's action for this to work?
• Before galleries open each morning, an operations lead wants one habitual check that covers pipeline runs, dataflow refreshes, and notebook jobs across every branch, catching anything that failed or is still stuck running overnight, without relying on any single item's individual alert having been configured. Which practice best satisfies this?
• A branch's collection-management system starts exporting its nightly catalogue extract with two columns swapped after an unannounced vendor update, and the archive network's pipeline Copy activity maps source columns to the warehouse table by position. The next morning the pipeline run shows a green succeeded status, but curators report accession numbers landing in the condition-notes field. What should the engineer check first to confirm this is the root cause?
• A pipeline uses a ForEach activity to loop over every branch's inter-branch loan-tracking feed and load it into a Lakehouse table. Overnight the whole pipeline run fails, and the engineer needs to find which specific branch's iteration caused the failure and why. Where should they look?
• A pipeline's Notebook activity, which runs a PySpark notebook that cleans OCR text from scanned oral-history transcripts, fails. The pipeline's own run history simply shows the activity as failed with a generic message that the notebook run did not complete successfully. What should the engineer do to find the actual root cause?
• A nightly pipeline loads digitized artifact images into a Lakehouse and then runs two downstream activities that generate thumbnails from those images. One night the image-copy activity fails, and in the pipeline's monitoring canvas the two thumbnail activities appear grayed out instead of red. How should the engineer interpret this?
• A branch renames a column in the spreadsheet it uses to track object loans between museum branches, and the following night a Dataflow Gen2 that consumes that spreadsheet fails to refresh. Where should the engineer look to identify which query step broke and why?
• A Dataflow Gen2 transforms nightly visitor-ticket counts from a branch's kiosk export, converting a text column to a whole-number type. The refresh itself completes, but several rows now show error values in that column instead of numbers, because a few kiosks exported stray characters alongside the digits. What is the best way to identify exactly which source rows are causing this?
• A Dataflow Gen2 that pulls a legacy card-catalog database, reachable only through an on-premises data gateway at one regional branch, fails to refresh three nights out of five, with no changes to the query logic. On the nights it fails, the refresh history shows the failure occurring almost immediately, before any transformation step begins. What does this pattern most strongly suggest?
• A Dataflow Gen2 merges a digitized-object metadata query with a conservation-status query using a shared object ID, joining them to enrich each object's record. After a source system change, the merge step starts returning zero matches instead of the expected joined rows, even though both tables clearly still contain matching IDs when viewed separately. What is the most likely cause?
• A PySpark notebook processes a large nightly batch of OCR text extracted from scanned oral-history transcripts. One cell calls .collect() to pull the entire distributed dataset into the driver for a quick inspection, and that cell now fails on especially large batches while earlier cells that only build and cache Spark DataFrames succeed. What does this pattern point to?
• A notebook's very first cell reads from a Lakehouse table that holds newly digitized artifact metadata, and it fails immediately, reporting that the table cannot be found, even though the engineer can see the table listed in the Lakehouse explorer. The notebook that is supposed to create this table each night runs later in the same pipeline, after this one. What is the most likely root cause?
• A notebook appends a nightly batch of newly digitized-object metadata to an existing Delta table in a Lakehouse. After the upstream extraction process starts including a new provenance field, the append fails with an error indicating the incoming data's schema does not match the target table. What is happening?
• One morning a scheduled notebook fails and its run detail shows no cells executed at all; the failure occurred while the notebook was still trying to attach to a Spark session. On other mornings the same notebook runs a specific cell that then errors out partway through processing. How should an engineer route these two different failures?
• An Eventstream carrying humidity and light readings from gallery conservation sensors feeds a table in an Eventhouse. After a branch upgrades its sensor firmware, the field names in the incoming JSON payload change slightly, and the Eventhouse's ingested row count drops sharply overnight even though the sensors are still transmitting. Where should the engineer look to confirm the cause?
• An analyst runs a KQL query against an Eventhouse table holding vibration sensor readings from a vault, filtering for the last five minutes, and gets back an unexpectedly empty result. The Eventhouse itself reports as healthy, and querying for the last hour instead returns plenty of rows, just none in the most recent few minutes. What is the most likely explanation?
• An Eventhouse has an update policy that runs a function to transform raw gallery sensor readings into a cleaned, derived table used for conservation dashboards. After the raw table's column names change, the derived table stops receiving new rows, while the raw table itself keeps filling normally. What has most likely happened?
• An Eventstream carries live ticket-scanner events from a museum branch's turnstiles through a filter step and into an Eventhouse destination. Visitors keep scanning tickets all afternoon, but the destination stops receiving any new events partway through the day. What is the best way to pinpoint where in the Eventstream the data flow actually stops?
• An Eventstream includes a filter step meant to drop test events generated by a branch's turnstile hardware during maintenance, keeping only real visitor scans. After the vendor updates the turnstile firmware and renames one of the fields the filter condition checks, the filter step's live preview shows every single event being dropped, including legitimate visitor scans. What has happened?
• An Eventstream's live preview shows visitor-flow events from a branch's people-counting sensors passing successfully through every node, including the final node right before the destination. Yet the target Eventhouse table shows no new rows arriving. What does this pattern suggest?
• A scheduled T-SQL stored procedure in a Fabric Warehouse aggregates nightly visitor-ticketing totals by pulling from a view that finance recently redefined, dropping a column the procedure references. The next run fails with an error naming that specific column as invalid. What should the engineer conclude?
• A T-SQL query against a Fabric Warehouse table of donor-gift records implicitly converts a text donation-amount column to a numeric type for a summary report. The query fails with a conversion error, and the engineer confirms the vast majority of rows convert without issue. What should the engineer check first?
• A T-SQL query that has run successfully for months against a Warehouse table of loan-tracking records suddenly fails, reporting that the object cannot be found, even though the table clearly still exists and other queries the same engineer runs work fine. After a recent reorganization, tables were moved into a new schema. What is the most likely explanation?
• An analyst's T-SQL query against a Warehouse table of exhibit loan agreements occasionally times out or fails with a concurrency-related error, but only during the narrow window each night when a pipeline is bulk-loading new loan records into that same table. Running the identical query at any other time of day works without issue. What should the engineer conclude?
• A national archive's Lakehouse stores humidity, light, and vibration telemetry streamed from conservation sensors in ten vault rooms, with each sensor writing a new file every few seconds. Curators notice that queries against the sensor table have grown steadily slower even though the table's total row count is only growing at a steady, predictable rate. What is the most likely cause?
• A digitisation team writes millions of rows of scanned-document metadata into a Fabric Lakehouse table using Spark, and the table is later queried heavily through both Power BI reports and the SQL analytics endpoint. Which optimization, applied automatically when the Spark job writes the table, is specifically intended to speed up exactly this kind of downstream read?
• A museum network's Lakehouse holds a decade of visitor-flow and ticketing history across dozens of branches. Analysts almost always filter their reports to a single month or a single quarter, yet every query currently scans the full ten years of files before applying that filter. Which change to the table's physical layout would most directly reduce the amount of data scanned for these queries?
• A data engineer runs table-compaction (OPTIMIZE) on a large digitised-artifact metadata table every night, and the underlying storage size keeps growing even though the number of active files stays roughly the same after each run. What is the most likely reason, and what should the engineer do next?
• To speed up queries, an engineer partitions a branch loan-tracking table by branch code, loan date, and exact loan time, resulting in thousands of partitions each holding only a handful of rows. Query performance gets worse, not better, after this change. What is the most likely explanation?
• A nightly pipeline loads collection updates from twelve independent regional branch archives into a central Lakehouse using a single ForEach activity that processes branches one at a time. The pipeline finishes correctly every night but now runs long enough to delay the morning reporting refresh. The branches have no dependency on each other. What change would most directly reduce the pipeline's total run time?
• A pipeline currently uses a Dataflow Gen2 to move a straightforward, unmodified extract of daily exhibit-ticket sales from a source database into a Warehouse staging table — no transformation logic is applied at any step. The load is one of the slowest steps in the nightly schedule. What change is most likely to improve its performance, without changing what data ends up in the staging table?
• After a one-time bulk load adds several years of archived loan-tracking records into an existing Fabric Warehouse table, a T-SQL join query that used to return quickly now takes noticeably longer, even though no query text changed. What is the most likely fix?
• A T-SQL report against the Warehouse computes, for every branch, the number of loan requests that occurred after that branch's most recent inspection date, using a correlated subquery inside the SELECT list that re-evaluates the inspection lookup for every branch row. The query is one of the slowest in the nightly report suite. Which rewrite is most likely to improve its performance while returning the same results?
• Conservators run frequent KQL queries against an Eventhouse table holding vault humidity and vibration telemetry, almost always looking at the last 48 hours of readings, even though the table retains several years of history for compliance. Queries against the recent window are slower than conservators expect. Which Eventhouse setting is designed specifically to speed up queries like this?
• An Eventstream ingests motion and vibration events from gallery sensors across every branch. On ordinary days it keeps up comfortably, but during peak visitor events, when sensor event volume spikes sharply, downstream processing starts lagging noticeably behind real time. What change to the Eventstream is most directly aimed at handling this kind of throughput spike?
• A notebook that runs PySpark cleanup on OCR output from digitised oral-history transcripts is assigned a fixed-size Spark pool sized for the largest transcript batch the archive ever processes, even though most nightly batches are far smaller. Utilization metrics show the pool sitting mostly idle on typical nights, while capacity costs stay constantly high. What change addresses this most directly?
• A PySpark job joins a very large fact table of digitised-object records against a small lookup table of the network's dozen branch codes and names. The job spends most of its time shuffling the huge fact table across the cluster just to match it with a handful of lookup rows. Which technique is most directly aimed at eliminating that unnecessary shuffle?
• A notebook builds a cleaned DataFrame of oral-history transcript metadata through several PySpark transformation steps, then uses that same DataFrame as the input to three separate downstream aggregations. Spark re-executes the entire chain of upstream transformations from scratch for each of the three aggregations, tripling the total runtime. What change would most directly avoid that repeated recomputation?
• A PySpark job aggregates a modest-sized daily visitor-count dataset by branch and hour, but after the group-by-and-aggregate step, Spark produces thousands of tiny output partitions, most holding only a few rows, and the job spends more time on task scheduling overhead than on actual computation. What is the most likely cause?
• An engineer notices that a Lakehouse table of ticket-scan events, which suffers from the small-file problem after months of frequent streaming writes, is slow not only in Spark notebooks but also when analysts query it through the SQL analytics endpoint using plain T-SQL. Why does the same file-layout issue affect both engines?
• A dashboard repeatedly runs the same KQL summarize-and-aggregate query against a large, continuously-growing Eventhouse table of vibration-sensor readings to show a rolling hourly average per vault. Each dashboard refresh re-scans and re-aggregates the full underlying table, which is becoming slow as history accumulates. Which Eventhouse feature is designed to precompute and store this kind of recurring aggregation so refreshes don't have to reprocess raw data from scratch each time?
• An engineer who previously worked with Azure Synapse Analytics dedicated SQL pools wants to define a hash distribution key on a large Fabric Warehouse fact table to control how its rows are distributed across compute nodes, the way they used to in Synapse. What should they understand about doing this in a Fabric Warehouse?
• A nightly pipeline copies a very large volume of digitised-object metadata directly from an on-premises collection-management database into a Fabric Warehouse table using a single Copy activity, and the load consistently runs far longer than the volume alone seems to justify. What change is most likely to improve throughput for this kind of large direct load into a Warehouse?
• Rather than compacting small files after the fact with a scheduled OPTIMIZE job, an engineer wants Spark to avoid generating excessive small files in the first place when it writes results from a nightly transformation of loan-tracking records into a Lakehouse table. Which approach addresses the problem at write time rather than after the fact?
• Analysts frequently filter a large Lakehouse table of digitised-object records by object category, but the table isn't partitioned by category because category values change over time and partitioning by it would create excessive small partitions. Query performance on category-filtered searches is still poor despite the table being regularly compacted. Which additional compaction option would most directly help category-filtered queries skip irrelevant data without partitioning the table by category?
• A report queries a wide Lakehouse table of digitised-object metadata that has over a hundred columns, including large free-text conservation-note fields, but the report only ever needs five specific columns such as object ID, category, and current branch. The query currently uses SELECT and is notably slower than reports of similar row count that select only a few columns. What is the most direct explanation, and the most direct fix?