Exam Bundle

That observation is a valid indicator for “Only a few tasks take much longer than others in one shuffle stage” when using Examine Spark task duration distributions and input/shuffle sizes. The scenario instead tests “A job has many tiny tasks and scheduling overhead dominates”; the relevant result is The stage has excessive short tasks relative to meaningful computation.

This signal belongs to the adjacent case “A task fails with a missing dependency after a runtime upgrade,” whose remedy is Inspect driver/executor stack traces for ClassNotFound or ImportError. The observed result for the present requirement is The stage has excessive short tasks relative to meaningful computation.

The suspected stage shows substantial disk spill and associated slowdown would demonstrate the separate action Inspect Spark UI spill metrics and task timelines, which addresses “A stage repeatedly spills large intermediates to disk.” It does not verify the requirement “A job has many tiny tasks and scheduling overhead dominates” following Review the Spark UI task count and scheduling timeline.

Publish a validated row-count metric and alert on the threshold is appropriate when “A job succeeded but produced no records and needs a semantic data-quality alert”; it would be verified by The empty output triggers quality alert despite success run state. It does not directly satisfy “A daily row-count threshold breach should notify the data steward automatically,” whose relevant acceptance evidence is An alert notification fires when the result crosses the configured threshold.

Inspect SQL query profile and pruning/scan metrics belongs to the separate case “A lakehouse table read suddenly becomes slow after a data layout change,” where Changed scan and shuffle costs explain the regression would be expected. That differs from the condition in this stem: “A daily row-count threshold breach should notify the data steward automatically.”

Define explicit null/no-data handling in the alert query and condition belongs to the separate case “The alert query produces no row because ingestion failed and this must be distinguished from zero,” where The failure-to-ingest case raises the intended no-data signal would be expected. That differs from the condition in this stem: “A daily row-count threshold breach should notify the data steward automatically.”

The result includes the desired failed runs without missing later pages pertains to a different control (Use the Jobs API run listing with relevant filters and pagination) and a different need (Automation needs to list recent failed runs for a configured job). It cannot establish that Create a Lakehouse/SQL alert on the scheduled aggregate query met the stated goal.

This signal belongs to the adjacent case “A noisy anomaly threshold raises hundreds of repeated false alarms,” whose remedy is Tune threshold, schedule and deduplication policy based on baseline metrics. The observed result for the present requirement is An alert notification fires when the result crosses the configured threshold.

This signal belongs to the adjacent case “Late-data drops rise after a watermark configuration change,” whose remedy is Inspect watermark position and dropped-late-record progress counters. The observed result for the present requirement is An alert notification fires when the result crosses the configured threshold.

Publish a validated row-count metric and alert on the threshold is appropriate when “A job succeeded but produced no records and needs a semantic data-quality alert”; it would be verified by The empty output triggers quality alert despite success run state. It does not directly satisfy “The alert query produces no row because ingestion failed and this must be distinguished from zero,” whose relevant acceptance evidence is The failure-to-ingest case raises the intended no-data signal.

The alternative Add supported expectations and inspect corresponding event metrics has a distinct purpose—A new pipeline produces no useful quality metrics for rejected data. It would not establish the required outcome (The alert query produces no row because ingestion failed and this must be distinguished from zero); the necessary verification concerns The failure-to-ingest case raises the intended no-data signal.

This action solves a different issue: “A noisy anomaly threshold raises hundreds of repeated false alarms.” Its normal signal is The same benign variance no longer floods the channel while true anomalies alert. Here the requested behavior is “The alert query produces no row because ingestion failed and this must be distinguished from zero,” so that signal is not sufficient.

The log identifies the failing flow and recorded exception would demonstrate the separate action Inspect pipeline event-log ERROR events for the affected update ID, which addresses “An update failed and engineers need the exact failure text and originating flow.” It does not verify the requirement “The alert query produces no row because ingestion failed and this must be distinguished from zero” following Define explicit null/no-data handling in the alert query and condition.

This signal belongs to the adjacent case “A daily row-count threshold breach should notify the data steward automatically,” whose remedy is Create a Lakehouse/SQL alert on the scheduled aggregate query. The observed result for the present requirement is The failure-to-ingest case raises the intended no-data signal.

Only the intended failed task/path reruns and downstream outputs are correct pertains to a different control (Use supported Jobs task repair or selective pipeline retry where applicable) and a different need (A pipeline failure needs targeted replay without rerunning unaffected upstream tasks). It cannot establish that Define explicit null/no-data handling in the alert query and condition met the stated goal.

Inspect driver/executor stack traces for ClassNotFound or ImportError belongs to the separate case “A task fails with a missing dependency after a runtime upgrade,” where The logs identify the missing package or incompatible class would be expected. That differs from the condition in this stem: “An alert should evaluate only after daily ETL completes rather than on stale intermediate data.”

This action solves a different issue: “A CLI script must use a production workspace rather than a developer's default profile.” Its normal signal is CLI responses identify the correct production workspace. Here the requested behavior is “An alert should evaluate only after daily ETL completes rather than on stale intermediate data,” so that signal is not sufficient.

Create a Lakehouse/SQL alert on the scheduled aggregate query belongs to the separate case “A daily row-count threshold breach should notify the data steward automatically,” where An alert notification fires when the result crosses the configured threshold would be expected. That differs from the condition in this stem: “An alert should evaluate only after daily ETL completes rather than on stale intermediate data.”

That observation is a valid indicator for “Late-data drops rise after a watermark configuration change” when using Inspect watermark position and dropped-late-record progress counters. The scenario instead tests “An alert should evaluate only after daily ETL completes rather than on stale intermediate data”; the relevant result is The alert reflects the completed day's data, not an earlier snapshot.

This signal belongs to the adjacent case “An engineer needs a persistent queryable audit of pipeline events rather than dashboard screenshots,” whose remedy is Expose or query the supported event-log table through SQL. The observed result for the present requirement is The alert reflects the completed day's data, not an earlier snapshot.

The designated on-call destination receives the test alert pertains to a different control (Configure the intended alert destination or notification channel) and a different need (A failing alert must route to a maintained distribution list without hardcoded personal addresses). It cannot establish that Schedule alert evaluation after the ETL dependency succeeds met the stated goal.

Define explicit null/no-data handling in the alert query and condition is appropriate when “The alert query produces no row because ingestion failed and this must be distinguished from zero”; it would be verified by The failure-to-ingest case raises the intended no-data signal. It does not directly satisfy “A failing alert must route to a maintained distribution list without hardcoded personal addresses,” whose relevant acceptance evidence is The designated on-call destination receives the test alert.

The alternative Publish a governed aggregate metric view or anonymized report has a distinct purpose—An account's daily cost should be visualized without disclosing sensitive team names to viewers. It would not establish the required outcome (A failing alert must route to a maintained distribution list without hardcoded personal addresses); the necessary verification concerns The designated on-call destination receives the test alert.

The alternative Configure a duration-based warning or supported monitoring alert has a distinct purpose—A job unexpectedly exceeds its runtime SLA without failing. It would not establish the required outcome (A failing alert must route to a maintained distribution list without hardcoded personal addresses); the necessary verification concerns The designated on-call destination receives the test alert.

An alert notification fires when the result crosses the configured threshold pertains to a different control (Create a Lakehouse/SQL alert on the scheduled aggregate query) and a different need (A daily row-count threshold breach should notify the data steward automatically). It cannot establish that Configure the intended alert destination or notification channel met the stated goal.

The result includes the desired failed runs without missing later pages pertains to a different control (Use the Jobs API run listing with relevant filters and pagination) and a different need (Automation needs to list recent failed runs for a configured job). It cannot establish that Configure the intended alert destination or notification channel met the stated goal.

This signal belongs to the adjacent case “A job unexpectedly exceeds its runtime SLA without failing,” whose remedy is Configure a duration-based warning or supported monitoring alert. The observed result for the present requirement is The designated on-call destination receives the test alert.

The alternative Define explicit null/no-data handling in the alert query and condition has a distinct purpose—The alert query produces no row because ingestion failed and this must be distinguished from zero. It would not establish the required outcome (A noisy anomaly threshold raises hundreds of repeated false alarms); the necessary verification concerns The same benign variance no longer floods the channel while true anomalies alert.

This action solves a different issue: “A SQL query repeatedly scans a large table even for a narrow date predicate.” Its normal signal is The scan plan shows poor pruning and the responsible predicate/layout. Here the requested behavior is “A noisy anomaly threshold raises hundreds of repeated false alarms,” so that signal is not sufficient.

Inspect streaming state operator metrics and batch execution durations belongs to the separate case “A microbatch duration spikes after an increase in state cardinality,” where State rows and processing time increase together for affected batches would be expected. That differs from the condition in this stem: “A noisy anomaly threshold raises hundreds of repeated false alarms.”

An alert notification fires when the result crosses the configured threshold would demonstrate the separate action Create a Lakehouse/SQL alert on the scheduled aggregate query, which addresses “A daily row-count threshold breach should notify the data steward automatically.” It does not verify the requirement “A noisy anomaly threshold raises hundreds of repeated false alarms” following Tune threshold, schedule and deduplication policy based on baseline metrics.

This signal belongs to the adjacent case “A cost anomaly occurs during a period of high workload retries,” whose remedy is Correlate spend time series with Jobs run and retry records. The observed result for the present requirement is The same benign variance no longer floods the channel while true anomalies alert.

That observation is a valid indicator for “A streaming sink occasionally retries the same batch after transient errors” when using Correlate batch IDs and sink commit logs across retries. The scenario instead tests “A noisy anomaly threshold raises hundreds of repeated false alarms”; the relevant result is The same benign variance no longer floods the channel while true anomalies alert.