Home / DE Professional practice test / Monitoring and Alerting

Free · 8 questions with explanations

Monitoring and Alerting: Databricks Data Engineer Professional Practice Questions

Exam-style questions on Monitoring and Alerting. Pick your answer, then open the explanation to see why it's right — and why the other options are wrong.

Updated November 16, 2026

1 Monitoring and Alerting Hard

A teammate is testing a new SQL Alert definition using the alert preview UI, but the underlying query depends on a parameter. They notice the preview cannot be run as expected. Which statement best matches current SQL Alert preview behavior?

  1. AAlert preview supports parameters only if the parameter is defined as a dashboard widget.
  2. BAlert preview does not support parameters, so parameterized queries cannot be previewed directly.
  3. CAlert preview supports parameters only when the query is executed from a notebook.
  4. DAlert preview requires enabling serverless compute; parameters are otherwise disabled.
Show answer & explanation

Correct answer: B

WHY B: Databricks documents a limitation in SQL Alerts preview: it does not support parameters. That means parameterized queries cannot be validated via preview in the usual way without refactoring/removing parameters or testing the query separately. WHY NOT A/C/D: These are not documented conditions for enabling parameters in alert preview.

2 Monitoring and Alerting Easy

Your team wants a lightweight script in CI that checks whether a specific Lakeflow Job run failed overnight and posts a summary to Slack. Which statement best describes how the Databricks CLI relates to the REST APIs for this use case?

  1. AThe Databricks CLI is a front end for dbutils; its jobs and pipelines commands run on an attached cluster.
  2. BThe Databricks CLI speaks Spark Connect over gRPC; its jobs and pipelines commands bypass the REST APIs.
  3. CThe Databricks CLI wraps the Workspace API for files and repos; jobs and pipelines need the REST APIs directly.
  4. DThe Databricks CLI is a wrapper over the REST APIs; its jobs and pipelines commands call the same endpoints.
Show answer & explanation

Correct answer: D

WHY D: Databricks documents that the CLI wraps the REST API and exposes commands that correspond to API operations (for example databricks jobs get-run maps to the Jobs API runs/get endpoint). That makes it suitable for automation, including monitoring and inspection flows such as checking an overnight Lakeflow Job run from CI, without requiring UI interaction. WHY NOT A: This confuses the CLI with dbutils, the notebook utility library. The CLI is a local tool that calls the workspace over HTTPS; it is not restricted to running inside notebooks and needs no attached cluster. WHY NOT B: This confuses the CLI with Databricks Connect, which runs Spark code remotely over Spark Connect (gRPC). The CLI does not bypass the REST APIs; it is built on them, so UI scraping or another protocol is unnecessary for programmatic workflows. WHY NOT C: The CLI is not limited to workspace files and repos. The Jobs API and Pipelines API are part of the workspace REST API surface and can be used programmatically either directly or via the CLI's jobs and pipelines command groups.

3 Monitoring and Alerting Medium

You are the owner of a Unity Catalog-enabled Lakeflow Spark Declarative Pipeline and want to query its event log directly in SQL without publishing the event log table name. Which query pattern is supported by default?

  1. ASELECT * FROM event_log(<pipelineId>)
  2. BSELECT * FROM system.event_log(<pipelineId>)
  3. CSELECT * FROM pipeline_event_log(<pipelineId>)
  4. DSELECT * FROM live.event_log(<pipelineId>)
Show answer & explanation

Correct answer: A

WHY A: Databricks documents a function-style access pattern to query the hidden event log table as the pipeline owner: SELECT * FROM event_log(<pipelineId>). By default the underlying hidden table name is formatted like event_log_{pipeline_id}. WHY NOT B: The event log is not queried via the system catalog in that way; it is associated with the pipeline's own event log table/function. WHY NOT C: pipeline_event_log is not a documented function; the table-valued function is named event_log. WHY NOT D: live was the legacy keyword for referencing datasets inside pipeline source code; it is not a schema that exposes an event log function, so this is not a documented query interface.

4 Monitoring and Alerting Medium

You configure a job-level notification for Failure. A task fails but is retried and eventually succeeds, and you do not receive a failure notification. What is the most accurate explanation?

  1. AJob-level failure notifications are not sent when failed tasks are retried by the job.
  2. BJob-level failure notifications are sent for notebook tasks, not for JAR or Python tasks.
  3. CJob-level failure notifications are sent once three different tasks have failed in a run.
  4. DJob-level failure notifications need a webhook destination, as email is not supported.
Show answer & explanation

Correct answer: A

WHY A: The Databricks job notifications documentation states that job-level notifications aren't sent when failed tasks are retried. To be notified after each failed task attempt, configure task-level notifications instead of relying on job-level notifications. WHY NOT B: Notifications apply broadly to jobs and tasks of any type, not just notebook tasks. WHY NOT C: There is no 'three task failures' threshold in the notification behavior. WHY NOT D: Email notifications are supported; webhooks/system destinations are optional alternatives.

5 Monitoring and Alerting Easy

You create a SQL Alert that checks whether the count of NULLs in a key column exceeds a threshold. After the scheduled run, the alert shows a status of TRIGGERED. What does TRIGGERED indicate?

  1. AThe condition evaluated to false and the metric is within limits.
  2. BThe alert query failed to run for lack of warehouse permissions.
  3. CThe alert query succeeded and the condition evaluated to true.
  4. DThe alert was paused, so the scheduled evaluation did not execute.
Show answer & explanation

Correct answer: C

WHY C: In SQL Alerts, TRIGGERED indicates the alert condition was met (the condition evaluated to true) on the most recent evaluation. WHY NOT B: Query failures are represented as ERROR, not TRIGGERED. WHY NOT A: A condition that evaluates to false maps to OK. WHY NOT D: Paused alerts generally do not execute; TRIGGERED is a result state that comes from an evaluation that did run.

6 Monitoring and Alerting Easy

You need to build a chargeback dashboard that shows billable usage (DBUs) by day and identifies the user or run-as identity associated with the usage. Your workspace uses Unity Catalog. Which approach best aligns with Databricks observability best practices?

  1. AQuery the billing usage system tables in the system catalog, with USE CATALOG and SELECT granted as needed.
  2. BParse the cluster driver logs for usage, as system tables are meant for internal Databricks troubleshooting.
  3. CExport audit and billing data to an external database, as Databricks has no built-in usage and audit tables.
  4. DUse Delta Sharing to share the metastore to yourself, then query usage from the recipient workspace.
Show answer & explanation

Correct answer: A

WHY A: Databricks provides system tables in the system catalog (Unity Catalog) for observability use cases such as billable usage, auditing, and workload monitoring, and they are the intended source for dashboards and analyses. Access is governed via Unity Catalog privileges (for example USE CATALOG and SELECT on the relevant schemas/tables). WHY NOT B: Driver logs are useful for debugging, but they are not the primary structured source for organization-wide billing and audit analytics. System tables are a customer-facing feature, not an internal Databricks tool. WHY NOT C: You can export the data if you want, but Databricks does provide built-in system tables for usage and auditing, so an external database is not required. WHY NOT D: Delta Sharing is for sharing datasets across parties; it is not the standard mechanism for querying your own workspace usage and audit data.

7 Monitoring and Alerting Easy

You want to share detailed query execution information with a teammate to help debug a performance regression. You both have access to the workspace, but your teammate wasn’t present when you ran the query. What is the best way to share the detailed query profile information according to the Query Profile UI capabilities?

  1. ADownload the query profile JSON from the Query Profile UI and share that file with your teammate.
  2. BSend your teammate screenshots; the Query Profile UI has no export or download of profile details.
  3. CShare the Spark UI URL with your teammate; Query Profile details are viewable in the Spark UI alone.
  4. DShare EXPLAIN FORMATTED output with your teammate; it carries the same detail as the Query Profile.
Show answer & explanation

Correct answer: A

WHY A: The Query Profile UI supports downloading the query profile as JSON, which is the most faithful way to share the detailed operator graph and metrics with someone else. WHY NOT B: Screenshots lose structured detail, and the UI does provide a JSON download option. WHY NOT C: Spark UI can help for Spark jobs, but Query Profile has its own export (JSON) and can be accessed via Query History and pipeline contexts. WHY NOT D: EXPLAIN is useful, but it does not include the runtime metrics and operator timing details that the query profile captures.

8 Monitoring and Alerting Easy

A data engineer cannot query system.query- and system.billing-related system tables from a notebook. They get a permissions error even though they can query normal Unity Catalog tables. What is the most likely missing requirement?

  1. AThey need to attach the notebook to an all-purpose cluster, as job clusters are blocked from system tables.
  2. BThey need Unity Catalog privileges on the system catalog and SELECT on the relevant system schemas.
  3. CThey need to enable Change Data Feed (CDF) on the system tables before those tables can be read.
  4. DThey need an external location and a storage credential for the storage behind the system tables.
Show answer & explanation

Correct answer: B

WHY B: System tables live in the system catalog and are governed by Unity Catalog. A common cause of access errors is missing privileges such as USE CATALOG on system and SELECT on the specific schemas/tables, even when the user can query ordinary Unity Catalog tables. WHY NOT A: System tables are not restricted to interactive all-purpose clusters as a rule; the cluster type is not what produces a permissions error. Missing privileges are the typical blocker. WHY NOT C: Change Data Feed is a Delta table feature for your own data, not a prerequisite for reading system tables. WHY NOT D: External locations and storage credentials govern access to object storage paths. They are not how access to the system catalog is granted.

Take the full DE Professional practice test →