DP-750 Practice Questions — Page 10

Question 91

An analyst reports that a dashboard query against a serverless SQL warehouse returns instantly but "doesn't reflect the new data we just merged." You investigate:

- The underlying Delta table was updated by a `MERGE` 10 minutes ago.

- When the analyst re-runs the exact same query text, results are still instant and you cannot open a **query profile** for that run.

- You need to (1) force the query to actually execute so the new data is returned and (2) be able to open the query profile to analyze the execution.

Which action both forces fresh execution and makes a query profile available?

A.Make a trivial change to the query (for example, change or remove the `LIMIT`) so it bypasses the result cache; the next run executes and a query profile becomes available.
B.Run `OPTIMIZE` on the Delta table; this invalidates Azure Databricks query caching and forces re-execution with a profile.
C.Restart the SQL warehouse; the remote result cache is tied to the warehouse and is cleared on restart, so the next run re-executes.
D.Increase `spark.sql.shuffle.partitions`; this disables result caching for the session and produces a query profile.

Question 92

Open question ↗

A high-ingest Delta table (`sales.bronze_events`) receives thousands of small files per hour from streaming writes and frequent `MERGE` operations. Query latency has grown and storage cost is rising from obsolete pre-compaction files. You must design a weekly maintenance run that (a) improves read performance by compacting and clustering files, then (b) reclaims storage from files no longer referenced, while preserving the standard 7-day time-travel window.

Order the maintenance steps into the correct end-to-end sequence.

```mermaid

flowchart LR

subgraph Tiles

T1["DESCRIBE HISTORY / DESCRIBE DETAIL<br/>(baseline file count and state)"]

T2["OPTIMIZE sales.bronze_events<br/>(bin-pack small files; cluster layout)"]

T3["VACUUM sales.bronze_events<br/>(default 7-day / 168-hour retention)"]

T4["DESCRIBE DETAIL<br/>(verify fewer, larger files)"]

end

subgraph Sequence

Slot1["Step 1"]

Slot2["Step 2"]

Slot3["Step 3"]

Slot4["Step 4"]

end

T1 -.-> Slot1

T2 -.-> Slot2

T3 -.-> Slot3

T4 -.-> Slot4

```

Place each tile into the correct step slot (Step 1 → Step 4).

Question 93

Open question ↗

Your security and operations teams want all Azure Databricks **workspace audit/diagnostic logs** (cluster events, jobs, DBFS, Unity Catalog, SQL, and so on) streamed into a central **Log Analytics workspace** in Azure Monitor so they can run Kusto (KQL) queries and build workbooks. The Databricks workspace runs the Premium plan.

The Databricks workspace is deployed into your own virtual network (VNet injection), and the network security groups (NSGs) are configured to **deny all outbound traffic that is not required by Azure Databricks**.

What must you configure to deliver these logs to Log Analytics?

A.In the Azure portal, open the Azure Databricks service resource, create a **Diagnostic setting** that selects the log categories and sends them to the Log Analytics workspace; and add an outbound NSG rule allowing the **AzureMonitor** service tag.
B.Install the Log Analytics agent on every cluster driver node and point it at the workspace key, then query the `DatabricksClusters` table.
C.Enable Predictive Optimization on Unity Catalog; it automatically forwards audit logs to Azure Monitor.
D.Create an Event Grid subscription on the storage account that backs DBFS; Event Grid relays the logs into Log Analytics.

Question 94

Open question ↗

Your Azure Databricks workspace already streams diagnostic logs to a Log Analytics workspace, and platform metrics are available for the Databricks resource. You must configure two Azure Monitor alerts:

1. A **near-real-time** alert that fires when a platform metric (for example, a workspace-level metric value) crosses a static threshold, evaluated at a fixed frequency with a defined aggregation window.

2. An alert based on a saved **Kusto (KQL) query** over the Databricks diagnostic logs in the Log Analytics workspace, run on a schedule, that fires when the number of returned results exceeds 0 (for example, repeated job-failure rows).

You also need every fired alert to send email and SMS notifications to the on-call rotation.

Which **three** statements correctly describe how to build this in Azure Monitor? (Choose THREE.)

A.Requirement 1 is a **metric alert rule** — it evaluates a metric value against a threshold using an evaluation frequency and an aggregation window (for example, `az monitor metrics alert create --condition "avg <metric> > <threshold>"`).
B.Requirement 2 is a **log search alert rule** (scheduled query rule) — it runs the KQL query on a schedule and fires based on the result/row count (for example, `az monitor scheduled-query create`).
C.Both alerts use an **action group** to define what happens when the alert fires, such as sending email and SMS notifications.
D.Requirement 1 must be implemented as a log alert because Azure Databricks platform metrics cannot be used by metric alert rules.
E.Requirement 2 must be implemented as a metric alert because KQL queries can only run inside metric alert rules.
F.Notifications are configured directly on each alert rule's threshold and do not use action groups.

Question 95

Open question ↗

A team wants to aggressively reclaim storage on a busy production Delta table that has long-running jobs writing to it throughout the day. An engineer proposes the following maintenance step, to be run during business hours:

```sql

-- Disable the safety check, then vacuum with zero retention

SET spark.databricks.delta.retentionDurationCheck.enabled = false;

VACUUM prod.fact_transactions RETAIN 0 HOURS;

```

**Proposed solution:** Running `VACUUM` with a 0-hour retention on this table risks breaking time travel and can corrupt results for concurrent readers and in-flight writers, so it should not be run this way on a busy production table.

Does this solution meet the goal?

A.Yes
B.No