Free Microsoft Azure DP-300 Actual Exam Questions

B works since RESTORE isn’t supported on managed instances anyway.

I’d pick option B as well. Since managed instances don’t support RESTORE like on regular SQL Server, the way to restore is by creating a new database as a copy of a backup or existing database. The CREATE DATABASE AS COPY OF syntax fits this scenario perfectly because it’s designed for managed instances, letting you basically clone the DB without needing RESTORE commands. Definitely not A or C since those use RESTORE which isn’t supported here.

B Using % is the standard way to switch languages in a Databricks notebook, like %r for R or %sql for SQL. The other options don’t match how Databricks handles multi-language cells.

Probably B here too since % is used as a magic command prefix to switch between languages, which matches the question about R, Scala, and SQL in one notebook.

Geo-redundant is obvious, and 7 days looks like a default retention period.

Looks like the backup redundancy is Geo-redundant for sure from the image.

Maybe D, since it might give stats on how queries are interacting with the index during execution. C seems focused on locks, but D could show actual operations causing delays.

Good point on C showing detailed operational data. I’d add B isn’t enough since it just tracks how often indexes are used, not what’s happening during queries. C is the only one with in-depth lock and latch info, so C.

I think using Azure Monitor makes sense since it directly supports SQL DB metrics collection. For streaming, if the question doesn’t say anything about transforming or analyzing the data on the fly, Event Hubs alone should be enough to just capture and stream the metrics as they come. Stream Analytics seems like overkill here unless there’s a clear need for real-time querying or filtering before sending the data downstream. So Azure Monitor plus Event Hubs feels like the simplest, most fitting combo.

Azure Monitor is the go-to for SQL DB metrics. For streaming, Event Hubs fits since it handles large data streams without added processing, unlike Stream Analytics which is more for real-time query needs.

Are we assuming the storage account is general-purpose v2 or Blob Storage?

B, because the secondary replica needs the database in a restoring state to join the AG.

Option B but this question feels kinda vague, anyone else?

A imo, because you usually disable purge protection before changing keys for encryption.

It’s D

D, since compression targets indexes directly, not the entire table here.

I think C can be ruled out because it talks about altering the table with a data compression option, which might not apply here. D makes more sense since we’re focusing on the clustered index itself.

Start with creating the resource pool to set limits, then workload group, classifier function, and enable Resource Governor.

I think the key is recognizing that you first define the resource pool with the CPU and memory caps. Next, you create a workload group tied to that pool to manage session allocation. Then, the classifier function comes in, which checks incoming sessions (like from App1) and routes them to the right workload group based on conditions like app name. Finally, enabling Resource Governor applies all these settings. The step order matters because you can't classify sessions without having groups and pools set up first.

I’m with those saying the scan points to a missing index on OrderDate alone. Adding other columns might help, but the biggest gain is from eliminating that scan first. Also, the sort looks like a Top N Sort since it’s just taking the first few results after ordering. If we fix the scan, the sort will become cheaper too. So picking B for the missing index and A for the sort type makes sense here.

The query’s slow because it’s doing a scan, so the right missing index is probably on OrderDate alone (B). The sort operation is likely a Top N Sort (A) since it’s limiting results. That fits the plan shape.

If the test is on SQL Server, disabling automatic plan correction might involve trace flags or certain database scoped configurations, so options mentioning those could be right. Otherwise, options about Oracle or MySQL won’t apply.

I’d rule out any options that mention “last known good plan” caching or forcing a plan freeze. Usually, you want to disable plan freezing or anything like automatic plan correction to stop it from reverting. One trap could be picking something that only updates stats but doesn’t prevent reversion. The key is to look for the setting or command that keeps the optimizer from switching back automatically during your test.

Agree, start with cluster diagnostic logs, then streaming metrics, finally send to Azure Monitor.

First thing’s to enable diagnostic settings on the cluster to catch Spark events. After that, configure Structured Streaming metrics tracking, and lastly, send all those logs and metrics to Azure Monitor.

I agree with saying No for forced plan since it’s off in the config. Also, the statement about automatic index creation and dropping should be Yes because they’re clearly enabled. The question is pretty straightforward if you just focus on what’s toggled on or off in the settings shown. The confusion probably comes from mixing up current versus recommended states, but from what I see here, it’s definitely the current settings, so trust those for your answers.

I’d say No for the forced plan statement because it’s set to off in the current config. So even if a regression is detected, it won’t automatically force a plan. The other tuning options are clearly enabled though.

I think option D for PARTITION BY makes sense since yearly partitions help delete old data fast by dropping partitions. For DISTRIBUTION, C looks better because HASH on a key like CustomerID spreads data evenly.

Is there any info on the expected volume of data or how often queries will filter on date? That could affect the choice between hash and range distribution here.