DP-750 · Practice set 8 of 11
Compute Libraries and Access Permissions: 10 practice questions
10 free DP-750 practice questions on Compute Libraries and Access Permissions, with an explanation for every answer. Untimed. The full mock exam and the timed version are in the app.
-
Question 1 of 10
Which permission is required to install a compute-scoped library on a cluster?
- ACAN ATTACH TO
- BCAN MANAGE
- CNO PERMISSIONS
- DCAN RESTART
Show the answer
Library installation changes cluster configuration, so it requires CAN MANAGE; the lower levels provide usage or lifecycle capabilities without configuration control.
Next → 1 / 10 -
Question 2 of 10
What is the scope of a compute-scoped library?
- AEvery cluster in the workspace
- BEvery notebook and job using the cluster
- COne user account across all workspaces
- DOne notebook session only
Show the answer
Compute scope shares the dependency across workloads attached to one cluster; notebook scope is narrower, while workspace-wide and account-wide propagation do not occur.
Next → 2 / 10 -
Question 3 of 10
Which permission lets an operator terminate, start, and restart compute without editing its configuration?
- AREAD VOLUME
- BNO PERMISSIONS
- CCAN ATTACH TO
- DCAN RESTART
Show the answer
CAN RESTART adds lifecycle control to the attach capabilities but stops short of the configuration authority granted by CAN MANAGE.
Next → 3 / 10 -
Question 4 of 10
Why should a production PyPI installation specify an exact package version?
- ATo convert the package into a Java JAR
- BTo bypass every Standard-mode allowlist approval check
- CTo grant CAN MANAGE to the installer
- DTo make repeated installations reproducible
Show the answer
Version pinning prevents a later cluster start from silently selecting a newer package; it does not alter package type, security approval, or compute permissions.
Next → 4 / 10 -
Question 5 of 10
What additional access is needed when a library file is stored in a Unity Catalog volume?
- ACREATE CATALOG on the metastore
- BCAN RESTART on the volume
- CWRITE VOLUME on every volume in the catalog
- DREAD VOLUME on that volume
Show the answer
The installer identity must be able to read the library file from its volume; compute configuration authority and allowlist approval remain separate requirements.
Next → 5 / 10 -
Keep the ones you got wrong
In the app, every question you miss comes back exactly when you’re about to forget it.
-
Question 6 of 10
What capability does CAN ATTACH TO grant without granting lifecycle control?
- AResize the cluster and install shared libraries for all users
- BAttach notebooks and view Spark UI and compute metrics
- CModify permissions for other users
- DTerminate and restart the compute resource
Show the answer
CAN ATTACH TO is the usage tier: it supports attached work and monitoring, while lifecycle belongs to CAN RESTART and configuration belongs to CAN MANAGE.
Next → 6 / 10 -
Question 7 of 10
When is an init script preferable to a compute-scoped library?
- AWhen a dependency should affect only one notebook session
- BWhen startup needs system-level configuration that a library cannot provide
- CWhen the installer lacks permission to modify the cluster
- DWhenever a Python wheel is available from PyPI
Show the answer
Init scripts are the advanced fallback for startup-level system configuration; cluster libraries are preferred for ordinary packages and still require appropriate permissions.
Next → 7 / 10 -
Question 8 of 10
Two teams share a cluster but require incompatible versions of the same Python package. Each version is needed in only one team's notebook session, and the shared cluster configuration must remain unchanged. Which approach best fits?
- AUse one unpinned compute-scoped PyPI package
- BPut both wheels in a volume and install both cluster-wide
- CUse notebook-scoped installations for the separate sessions
- DInstall both versions as compute-scoped libraries
Show the answer
Notebook scope isolates the conflicting dependencies to their sessions while preserving the shared cluster environment; every compute-scoped alternative affects all attached workloads.
Next → 8 / 10 -
Question 9 of 10
A Standard-mode cluster must install an approved JAR stored in a Unity Catalog volume. The operator can manage the cluster, but installation still fails an access check. Which complete condition should be verified?
- AThe path is allowlisted, which automatically grants file access
- BThe path is allowlisted and the installer has READ VOLUME
- CThe JAR is unpinned and the cluster has NO PERMISSIONS
- DThe operator has CAN RESTART and the volume has WRITE VOLUME
Show the answer
Standard-mode approval and data authorization are independent: the path must pass the allowlist and the installer must be able to read the volume, in addition to managing the cluster.
Next → 9 / 10 -
Question 10 of 10
An analyst must run notebooks and inspect Spark execution. A team lead must also recover a stopped cluster, while only the platform owner may change configuration or libraries. Which least-privilege mapping is correct?
- AAnalyst: NO PERMISSIONS; lead: CAN ATTACH TO; owner: CAN RESTART
- BAnalyst: CAN ATTACH TO; lead: CAN RESTART; owner: CAN MANAGE
- CAnalyst: CAN RESTART; lead: CAN MANAGE; owner: CAN ATTACH TO
- DAnalyst: CAN MANAGE; lead: CAN ATTACH TO; owner: CAN RESTART
Show the answer
The tiers build from usage to lifecycle to configuration, so this mapping gives each role exactly the required capability without elevating the analyst or lead.
Next → 10 / 10 -
You’ve finished this set
That’s 10 questions on Compute Libraries and Access Permissions. In the app the ones you miss come back exactly when you’re about to forget them.
The whole course, on your phone
Lessons you can read, audio you can listen to on the way to work, and practice that remembers what you got wrong.