Skip to content

Data Engineer Professional · Domain 6: Ensuring Data Security and Compliance · 8% of the exam

Task 6.1: Apply least-privilege access control lists (ACLs) to secure Unity Catalog securable objects and workspace resources.

Unity Catalog privileges for each task, ownership and MANAGE, job, pipeline, warehouse and secret scope permissions, and service principals as run identities.

Study it

Sample questions

Try each one before opening the answer. Every option is explained, with the Databricks documentation page that proves it.

Question 1 · choose 1

A data engineer created the service principal etl-sp and has CAN MANAGE on a production job. When they try to set the job's Run as to etl-sp, the service principal is not offered. Security wants the smallest change that fixes this. What should be granted?

  1. AIS OWNER on the job, transferred to the engineer from its creator
  2. BThe workspace admin role to the engineer
  3. CCAN MANAGE on the job for the service principal etl-sp
  4. DThe Service Principal User role on etl-sp to the engineer
Show the answer and why
  • AIS OWNER on the job, transferred to the engineer from its creator

    Incorrect

    CAN MANAGE already allows changing Run as; the missing piece is the role on the service principal.

  • BThe workspace admin role to the engineer

    Incorrect

    Admin rights are far broader than needed for this one change.

  • CCAN MANAGE on the job for the service principal etl-sp

    Incorrect

    Permissions for the service principal on the job do not let the engineer select it as the run identity.

  • DThe Service Principal User role on etl-sp to the engineer

    Correct

    Run as can be set to a service principal on which the user has the Service Principal User role; the manager role does not include it.

Changing a job's Run as needs CAN MANAGE or IS OWNER on the job and the Service Principal User role on the service principal. The creator of a service principal becomes its manager, but managers do not inherit the user role and must assign it to themselves.

Question 2 · choose 1

An operations team must be able to start and cancel runs of a production job, but must not change its tasks or settings. Which job permission should they get?

  1. ACAN MANAGE RUN
  2. BCAN MANAGE
  3. CCAN VIEW
  4. DIS OWNER
Show the answer and why
  • ACAN MANAGE RUN

    Correct

    CAN MANAGE RUN lets users trigger and cancel job runs.

  • BCAN MANAGE

    Incorrect

    CAN MANAGE also allows editing the job definition and permissions.

  • CCAN VIEW

    Incorrect

    CAN VIEW shows run results but does not allow starting runs.

  • DIS OWNER

    Incorrect

    IS OWNER brings full control of the job, far more than running it.

Job permissions are CAN VIEW, CAN MANAGE RUN, CAN MANAGE and IS OWNER. Databricks advises granting CAN MANAGE or IS OWNER only to users trusted to modify production code.

Question 3 · choose 1

A team wants the group data_eng, rather than any one person, to be the Run as identity of a production job. When an engineer who belongs to data_eng opens the job's Run as setting, the group is not offered. data_eng is a workspace-local group. What should the team do?

  1. ASet Run as to the built-in users group, which includes data_eng
  2. BGrant data_eng CAN MANAGE on the job so that its runs use the group
  3. CRecreate data_eng as an account group assigned to the workspace
  4. DSet Run as to a workspace admin who is also a member of data_eng
Show the answer and why
  • ASet Run as to the built-in users group, which includes data_eng

    Incorrect

    The broad built-in groups admins, users and account users cannot be used as a run-as identity.

  • BGrant data_eng CAN MANAGE on the job so that its runs use the group

    Incorrect

    CAN MANAGE lets the group edit and run the job; it does not set the group as Run as.

  • CRecreate data_eng as an account group assigned to the workspace

    Correct

    Run as accepts an account group that is assigned to the workspace; workspace-local groups cannot be used as a run-as identity.

  • DSet Run as to a workspace admin who is also a member of data_eng

    Incorrect

    The team wants the group itself, not one person, in Run as.

Run as can be set to a group. The group must be an account group that is assigned to the workspace; workspace-local groups cannot be used as a run-as identity, and neither can the built-in groups admins, users and account users.

Question 4 · choose 1

A team is designing secret scopes for its pipelines' database passwords. What does Databricks recommend?

  1. ACreate one scope per person who runs a pipeline, named after them
  2. BRely on notebook redaction to stop readers from seeing values
  3. CStore the passwords as job tags so that only the job sees them
  4. DAlign scopes to roles or applications and grant access carefully
Show the answer and why
  • ACreate one scope per person who runs a pipeline, named after them

    Incorrect

    Per-person scopes are not the recommended alignment.

  • BRely on notebook redaction to stop readers from seeing values

    Incorrect

    Redaction is best effort and does not stop authorized users.

  • CStore the passwords as job tags so that only the job sees them

    Incorrect

    Job tags are not meant for sensitive values.

  • DAlign scopes to roles or applications and grant access carefully

    Correct

    Scopes should match roles or applications, and anyone granted access can read the secrets even though outputs are redacted.

Databricks recommends aligning secret scopes to roles or applications. Workspace admins, secret creators and users granted access can read secrets; redaction in outputs cannot fully prevent that.

Question 5 · choose 1

Analysts must attach notebooks to a shared classic cluster but must not restart, resize or edit it. Which compute permission fits?

  1. ACAN RESTART
  2. BCAN ATTACH TO
  3. CCAN MANAGE
  4. DNo permission, because attaching needs only data privileges
Show the answer and why
  • ACAN RESTART

    Incorrect

    CAN RESTART adds starting, restarting and terminating the compute.

  • BCAN ATTACH TO

    Correct

    CAN ATTACH TO allows attaching notebooks and viewing the Spark UI and metrics.

  • CCAN MANAGE

    Incorrect

    CAN MANAGE also allows editing, resizing and changing permissions.

  • DNo permission, because attaching needs only data privileges

    Incorrect

    Attaching to compute requires at least CAN ATTACH TO.

Compute permissions grow from CAN ATTACH TO to CAN RESTART to CAN MANAGE; only CAN MANAGE can edit, resize or change permissions.

Practice domain 6 →Practice all domains →