M2.7 module quiz
Ten questions. 80% to pass (8 of 10). Open book, unlimited retakes.
Answers are collapsed under each question. Answer first, then check.
Q1
Databricks connects to ZopNight as:
A. A standalone connection on all three clouds, with its own credentials row
B. Always through the underlying cloud account that hosts the workspace
C. A standalone connection on AWS and GCP; through the Azure subscription on Azure
D. Through a cloud marketplace integration registered at onboarding
Show answer
Correct: C. The Azure exception is the detail that surprises people: there is no separate Databricks account there, and workspaces are discovered under the Azure account via its service principal.
Q2
The six discovered Databricks resource types include the workspace plus:
A. Notebooks, dashboards, queries, alerts, and stored secrets
B. Users, groups, tokens, policies, permissions
C. Tables, views, schemas, catalogs, volumes
D. Clusters, pools, SQL warehouses, jobs, model-serving endpoints
Show answer
Correct: D. Jobs and model-serving endpoints are discovered but not schedulable, which is the distinction that matters when planning what a schedule can actually cover.
Q3
Databricks auto-termination is:
A. A direct replacement for ZopNight scheduling on that cluster
B. Incompatible with scheduling, so only one may be set
C. The baseline; ZopNight scheduling adds a daily lifecycle on top
D. Configured through ZopNight rather than in Databricks
Show answer
Correct: C. Auto-term stops an idle cluster; a schedule decides whether the window is open at all. The same two-layer relationship as Snowflake auto-suspend.
Q4
Stopping a Databricks cluster is:
A. Terminal: it terminates the cluster
B. A reversible pause that resumes on the next query
C. A scale-to-zero that keeps the cluster allocated
D. A suspend, which retains the configuration intact
Show answer
Correct: A. Which is why cluster stop is treated as destructive relative to a Snowflake warehouse suspend, and why start latency matters when planning the window.
Q5
SQL Warehouse scheduling offers three patterns. The main axis they differ on is:
A. Cost measured against the published credit rate for that particular Databricks edition
B. How much of the day the warehouse is available versus how much cold-start latency users absorb
C. Provider support for the warehouse size
D. Warehouse size and its per-hour burn rate
Show answer
Correct: B. Availability against latency is the real trade. Sizing for actual usage is a separate lever that composes with whichever pattern you pick.
Q6
A scheduled cluster stop that breaks a dependent job is resolved by:
A. Disabling the ZopNight schedule on that cluster entirely and permanently from now on
B. Increasing the auto-termination threshold on the cluster
C. One of: changing the job schedule, overlapping the cluster schedule, or moving to job clusters
D. Deleting the job and recreating it on a later schedule
Show answer
Correct: C. Job clusters are the structurally cleanest of the three, because they remove the dependency rather than coordinating around it.
Q7
ZopNight detects job-schedule conflicts:
A. After the first failure
B. Never
C. Only on all-purpose clusters
D. At schedule attachment time
Show answer
Correct: D. Catching it at attachment is what turns a 3 AM failed job into a warning while you are still configuring.
Q8
Instance pools are optimized by:
A. Scaling minimum idle instances down outside working hours
B. Deleting the pools entirely outside normal working hours
C. Converting the pools into job clusters
D. Sharing the pools across workspaces
Show answer
Correct: A. A pool holding idle VMs to reduce start latency is paying for that latency reduction around the clock unless the minimum is scheduled.
Q9
Start/stop on Databricks executes over:
A. The underlying cloud account’s own credential
B. The workspace’s OAuth M2M service principal
C. A ZopNight-managed token minted per workspace
D. The connecting user’s personal access token
Show answer
Correct: B. Which is why standalone AWS and GCP connections work without an underlying cloud account being connected at all.
Q10
Databricks recommendations span rule families on:
A. All three clouds, as a provider-parameterised set
B. AWS only, since that is the primary cloud there
C. Azure only, riding the subscription
D. AWS and Azure only, not on GCP
Show answer
Correct: A. The same set of checks is instantiated per cloud: auto-termination, warehouse auto-stop, always-on serving, pool min-idle, oversizing, autoscaling and Photon settings, job placement, spot eligibility, cluster policy, cost tags, and orphans.
What’s next
Back to Databricks scheduling.