M6.6 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
Which area answers “how many replicas is this service running”?
A. Cost
B. Ship
C. Explore
D. Optimize
Show answer
Correct: B. Replicas are part of the shape you deployed. Explore describes the cloud resources underneath and Cost what they charge, but neither knows what you asked the platform to run.
Q2
The running version is broken and the fix is not merged. The right verb:
A. Rollback
B. Redeploy
C. Update the config
D. Turn auto-deploy off
Show answer
Correct: A. Rollback restores the last good version without rebuilding. Redeploy would ship the same broken inputs again.
Q3
A rollback adds what to the deploy history?
A. Nothing at all: the failed entry is amended in place
B. A new entry pointing at the earlier version
C. A note appended to the original deploy entry
D. It removes the bad deploy from the history
Show answer
Correct: B. The history is append-only, so it records what ran and when rather than a tidied account of events.
Q4
Every write is checked against three things. Which set?
A. Organisation write access setting, your live role, your token’s capability
B. Your plan tier, the resource’s region, and your role at login time
C. Your role, the image registry’s reachability, and the environment
D. Your token, the project it targets, and the time of day the call is made
Show answer
Correct: A. All three, on every call. A token can deliberately carry less than the person holding it.
Q5
Why is a permission refusal not worth retrying?
A. Because the server is rate limiting the connection
B. Because the token was consumed by the refused call
C. Because none of the three checks change by asking again
D. Because the assistant caches the refusal for the session
Show answer
Correct: C. A refusal describes configuration, not a transient failure.
Q6
Your assistant lists fewer tools than the public catalogue. Why?
A. The public catalogue is out of date and still lists retired tools
B. Your client needs updating before it can see the newer tools
C. Some of those tools have been deprecated and are now hidden from you
D. Your list is filtered to what your token, role and organisation allow
Show answer
Correct: D. Nothing advertised to you will fail on permissions, and nothing you could call is missing. It is the authoritative list for you.
Q7
Blast radius reports no dependents. What does that mean?
A. Nothing depends on the resource at all
B. Nothing the platform can see depends on it
C. It is safe to delete without checking
D. Discovery is stale and needs refreshing
Show answer
Correct: B. A hardcoded address in somebody’s script is a real dependency and an invisible one. The report lowers risk rather than removing it.
Q8
Why is a preview available at every access level?
A. Because it is cheaper than a write call
B. Because it runs against a sandbox copy
C. Because tier 1 admits it for every role
D. Because it changes nothing, so it is a read
Show answer
Correct: D. Checking first is never gated behind write access, which is what makes it a habit rather than a privilege.
Q9
The build finished and the rollout never completed. Where is the reason?
A. The build output
B. The audit log
C. The running replica’s log
D. The service configuration
Show answer
Correct: C. The builder has nothing left to say once the image exists. Something in the container is refusing to stay up, and it is usually saying so.
Q10
A log answer reports coverage as windowed. What follows?
A. Nothing was printed during that build, so there is no error
B. You need to ask for a larger tail before concluding anything
C. The service has no log collection configured for it at all
D. Older output exists and was not returned, so absence proves nothing
Show answer
Correct: D. Windowed means the cap was reached. A long or retried build is the most likely case, and it is exactly when a wrong root cause gets agreed.
What’s next
Back to Ship and operate through MCP.