M2.3 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
The auto-remediation allowlist and the Safety Gate differ in that:
A. The allowlist is keyed on rule id and is resource-blind; the gate is resource-aware
B. They are two names for the same check, one used in the catalog and one in the executor logs
C. The gate runs first and filters the resource set before the allowlist is consulted at all
D. The allowlist is customer-editable per org, while the gate is fixed by the platform
Show answer
Correct: A. The allowlist answers “does this rule support auto”; the gate answers “is it safe on this resource”. The rule-grouped pill reflects capability, the per-recommendation pill reflects the gate’s verdict.
Q2
Under the Safety Gate, stop_compute is:
A. Allowed when the resource is tagged non-prod
B. Allowed on dev accounts
C. Allowed with approval
D. Always demoted to advisory
Show answer
Correct: D. No cloud API exposes the workload inside a generic compute instance, so a one-shot stop can never be proven safe. The gate reads control-plane fields only, never tags, because tags are customer-set and can be stale.
Q3
For delete_resource, an absent orphan-evidence flag is treated as:
A. Equivalent to false, since an absent association is an empty one
B. Evidence of orphan status, because nothing is attached to it
C. Not evidence, so the recommendation is demoted
D. A reason to retry later, once the discoverer re-reads it
Show answer
Correct: C. Fail closed. In-use is cheap to prove; orphan must be proven. This is why RC-002 stays advisory: AWS gives no real last-detach signal from a single DescribeVolumes call.
Q4
The four authoritative orphan-evidence keys include:
A. resource_age_days above the configured dwell threshold for that rule
B. tag_environment set to dev, since dev resources are safe to reclaim
C. last_login older than the rule’s configured inactivity window
D. ip_associated explicitly false
Show answer
Correct: D. The others are unattached with a detached_since dwell, source_disk_gone, and protected_items_count == 0. All four are control-plane facts, never tags or name tokens.
Q5
The 3-step remediation workflow is:
A. detect, then notify, then apply the change once acknowledged
B. precondition → action → validate
C. plan, then approve, then execute against the target resource
D. snapshot, then mutate, then verify the resulting state
Show answer
Correct: B. With an optional approval step between precondition and action for destructive operations. Halt-on-first-failure means a failed precondition never reaches the cloud.
Q6
“Guided” remediation means:
A. A ZopNight engineer performs it on the customer’s behalf after review
B. It runs only during a declared maintenance window agreed in advance
C. A type-to-confirm review modal before execution
D. It requires two separate approvers before the executor will run it
Show answer
Correct: C. A human decision point rather than an approval queue. Single-replica K8s scale-ups and orphan EBS snapshot deletes are guided for different reasons: one is a deliberate customer choice, the other is irreversible.
Q7
The three error classes are:
A. fatal, warning and info, mirroring log levels
B. user_action, transient, system
C. client, server and network, by failure origin
D. retryable, permanent and unknown outcomes
Show answer
Correct: B. user_action is yellow and shows a fix hint; transient is blue and offers retry; system is red and offers a diagnostic to copy. Only transient sets the retryable flag.
Q8
ZopNight never auto-mutates customer databases because:
A. Database mutations risk data rather than availability, and the asymmetry is permanent
B. Cloud providers forbid automated mutation of managed database services entirely
C. Databases cannot be discovered reliably enough to act on automatically
D. Database APIs are unreliable and frequently time out mid-operation
Show answer
Correct: A. The denylist is hardcoded rather than configurable, and it explicitly covers RDS, Aurora, ElastiCache, Cloud SQL, Azure SQL and the Postgres and MySQL families.
Q9
Terminal success and failure emails are:
A. On by default for every single organisation created
B. Sent only to Admins on the org
C. Not available on any plan
D. Gated behind an environment variable, to avoid noise
Show answer
Correct: D. Approval-required notifications are always sent, guarded so a reconciler replay cannot re-send. Terminal notifications are opt-in because most orgs do not want an email per successful remediation.
Q10
Schedule-class recommendations are remediated by:
A. The workflow engine, which queues the change and waits for an approval first
B. A deep-link to schedule creation with the resource prefilled and auto-attached on save
C. A guided modal that walks through creating the schedule by hand
D. An immediate stop, with the schedule offered afterwards
Show answer
Correct: B. Their fix is a recurring window rather than a one-shot action, so they use the schedule-attach path rather than the remediation workflow. That is also why they were carved into their own category.
What’s next
Back to Auto-remediation.