Zonal Cluster Operations Quiz
Related Document: Zonal Cluster Operations
Multiple Choice Questions
1. What does the seven-day EKS native rollback window mean?
- A) The time rollback must take
- B) Eligibility to initiate rollback after upgrade completion
- C) Maximum node lifetime
- D) The period for automatic add-on restoration
Show Answer
Answer: B) Eligibility to initiate rollback after upgrade completion
It is the window to initiate rollback to the immediately previous minor version. Creation version, support status, subsequent upgrades, and compatibility still matter. Auto Mode rolls back nodes first; add-ons, applications, and data changes are not automatically restored.
2. What must be checked after setting an NLB target-group weight to zero?
- A) All existing connections terminate immediately
- B) TargetGroupBinding moves to another cluster
- C) New-flow reduction and existing-flow draining separately
- D) ARC rewrites every other cluster’s weights automatically
Show Answer
Answer: C) New-flow reduction and existing-flow draining separately
Ordinary weight changes affect new flows, but the current NLB guide says weight zero also closes existing connections after a short period. Do not assume natural draining alone; test reconnection and retry behavior. Check NewFlowCount, ActiveFlowCount, and errors before changing nodes. TGB has no weight field, and EKS zonal shift does not automatically change weights across clusters.
3. Which combination configures Kafka KIP-392 correctly?
- A) broker.rack alone automatically configures all consumers
- B) RackAwareReplicaSelector, broker.rack, and matching consumer client.rack
- C) Only unclean.leader.election.enable=true
- D) Different AZ-name and AZ-ID strings are interchangeable
Show Answer
Answer: B) RackAwareReplicaSelector, broker.rack, and matching consumer client.rack
The complete replica.selector.class is org.apache.kafka.common.replica.RackAwareReplicaSelector. The Strimzi Kafka CR configures brokers; ordinary application consumers need separate configuration. Selection falls back to the leader when no suitable local replica exists.
4. What is the priority for GLIDE AZ_AFFINITY_REPLICAS_AND_PRIMARY?
- A) Primary only
- B) Local replicas → local primary → replicas or primary in other AZs
- C) Remote replicas only
- D) Always error when no local node exists
Show Answer
Answer: B) Local replicas → local primary → replicas or primary in other AZs
Server AZ metadata must match client_az. Evaluate tolerance for stale replica reads; read percentage alone is insufficient. HotelTrader’s reported improvements include request batching as well as AZ-aware routing.
5. Which statement about Aurora’s default reader endpoint is correct?
- A) It distributes each SQL query to a different replica
- B) It guarantees a reader in the same AZ
- C) It balances connections and can use the writer if no replicas exist
- D) It requires the JDBC Wrapper to connect
Show Answer
Answer: C) It balances connections and can use the writer if no replicas exist
It does not guarantee AZ preference. Per-AZ READER custom endpoints need membership and fallback management. JDBC fastestResponse uses response time, not a strict AZ constraint. Feature request #1139 was closed in 2025.
6. Which statement about discovering a pod’s AZ is incorrect?
- A) Ordinary Pod-create admission always knows the scheduler-selected node
- B) The AWS MSK Kyverno example handles Pod/binding requests
- C) spec.nodeName can be passed to an initialization component through the Downward API
- D) IMDS access depends on environment and security configuration
Show Answer
Answer: A) Ordinary Pod-create admission always knows the scheduler-selected node
A node normally has not been selected at Pod creation. Use binding-time injection or post-scheduling lookup. The Downward API does not directly query node labels, and Strimzi does not configure unrelated application consumers automatically.