Kubernetes & Amazon EKS Training

English

한국어
中文
日本語
Español

English

한국어
中文
日本語
Español

Appearance

Zonal Cluster Operations Quiz ​

Related Document: Zonal Cluster Operations

Multiple Choice Questions ​

1. What is the eligibility window for Amazon EKS's native Kubernetes version rollback (GA'd July 2026)? ​

  • A) 24 hours
  • B) 7 days
  • C) 30 days
  • D) Unlimited
Show Answer

Answer: B) 7 days

Explanation: EKS native rollback can revert one minor version at a time, within 7 days of the upgrade. Clusters created at the target version, more than 7 days elapsed, or clusters already re-upgraded are not eligible.

2. What mechanism is used to drain traffic out of a zone during a Zonal In-Place upgrade? ​

  • A) kubectl drain
  • B) Adjusting Target Group weight
  • C) Waiting for DNS TTL to expire
  • D) Recreating the cluster
Show Answer

Answer: B) Adjusting Target Group weight

Explanation: Instead of touching anything inside the cluster, you adjust the weight of the Target Group bound via TargetGroupBinding to reduce or stop traffic to a given zone. For unplanned situations like an AZ outage, ARC Zonal Shift performs this role automatically.

3. What must be set on Kafka brokers to enable KIP-392 (Follower Fetching)? ​

  • A) auto.leader.rebalance.enable=true
  • B) replica.selector.class=RackAwareReplicaSelector
  • C) unclean.leader.election.enable=true
  • D) min.insync.replicas=2
Show Answer

Answer: B) replica.selector.class=RackAwareReplicaSelector

Explanation: Brokers need replica.selector.class set to RackAwareReplicaSelector and a broker.rack (AZ ID) assigned. On the consumer side, the client.rack property must be set to the consumer's own AZ ID so fetches get redirected to a same-rack follower.

4. Which Valkey GLIDE ReadFrom strategy is recommended for workloads that are over 99% reads? ​

  • A) PRIMARY
  • B) PREFER_REPLICA
  • C) AZ_AFFINITY_REPLICAS_AND_PRIMARY
  • D) Random distribution
Show Answer

Answer: C) AZ_AFFINITY_REPLICAS_AND_PRIMARY

Explanation: It prefers a same-AZ replica first, falls back to the same-AZ primary, and only reaches into other AZs as a last resort. For read-dominant workloads this is the recommended balance of cost savings and availability — HotelTrader cut inter-AZ transfer costs by 95% after adopting it.

5. Which statement about Amazon Aurora's default reader endpoint is correct? ​

  • A) It automatically prioritizes replicas in the same AZ
  • B) It is round-robin DNS with no AZ awareness
  • C) It always routes to the primary
  • D) It cannot be used without the AWS Advanced JDBC Wrapper
Show Answer

Answer: B) It is round-robin DNS with no AZ awareness

Explanation: Aurora's default reader endpoint has no AZ affinity. You can work around this with per-AZ custom endpoints or the AWS Advanced JDBC Wrapper's fastestResponse strategy, but true AZ affinity itself remains an open feature request in the aws-advanced-jdbc-wrapper repository.

6. Which statement about how a pod can determine its own AZ is INCORRECT? ​

  • A) It can look this up directly via EC2 IMDS
  • B) A Kyverno mutating policy can copy a node label onto a pod annotation
  • C) The Kubernetes Downward API injects the node's zone label into the pod by default
  • D) An operator like Strimzi can provide rack-awareness as a built-in feature
Show Answer

Answer: C) The Kubernetes Downward API injects the node's zone label into the pod by default

Explanation: The Downward API does not automatically inject a node's topology.kubernetes.io/zone label into a pod. That's why one of the other approaches — direct IMDS lookup, Kyverno-based admission-time label copying, or an operator's built-in support like Strimzi's — is needed.