EKS Auto Mode Scaling Behavior Quiz

Related Document: Scaling Behavior

Multiple Choice Questions

1. What is the behavior of NodePool's consolidationPolicy: WhenEmptyOrUnderutilized?

• A) Only removes empty nodes
• B) Consolidates both empty and underutilized nodes
• C) Always keeps all nodes
• D) Only removes nodes at specific times

Show Answer

Answer: B) Consolidates both empty and underutilized nodes

Explanation: The WhenEmptyOrUnderutilized policy consolidates not only empty nodes but also underutilized nodes to optimize costs. This allows consolidating workloads from multiple underutilized nodes onto fewer nodes.

disruption:
  consolidationPolicy: WhenEmptyOrUnderutilized
  consolidateAfter: 1m  # Consolidate 1 minute after condition is met

Comparison:

• WhenEmpty: Only removes empty nodes (conservative)
• WhenEmptyOrUnderutilized: Consolidates empty + underutilized nodes (aggressive)

2. What is the kubectl command to check NodeClaim status?

• A) kubectl get nodes --show-claims
• B) kubectl get nodeclaims
• C) kubectl describe karpenter claims
• D) kubectl get ec2-nodes

Show Answer

Answer: B) kubectl get nodeclaims

Explanation: NodeClaim is a resource that represents the state of a node being provisioned.

# List NodeClaims
kubectl get nodeclaims

# Detailed information for specific NodeClaim
kubectl describe nodeclaim <name>

# View NodeClaims with node information
kubectl get nodeclaims -o wide

3. Under what condition does node provisioning start when a Pending Pod occurs in Auto Mode?

• A) When Pod has been Pending for more than 5 minutes
• B) When no node satisfies the NodePool's requirements
• C) When total node count falls below threshold
• D) When manual scale-up command is executed

Show Answer

Answer: B) When no node satisfies the NodePool's requirements

Explanation: Auto Mode analyzes the requirements of Pending Pods and immediately provisions a new node if no suitable node exists.

Provisioning Flow:

1. Pod enters Pending state
2. Karpenter analyzes Pod's resource requests, nodeSelector, affinity
3. Checks requirements of suitable NodePools
4. Selects optimal instance type
5. Launches EC2 instance (40-90 seconds)

4. In which cases does Consolidation NOT occur?

• A) When node has do-not-disrupt annotation
• B) When node only has DaemonSet Pods
• C) When consolidateAfter time has not elapsed
• D) All of the above

Show Answer

Answer: D) All of the above

Explanation: Consolidation does not occur in the following situations:

1. do-not-disrupt annotation: Nodes or Pods with this annotation are excluded from consolidation
2. Only DaemonSet Pods: DaemonSets run on all nodes, so treated as empty node
3. consolidateAfter not elapsed: Must wait specified time after condition is met

metadata:
  annotations:
    karpenter.sh/do-not-disrupt: "true"

5. What situations trigger Drift detection?

• A) When NodeClass spec changes
• B) When new AMI becomes available
• C) When security groups change
• D) All of the above

Show Answer

Answer: D) All of the above

Explanation: Drift detection is triggered when the current state of a node differs from the desired state:

• NodeClass changes: AMI family, subnets, security groups, etc. changed
• New AMI: EKS optimized AMI updated
• Security group changes: Referenced security groups modified

When Drift is detected, nodes are replaced sequentially.

6. Which AMI family is recommended for optimizing node provisioning speed?

• A) AL2023
• B) Bottlerocket
• C) Ubuntu
• D) Amazon Linux 2

Show Answer

Answer: B) Bottlerocket

Explanation: Bottlerocket is an OS designed specifically for containers, providing faster boot times than AL2023.

Boot Time Comparison:

• AL2023: 20-40 seconds
• Bottlerocket: 15-25 seconds

Additional Bottlerocket benefits:

• Smaller attack surface
• Immutable file system
• Automatic security updates