Back to all scenarios
Scenario #411
Scaling & Load
Kubernetes v1.25, AWS EKS with PriorityClasses

Cluster Autoscaler Ignored Pod PriorityClasses

Low-priority workloads blocked scaling of high-priority ones due to misconfigured Cluster Autoscaler.

Find this helpful?
What Happened

High-priority pods remained pending, even though Cluster Autoscaler was active.

Diagnosis Steps
  • 1kubectl get pods --all-namespaces | grep Pending showed stuck critical workloads.
  • 2CA logs indicated scale-up denied due to resource reservation by lower-priority pods.
Root Cause

Default CA config didn't preempt lower-priority pods.

Fix/Workaround
• Enabled preemption.
• Re-tuned PriorityClass definitions to align with business SLAs.
Lessons Learned

CA doesn’t preempt unless explicitly configured.

How to Avoid
  • 1Validate PriorityClass behavior in test environments.
  • 2Use preemptionPolicy: PreemptLowerPriority for critical workloads.
Previous ScenarioNext Scenario