Amazon EKS insights

CommercialVersions: [current]67 metrics

OpenTelemetry·

Karpenter nodes disrupted earlier than expireAfter duration by drift or consolidation

Missing disruption budget allows unrestricted node disruption outside schedule

Blocking pods forcefully terminated at end of terminationGracePeriod

do-not-disrupt annotation does not prevent forceful termination

terminationGracePeriod longer than budget schedule duration causes undefined behavior

Kubelet crashes or stops causing node NotReady state

Disk pressure causes node NotReady from full disk or inode exhaustion

Memory exhaustion triggers MemoryPressure and OOM kills on nodes

Containerd runtime failure prevents pod execution and node readiness

Network connectivity loss between node and EKS API server causes NotReady

Node version mismatch after cluster upgrade causes compatibility issues

Intermittent NotReady node flapping from heartbeat delays or network instability

All nodes NotReady from control plane or security group issues

Newly joined node stuck NotReady from authentication or CNI initialization failures

Insufficient CPU or memory prevents pod scheduling

Node selector mismatch blocks pod scheduling

Node taints prevent pod scheduling without matching tolerations

Volume AZ constraint prevents scheduling when no nodes available in volume's zone

Pod topology spread constraints block scheduling across zones

Maximum pod limit per node prevents scheduling

Cordoned nodes block all new pod scheduling

Aggressive NLB health check failover removes healthy capacity during network propagation

DNS TTL extremes amplify or prolong regional outage impact

DynamoDB DNS race condition causes cascading AWS control plane failure

EC2 control plane congestive collapse from DynamoDB dependency failure

IAM control plane degradation creates diagnostic feedback loop during outages

Thundering herd risk during mass service recovery after control plane restoration

Control plane failure causes data plane unavailability despite intact infrastructure

DNS endpoint resolution failure causes cascading control plane outage

EC2 lease expiration prevents EKS/ECS container launches during recovery

Amazon EKS insights requires a free account. Sign in with Google or GitHub to access the full knowledge base.

Sign in to access