Skip to main content

EKS Node Monitoring Agent

Written: 2025-08-26 | Updated: 2026-02-13 | Reading time: ~7 min

Overview​

The EKS Node Monitoring Agent (NMA) is an AWS-provided node health monitoring tool that automatically detects and reports hardware and system-level issues on EKS cluster nodes. GA in 2024, it works alongside Node Auto Repair for improved cluster stability.

Key Features​

  • Log-based problem detection: Real-time system log analysis with pattern matching
  • Automatic event generation: Creates Kubernetes Events and Node Conditions on detection
  • CloudWatch integration: Centralized monitoring via CloudWatch
  • EKS Add-on support: Simple installation and management

Architecture​

DaemonSet-based deployment monitoring: Container Runtime, Storage, Networking, Kernel, and Accelerated Hardware (GPU/Neuron). Uses controller-runtime for K8s-native integration.

Node Conditions​

  • ContainerRuntimeReady, StorageReady, NetworkingReady, KernelReady, AcceleratedHardwareReady

Detectable Issues​

Conditions (auto-repair targets): DiskPressure, MemoryPressure, PIDPressure, NetworkUnavailable, KubeletUnhealthy, ContainerRuntimeUnhealthy

Events (warning only): Kernel soft lockup, I/O delays, filesystem errors, network packet loss, hardware error indicators

Limitations​

  • Not a metrics collection tool — log-based analysis only
  • Cannot detect sudden hardware failures or complete network disconnection
  • Limited Prometheus endpoint (port 8080)

Best Practices​

Use NMA as L1 (state detection) in a multi-layer monitoring stack: L1 NMA → L2 Container Insights/Prometheus → L3 Node Auto Repair → L4 Unified Dashboard.

Resource usage: CPU 10m-250m, Memory 30Mi-100Mi.