Diagnose EKS Node Issues Faster with AWS DevOps Agent and Custom MCP

This article demonstrates how to extend AWS DevOps Agent with custom Model Context Protocol (MCP) servers to diagnose EKS node issues by accessing operating system-level diagnostics unavailable through native integrations.

• AWS DevOps Agent autonomously investigates production incidents but cannot access node OS data like iptables, CNI config, and kernel logs
• Custom MCP servers bridge this gap by exposing structured access to unreachable data sources
• Reference implementation uses SSM Automation to safely collect 20+ node-level log sources and upload to S3
• Amazon Bedrock AgentCore Gateway exposes Lambda-backed MCP server to AWS DevOps Agent with OAuth authentication
• Demonstration shows agent identifying DNS resolution failure caused by iptables DROP rules through parallel investigation tasks
• Pattern generalizes to EC2, ECS, on-premises servers, network devices, and third-party tools via API integrations
• Requires SSM Agent on nodes; not a replacement for continuous monitoring or log shipping

This approach enables autonomous root cause analysis for infrastructure issues by systematically correlating evidence across multiple diagnostic layers without manual SSH access.