Skip to content

Harrier EMR MCP

Harrier is a headless MCP server that gives AWS DevOps Agent evidence-backed diagnosis for Amazon EMR, Spark, and Kubernetes failures.

It helps operators answer:

  • Is this an infrastructure, data, Spark runtime, Kubernetes, observability, or configuration problem?
  • What evidence supports that triage call?
  • Which checks passed, failed, or remain inconclusive?
  • What should Harrier or the operator investigate next?
  • Is there a PR-ready recommendation?

Start Here

  1. Installation
  2. Local quickstart
  3. MCP tool contracts
  4. AWS permissions
  5. DevOps Agent integration

Supported Runtimes

Runtime Status Evidence
EMR on EC2 Supported EMR metadata, YARN app IDs, S3 logs, CloudWatch metrics
EMR Serverless Supported Serverless job metadata, S3 logs, CloudWatch logs, metrics
EMR on EKS Supported EMR Containers metadata, S3 logs, CloudWatch logs, optional Kubernetes pod status

Report Philosophy

Harrier’s operator-facing report is initial triage, not final RCA. It is meant to quickly show the investigation path, the visual check map, evidence cards, bounded log excerpts, inconclusive checks, and next steps.