Skip to main content

Troubleshooting Guide

This guide provides solutions to common issues encountered when working with MCP Server with LangGraph.

Quick Navigation


Kubernetes-Specific Guides


Quick Diagnostic Commands

Check Service Health

# Local development
curl http://localhost:8000/health

# Kubernetes
kubectl get pods -n mcp-server
kubectl logs -n mcp-server <pod-name>

Enable Debug Logging

# Set environment variable
export LOG_LEVEL=DEBUG

# Or in .env file
LOG_LEVEL=DEBUG

View Traces

# Open LangSmith dashboard (if configured)
open https://smith.langchain.com

# Or check local Jaeger
open http://localhost:16686

Common Quick Fixes

IssueQuick Fix
ModuleNotFoundErroruv sync --extra dev
Pod CrashLoopBackOffCheck logs: kubectl logs <pod> --previous
401 UnauthorizedVerify JWT token is valid and not expired
503 Service UnavailableCheck upstream service health
OOM KilledIncrease memory limits in deployment
Tests fail in CIEnsure uv sync --extra dev in workflow

Getting More Help

If you’re still experiencing issues:
  1. Check Logs: Enable debug logging with LOG_LEVEL=DEBUG
  2. Review Documentation: See our comprehensive docs
  3. Search Issues: Check GitHub Issues
  4. Ask Community: Post in GitHub Discussions
  5. Report Bug: Create an issue with:
    • Error message
    • Steps to reproduce
    • Environment details (python --version, kubectl version, etc.)
    • Relevant logs

Troubleshooting Guides Index

GuideDescription
InstallationPython, uv, dependencies
DeploymentKubernetes, Docker, Cloud Run
AuthenticationJWT, Keycloak, OpenFGA
PerformanceMemory, CPU, latency
CI/CDGitHub Actions, builds, tests
DatabasePostgreSQL, Redis, sessions
Pod FailuresKubernetes pod issues