First Workflow Issue: Possible Environments Variable Issue while using MCP Client

Questions
, , ,
2

Steps we took to resolve the “Lost connection to the server” error encountered when executing an MCP Client node that runs an npx command within a Dockerized n8n setup behind an NGINX reverse proxy.

:hammer_and_wrench: Issue Summary

  • Deployment Details:

    • n8n Version: 1.90.2
    • Deployment Method: Docker Compose
    • Reverse Proxy: NGINX with Let’s Encrypt SSL
    • Domain: automation.(MY DOMAIN)
    • VPS Provider: DownDoggy.com (1 vCPU, 1 GB RAM)[n8n

  • Problem Encountered:

    • Running an MCP Client node that executes npx -y @apify/actors-mcp-server --actors (MY ACTOR) resulted in the n8n UI displaying:

“Problem running workflow – Lost connection to the server”

  • Despite the UI error, the backend process continued running.

:magnifying_glass_tilted_left: Root Cause

The issue stemmed from the NGINX reverse proxy not being configured to handle WebSocket connections properly. n8n relies on WebSockets for real-time communication between the server and the browser UI. Without the appropriate NGINX configuration, the WebSocket connection was not established, leading to the UI losing connection during long-running workflows.

:white_check_mark: Solution

We updated the NGINX configuration to support WebSocket connections and increased timeout settings to accommodate long-running processes.

Updated NGINX Configuration:

nginx

CopyEdit

server {
    listen 443 ssl;
    server_name (MY DOMAIN);

    ssl_certificate /etc/letsencrypt/live/automation.(MY DOMAIN)/fullchain.pem;
    ssl_certificate_key /etc/letsencrypt/live/automation.(MY DOMAIN)/privkey.pem;
    include /etc/letsencrypt/options-ssl-nginx.conf;
    ssl_dhparam /etc/letsencrypt/ssl-dhparams.pem;

    location / {
        proxy_pass http://localhost:(PORT #);

        # WebSocket support
        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection "upgrade";

        # Standard headers
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header X-Forwarded-Proto $scheme;

        # Timeouts for long-running workflows
        proxy_read_timeout 3600;
        proxy_send_timeout 3600;
    }
}

server {
    listen 80;
    server_name automation.(MY DOMAIN);

    if ($host = (MY DOMAIN)) {
        return 301 https://$host$request_uri;
    }

    return 404;
}

Steps Taken:

  1. Edit NGINX Configuration:
  • Updated the /etc/nginx/sites-available/n8n file with the above configuration.
  1. Test and Reload NGINX:
  • Ran sudo nginx -t to test the configuration.
  • Executed sudo systemctl reload nginx to apply the changes.

Post-Implementation Results

  • The n8n UI no longer displays the “Lost connection to the server” error during long-running workflows.
  • The MCP Client node executes successfully, and the UI remains responsive throughout the process.

:folded_hands: Acknowledgments

A special thanks to the community members who highlighted the importance of proper WebSocket configuration in reverse proxies:

Their insights were instrumental in identifying and resolving the issue.