惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

F
Fortinet All Blogs
云风的 BLOG
云风的 BLOG
M
MIT News - Artificial intelligence
WordPress大学
WordPress大学
T
Tailwind CSS Blog
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
S
Secure Thoughts
博客园 - 【当耐特】
Know Your Adversary
Know Your Adversary
NISL@THU
NISL@THU
博客园 - 司徒正美
Last Week in AI
Last Week in AI
C
Cybersecurity and Infrastructure Security Agency CISA
P
Privacy & Cybersecurity Law Blog
C
CXSECURITY Database RSS Feed - CXSecurity.com
B
Blog
The GitHub Blog
The GitHub Blog
小众软件
小众软件
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
Spread Privacy
Spread Privacy
Martin Fowler
Martin Fowler
博客园 - 叶小钗
Security Archives - TechRepublic
Security Archives - TechRepublic
T
Tenable Blog
S
Securelist
博客园 - 三生石上(FineUI控件)
Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
Microsoft Security Blog
Microsoft Security Blog
Apple Machine Learning Research
Apple Machine Learning Research
罗磊的独立博客
T
Threat Research - Cisco Blogs
Application and Cybersecurity Blog
Application and Cybersecurity Blog
F
Full Disclosure
Cloudbric
Cloudbric
The Cloudflare Blog
Y
Y Combinator Blog
Hugging Face - Blog
Hugging Face - Blog
Microsoft Azure Blog
Microsoft Azure Blog
H
Hacker News: Front Page
腾讯CDC
L
Lohrmann on Cybersecurity
C
CERT Recently Published Vulnerability Notes
V2EX - 技术
V2EX - 技术
GbyAI
GbyAI
TaoSecurity Blog
TaoSecurity Blog
I
Intezer
The Last Watchdog
The Last Watchdog
G
GRAHAM CLULEY
Google Online Security Blog
Google Online Security Blog
T
The Blog of Author Tim Ferriss

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant Common SOC 2 Failures (Real World) Stop Vibe-Checking Your AI App: A Practical Guide to Evals How to Use SonarQube and SonarScanner Locally to Level Up Your Code Quality Your Next To-Do App Is Dead — I Replaced Mine with an OpenClaw AI Sign a Nostr event in 60 lines of Python using coincurve — no nostr-sdk, no nbxplorer, no rust toolchain ITGC Audit Explained Like You’re in Big 4 Patch Tuesday abril 2026: Microsoft parcha 163 vulnerabilidades y un zero-day en SharePoint Stop scraping everything: a better way to track competitor price changes Listing on MCPize + the Official MCP Registry while routing payments OUTSIDE the marketplace — how I kept 100% of my x402 revenue Building an AI-Powered Risk Intelligence System Using Serverless Architecture Why We Ripped Function Overloading Out of Our AI Toolchain Testing AI-Generated Code: How to Actually Know If It Works SaaS Churn Is Killing Your Business. Here Is What to Do About It (Without a Support Team) The Speed of AI Is No Longer Linear - And Self-Improving Models Are Why How to Implement RBAC for MCP Tools: A Practical Guide for Engineering Teams From Standard Quote to Persuasive Proposal: AI Automation for Arborists I built a CLI that scaffolds complete multi-tenant SaaS apps Axios CVE-2025–62718: The Silent SSRF Bug That Could Be Hiding in Your Node.js App Right Now The dashboard that ended our friendship Data Pipelines Explained Simply (and How to Build Them with Python) The Hidden Cost of AI Systems Nobody Talks About. undefined vs undeclared, and how typeof behaves Switching from file-based jobs to NATS/Kafka in Rust without changing code io_uring Adventures: Rust Servers That Love Syscalls Why Agentic AI is Killing the Traditional Database The POUR principles of web accessibility for developers and designers Quantum Neural Network 3D — A Deep Dive into Interactive WebGL Visualization How To Install Caveman In Codex On macOS And Windows Automation Pipeline Reliability: Why Your Workflow Breaks When Nobody Is Watching I Built an 'Open World' AI Coding Agent — It Works From ANY Folder From Freelancing to Product: A Tech Service Company's SaaS Transformation China's AI Giants: Adding Tencent Hunyuan & ByteDance Doubao to AI University (74 Providers) On the Vibe Coders and Their Lies clerk: Auto-Summarize Your Claude Code Sessions AI Weekly — 2026/04/10–04/17 | The Model Lockdown Is Here, but the Toolchain Is the Real Battleground AI 週報 — 2026/04/10–2026/04/17 模型封鎖潮來了,但工具鏈才是真戰場 Maybe this is how Open-Source apps are born... 🚀 Fine-Tune LLMs with LoRA and QLoRA: 2026 Guide tRPC v11 + Next.js App Router: End-to-End Type Safety Without the Boilerplate ShadCN UI in 2026: Why I Stopped Installing Component Libraries and Started Owning My Components SaaS Billing in React Server Components: Stripe + Supabase Without a Single `useEffect` Join our DEV Weekend Challenge — $1,000 in Prizes Across TEN winners! Submissions Due April 20 at 6:59 AM UTC. Implementing FSRS Spaced Repetition in Flutter + Supabase — Adding Memory Science to an AI Learning App "I Texted My Localhost From the Train — Claude Code Fixed the Bug Before I Got Home" I Built a Sales Prep AI and It Went Deeper Than Expected Design to Code #2: One JSON, Eleven Outputs Solving the 100M-Row Problem: A Summary Table Pattern for High-Volume Push Notification Logs Flutter Web With Wasm: What Actually Changes For Developers I Built 50 Royalty-Free Soundtracks for My Side Project in a Weekend Using AI Music Generation The Vibe Coding Security Checklist: 7 Things to Check Before You Ship Stop Letting Googlebot Guess Fix Your React App's SEO Right Desconstruindo o Streaming do LinkedIn: Como Criar um Engine de Extração de Vídeo de Alta Performance com HLS e FFmpeg (EDA Part-1) EDA (Exploratory Data Analysis) Explained With Real Life — Why Looking at Your Data Is the Most Important Step in Machine Learning Brand Relationship Management at Scale: Our 4-Touch Outreach System for 200+ Brands Why String.fromEnvironment() Might Return an Empty String in Dart JGuardrails 1.0.0 — Hardening Java LLM Apps Against Jailbreaks, Toxicity, and Prompt Injection Plan and Schedule a Full Week of Threads Content From One Claude Conversation Coding Cat Oran Ep3, Five Tables Changed Everything Updated: BFF Pattern I'm done watching freelancers get buried by 200 proposals. So I'm building the alternative. This is my first post BFS Algorithm in Java Step by Step Tutorial with Examples Tracking LLM Pricing Monthly: An Open Dataset for 22 AI Models How We Measure Content ROI on a Comparison Site: Revenue Attribution Without Perfect Data Introducing Nova AI Ops: The AI-Native Operating System for SRE Teams I built a free desktop video downloader for Windows — Grabbit How Talkie OCR Helps Vision-Impaired & Dyslexic Users Read the World Around Them VRCFaceTracking安装和iPhone面捕配置教程,有bug Even CrowdStrike Can't See Your Agents The Automation Gold Rush: What n8n Workflows and Claude Are Opening Up for Developers Right Now
MCP Connection Issues: Why My MCP Server Kept Dropping Connections and How I Fixed It (After 86 Production Outages)
KevinTen · 2026-06-25 · via DEV Community

KevinTen

MCP Connection Issues: Why My MCP Server Kept Dropping Connections and How I Fixed It (After 86 Production Outages)

Honestly, I thought I had everything figured out after 86 outages. Logging was done. Health checks were in place. Tool discovery had fuzzy matching. Authentication handled every possible location. Rate limiting was working.

But my MCP server kept dropping connections randomly.

Three days of debugging later, I found the problem — and it wasn't what I expected. It was all about SSE (Server-Sent Events) streaming. If you're building an MCP server, you need to read this. I saved you three days of headache.

The Problem: Random Connection Drops

Let me set the scene. My MCP server has been running in production for a while. Everything works fine most of the time. But randomly, clients would get disconnected mid-request. Sometimes it happened after 30 seconds. Sometimes after 2 minutes. Sometimes it worked perfectly.

The worst part? It only happened in production. Locally, everything worked fine. Classic "it works on my machine" situation.

Here's what the error looked like from the client side:

Error: disconnected before response completed

That's it. No additional info. Just ... disconnected.

My First Guesses (All Wrong)

Like any good developer, I started blaming the usual suspects:

  1. Reverse proxy timeout – I checked Nginx settings. Increased timeouts. Restarted. Still happening.
  2. Cloudflare idle timeout – Cloudflare has a 100-second idle timeout. But it was happening faster than that sometimes. Still, I checked. Didn't help.
  3. Memory leak – Profiled the app. Memory was stable. No leak.
  4. Thread deadlock – Looked at thread dumps. Nothing blocked. Everything looked normal.
  5. Database connection issues – Database connection pool was fine. Connections checked out fine.

I spent half a day going through this list. Nothing worked. The problem kept happening.

The Breakthrough: It's the SSE, Stupid

MCP uses SSE for the connection from server to client. The client sends the JSON-RPC request over HTTP POST, and the server streams responses back via SSE.

Here's what I missed: SSE connections are idle connections. If nothing is being sent, some proxies/load balancers will drop them.

Wait, but how long is idle? Depends on your hosting. On Fly.io, the default idle timeout is 75 seconds. On Cloudflare, it's 100 seconds. On Heroku, it's 55 seconds.

And here's the thing about MCP: most of the time, the connection is idle. The client connects, waits for the LLM to process the request, then the server sends the response. If the LLM takes more than your proxy's idle timeout ... boom. Connection dropped.

I know, I know. You're thinking "just increase the timeout." But you can't always do that. Some platforms don't let you change it. And even if you can, what if the LLM is processing a really big request that takes 5 minutes? You can't just set the timeout to 10 minutes – that creates other problems.

The Solution: Heartbeat Comments

SSE allows any comment line starting with : to be sent as a heartbeat. These comments don't interrupt the actual event stream, they just keep the connection alive.

Here's what it looks like on the wire:

: heartbeat
data: {"jsonrpc":"2.0","method":"tools/list","result":{"tools":[...]}

: heartbeat
: heartbeat
data: {"jsonrpc":"2.0","method":"tools/call","result":{"content":[...]}

The client just ignores the comment lines. But they count as activity, so the proxy doesn't drop the connection.

Brilliant, right? Simple, effective, almost no code.

Implementation in Java/Spring Boot

I'm building my MCP server with Java Spring Boot, so let me show you the complete implementation. It's about 30 lines of code.

First, the SSE emitter configuration with heartbeats:

import org.springframework.web.servlet.mvc.method.annotation.SseEmitter;
import java.io.IOException;
import java.util.concurrent.Executors;
import java.util.concurrent.ScheduledExecutorService;
import java.util.concurrent.ScheduledFuture;
import java.util.concurrent.TimeUnit;

public class McpSseConnection {
    private final SseEmitter emitter;
    private final ScheduledExecutorService scheduler;
    private ScheduledFuture<?> heartbeatTask;
    private static final long HEARTBEAT_INTERVAL = 30; // seconds

    public McpSseConnection() {
        // Use 0 timeout because we handle heartbeats ourselves
        this.emitter = new SseEmitter(0L);
        this.scheduler = Executors.newSingleThreadScheduledExecutor();
        startHeartbeat();
    }

    private void startHeartbeat() {
        heartbeatTask = scheduler.scheduleAtFixedRate(() -> {
            try {
                // Send a comment heartbeat – this keeps the connection alive
                emitter.send(SseEmitter.event().comment("heartbeat"));
            } catch (IOException e) {
                // Connection closed, stop heartbeats
                heartbeatTask.cancel(false);
                scheduler.shutdown();
            }
        }, HEARTBEAT_INTERVAL, HEARTBEAT_INTERVAL, TimeUnit.SECONDS);
    }

    public SseEmitter getEmitter() {
        return emitter;
    }

    public void complete() {
        heartbeatTask.cancel(false);
        scheduler.shutdown();
        emitter.complete();
    }

    public void completeWithError(Throwable ex) {
        heartbeatTask.cancel(false);
        scheduler.shutdown();
        emitter.completeWithError(ex);
    }
}

Then in your controller, use it like this:

import org.springframework.http.MediaType;
import org.springframework.web.bind.annotation.PostMapping;
import org.springframework.web.bind.annotation.RestController;
import org.springframework.web.servlet.mvc.method.annotation.SseEmitter;

@RestController
public class McpController {

    @PostMapping(value = "/mcp/messages", produces = MediaType.TEXT_EVENT_STREAM_VALUE)
    public SseEmitter handleMessages() {
        McpSseConnection connection = new McpSseConnection();
        SseEmitter emitter = connection.getEmitter();

        // Process the MCP request asynchronously
        mcpMessageProcessor.process(request, emitter)
            .whenComplete((result, error) -> {
                if (error != null) {
                    connection.completeWithError(error);
                } else {
                    connection.complete();
                }
            });

        return emitter;
    }
}

That's it. Thirty lines of code. Problem solved.

What About CORS?

Oh right, you're going to hit CORS preflight issues with SSE too. Let me save you another hour. Make sure your CORS configuration allows OPTIONS for the SSE endpoint and doesn't require auth for preflight:

import org.springframework.context.annotation.Bean;
import org.springframework.context.annotation.Configuration;
import org.springframework.web.cors.CorsConfiguration;
import org.springframework.web.cors.UrlBasedCorsConfigurationSource;
import org.springframework.web.filter.CorsFilter;

@Configuration
public class CorsConfig {

    @Bean
    public CorsFilter corsFilter() {
        CorsConfiguration config = new CorsConfiguration();
        config.setAllowCredentials(true);
        config.addAllowedOriginPattern("*"); // customize for your domains
        config.addAllowedHeader("*");
        config.addAllowedMethod("*");
        // Allow preflight to be cached for 1 hour
        config.setMaxAge(3600L);

        UrlBasedCorsConfigurationSource source = new UrlBasedCorsConfigurationSource();
        source.registerCorsConfiguration("/**", config);
        return new CorsFilter(source);
    }
}

The key point: don't require authentication for OPTIONS requests. Most proxies and browsers expect OPTIONS to be unauthenticated. If you require auth on OPTIONS, preflight will fail before the connection even starts.

What About proxy buffering?

Another gotcha! If you're using Nginx in front of your server, turn off proxy buffering for SSE endpoints:

location /mcp/ {
    proxy_pass http://backend;
    proxy_http_version 1.1;
    proxy_set_header Connection '';
    proxy_chunked_transfer_encoding on;
    proxy_buffering off;
    proxy_cache off;
}

If you leave proxy buffering on, Nginx will buffer the entire response before sending it to the client. That means no streaming – the client waits until everything is done. Which defeats the whole purpose of SSE streaming.

I learned this the hard way. Everything looked fine locally, but on production with Nginx, nothing worked until I added these settings.

Pros & Cons of This Approach

Let me be honest with you – this isn't perfect. Nothing is. Here's what works and what doesn't.

Pros

Simple – 30 lines of code, no complex infrastructure

Works everywhere – every proxy respects heartbeats

Low overhead – 30-second heartbeats is basically nothing for your server

Backward compatible – all MCP clients work with this, no changes needed

Cheap – doesn't consume any extra resources when the connection is active

Cons

Still consumes a connection – heartbeat keeps it alive, but you still have an open HTTP connection

Doesn't solve very long pauses – if your LLM takes 10 minutes to respond, you're still at risk of something dropping it

Extra scheduler thread – but one thread per connection isn't a big deal for personal use

When Should You Use This?

Use this approach if:

  • You're building a personal MCP server or small-scale service
  • Most requests complete within a few minutes
  • You don't want to mess with websockets
  • Your hosting provider has aggressive idle timeouts

Consider websockets instead if:

  • You need bidirectional communication
  • You expect long-lived connections (hours)
  • You have high traffic and want better connection multiplexing

Honestly, for most personal MCP servers, this approach is perfect. It's simple, it works, and you can implement it in 10 minutes. I've been running it for a month now – zero dropped connections since deploying.

What I Learned

After 86 outages and three days of debugging, here's what I'll remember:

  1. SSE connections are idle connections – if you expect pauses, send heartbeats. Period.
  2. Proxies exist to drop your connections – it's not personal, it's their job. Don't give them a reason.
  3. Local ≠ production – your proxy will break things that work fine locally. Test early.
  4. Simple solutions beat complex ones – heartbeats are simpler than websockets for this use case.
  5. Document your proxy config – I'm putting this in my repo README so the next person (me) doesn't spend three days debugging this again.

Your Turn

Have you built an MCP server? Did you run into connection dropping issues? Did you find a different solution? I'd love to hear about it in the comments.

And if this saved you three days of debugging – drop a ❤️! It helps other people find this article.


Check out the full implementation on GitHub: kevinten10/Papers – it's open source and you can copy-paste the heartbeat code directly into your own project.