From 664417c6dafe298bb0a8eb9e87208c2cd1f98ed6 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Mats=20T=C3=B6rnberg?= <mats.tornberg@gmail.com>
Date: Tue, 22 Sep 2026 11:21:52 +0200
Subject: [PATCH] fix(server): raise keepAliveTimeout above ingress-nginx's
 upstream idle timeout
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Node's default keepAliveTimeout (5s) is shorter than ingress-nginx's
default upstream keepalive idle timeout (60s). Under that mismatch,
nginx can forward a request onto a pooled connection Node has *just*
closed, which the browser sees as a 502 — and it never shows up in
this process's own traces because Node never received the request.
/api/rooms is polled every 5s from idle clients, landing almost
exactly on that 5s boundary, which is why it was the one showing up
for external users.

Also add abort/slow-request logging (tagged with the active trace ID
so it can be pivoted to from RUM) and a clientError handler, so a
genuine backend-side failure is now visible instead of indistinguishable
from the proxy-side race.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
---
 server.js | 33 +++++++++++++++++++++++++++++++++
 1 file changed, 33 insertions(+)

diff --git a/server.js b/server.js
index b3d05c5..11151fe 100644
--- a/server.js
+++ b/server.js
@@ -7,6 +7,11 @@ const fs = require('fs');
 const path = require('path');
 const crypto = require('crypto');
 const { WebSocketServer } = require('ws');
+const { trace } = require('@opentelemetry/api');
+
+function traceId() {
+  return trace.getActiveSpan()?.spanContext().traceId;
+}
 
 const PORT = Number(process.env.PORT || 8080);
 const HOST = process.env.HOST || '0.0.0.0';
@@ -92,6 +97,20 @@ function serveStatic(req, res) {
 }
 
 const server = http.createServer((req, res) => {
+  const start = process.hrtime.bigint();
+  const { method, url } = req;
+  res.on('close', () => {
+    const ms = Number(process.hrtime.bigint() - start) / 1e6;
+    // res.writableEnded false here means the client (or ingress) dropped the
+    // connection before we finished writing — the smoking gun for a 502 that
+    // originates on the proxy side rather than in this process.
+    if (!res.writableEnded) {
+      console.warn(`[http] ABORTED ${method} ${url} after ${ms.toFixed(1)}ms (status=${res.statusCode}) trace=${traceId()}`);
+    } else if (ms > 1000) {
+      console.warn(`[http] SLOW ${method} ${url} ${res.statusCode} ${ms.toFixed(1)}ms trace=${traceId()}`);
+    }
+  });
+
   if (req.url === '/healthz' || req.url === '/readyz' || req.url === '/health') {
     res.writeHead(200, { 'Content-Type': 'text/plain' });
     res.end('ok');
@@ -110,6 +129,20 @@ const server = http.createServer((req, res) => {
   serveStatic(req, res);
 });
 
+// Node's default keepAliveTimeout (5s) is shorter than ingress-nginx's default
+// upstream keepalive idle timeout (60s). Under that mismatch, nginx can pick a
+// pooled connection Node has *just* closed and forward a request onto it,
+// which comes back to the browser as a 502 — invisible from inside this
+// process because Node never saw the request. headersTimeout must stay above
+// keepAliveTimeout (Node enforces this).
+server.keepAliveTimeout = 65_000;
+server.headersTimeout = 66_000;
+
+server.on('clientError', (err, socket) => {
+  console.warn(`[http] clientError: ${err.message} (code=${err.code})`);
+  if (socket.writable) socket.end('HTTP/1.1 400 Bad Request\r\n\r\n');
+});
+
 // ---------------------------------------------------------------------------
 // Signaling
 //   client -> server: {type: offer|answer|candidate, to}  {type: pos|state}