Load balancer health checks: what they check and why endpoints matter

A server is technically running but in a degraded state (database disconnected, out of memory, job queue backed up). The load balancer keeps routing traffic to it because the process is alive.

Implement a real health check endpoint that verifies actual dependencies:

// Express health check endpoint
app.get('/health', async (req, res) => {
  const checks = {};
  let status = 200;

  try {
    await db.raw('SELECT 1'); // verify database connectivity
    checks.database = 'ok';
  } catch (err) {
    checks.database = 'error';
    status = 503;
  }

  try {
    await redis.ping();
    checks.redis = 'ok';
  } catch (err) {
    checks.redis = 'degraded'; // non-fatal
  }

  res.status(status).json({
    status: status === 200 ? 'ok' : 'degraded',
    checks,
    uptime: process.uptime(),
    timestamp: new Date().toISOString()
  });
});

Nginx upstream health check config:

upstream backend {
    server 10.0.0.1:3000;
    server 10.0.0.2:3000;
}

A process-alive check only confirms the web server is accepting TCP connections. A deep health check catches application-level failures that affect real users.