Backend process handles graceful shutdown

A backend process that ignores SIGTERM drops in-flight requests immediately when the orchestrator (Kubernetes, Railway, Fly.io, Heroku) restarts or redeploys it. Users mid-checkout, mid-upload, or mid-auth flow receive hard errors with no recovery path. CWE-460 (improper cleanup on thrown exception) encompasses improper shutdown handling. Platforms send SIGTERM and wait a grace period (typically 30 seconds) before SIGKILL; failing to handle it means every deploy causes unannounced request failures that are invisible in your error tracking.

Register SIGTERM and SIGINT handlers in your server entry point. Stop accepting new connections immediately, let in-flight requests finish, then close the database pool and exit.

// src/server.ts
const server = app.listen(3000)

async function shutdown(signal: string) {
  console.log(`${signal} received — shutting down`)
  server.close(async () => {
    await pool.end()
    process.exit(0)
  })
  // Force exit if drain takes too long
  setTimeout(() => process.exit(1), 30_000)
}

process.on('SIGTERM', () => shutdown('SIGTERM'))
process.on('SIGINT', () => shutdown('SIGINT'))

Test graceful shutdown locally: run your server, send kill -SIGTERM <pid>, and verify in-flight requests complete before the process exits.