Health Check Endpoint

Problem

Implement health check endpoints that accurately report whether your server and its dependencies are ready to handle traffic. A /healthz endpoint checks that the server is alive, while /readyz verifies that all downstream dependencies (database, cache) are actually reachable.

Background

In production, orchestrators like Kubernetes use health checks to decide whether to route traffic to a pod, restart it, or wait for it to become ready. A naive health check that always returns 200 hides real problems — your server might be "alive" but unable to serve requests because the database is down. Smart health checks catch this and let the system heal automatically.

There are two types: liveness probes (/healthz) confirm the process isn't deadlocked, and readiness probes (/readyz) confirm the server can actually do work.

Requirements

• GET /health — basic liveness check, always returns 200 (used by the harness for startup detection)
• GET /healthz — liveness probe: returns 200 if the server process is running
• GET /readyz — readiness probe: returns 200 only if both PostgreSQL and Redis are reachable
• When a dependency is down, /readyz returns 503 with details about which dependencies failed
• The response body must include a checks object showing the status of each dependency
• Health checks must complete within 2 seconds (use connection timeouts)

Your Server

• Your server receives PORT (default 3000), DATABASE_URL, and REDIS_URL environment variables
• PostgreSQL and Redis are provided as sidecars — they are up when your server starts
• The harness may stop a sidecar during testing to verify your health check detects the failure

API Contract

`GET /health`

Basic health check (used by harness for startup detection).

• Status: 200
• Body: { "status": "ok" }

`GET /healthz`

Liveness probe — is the process alive?

• Status: 200
• Body: { "status": "ok" }

`GET /readyz`

Readiness probe — can the server handle requests?

When all dependencies are healthy:

• Status: 200
• Body:

{
  class="text-pass">"status": class="text-pass">"ok",
  class="text-pass">"checks": {
    class="text-pass">"postgres": { class="text-pass">"status": class="text-pass">"ok" },
    class="text-pass">"redis": { class="text-pass">"status": class="text-pass">"ok" }
  }
}

When one or more dependencies are down:

• Status: 503
• Body:

{
  class="text-pass">"status": class="text-pass">"error",
  class="text-pass">"checks": {
    class="text-pass">"postgres": { class="text-pass">"status": class="text-pass">"ok" },
    class="text-pass">"redis": { class="text-pass">"status": class="text-pass">"error", class="text-pass">"message": class="text-pass">"Connection refused" }
  }
}

Useful APIs

The starter code provides a pool (pg.Pool) and redis (ioredis) client. Here's how to use them:

class=class="text-pass">"text-fg-subtle">// PostgreSQL: get a client from the pool and run a test query
const client = await pool.connect();
try {
  await client.query(class="text-pass">'SELECT 1');
  class=class="text-pass">"text-fg-subtle">// database is reachable
} catch (err) {
  class=class="text-pass">"text-fg-subtle">// database is down
} finally {
  client.release(); class=class="text-pass">"text-fg-subtle">// always release the client back to the pool
}

class=class="text-pass">"text-fg-subtle">// Redis: send PING command
await redis.ping()  class=class="text-pass">"text-fg-subtle">// returns class="text-pass">'PONG' if healthy, throws if down

class=class="text-pass">"text-fg-subtle">// Run multiple async checks in parallel (doesnclass="text-pass">'t short-circuit on failure)
const results = await Promise.allSettled([checkPostgres(), checkRedis()])
class=class="text-pass">"text-fg-subtle">// results[i].status === 'fulfilledclass="text-pass">' → results[i].value
class=class="text-pass">"text-fg-subtle">// results[i].status === 'rejectedclass="text-pass">'  → results[i].reason (no .value!)

class=class="text-pass">"text-fg-subtle">// Set a timeout on a promise
const withTimeout = (promise, ms) =>
  Promise.race([promise, new Promise((_, reject) => setTimeout(() => reject(new Error('timeout')), ms))])