Fix /health/ready res.status bug + add comprehensive health endpoint tests
CI / Test & Lint (push) Has been cancelled
CI / Security audit (push) Has been cancelled

The readiness probe was crashing with 'res.status is not a function' because
asyncHandler(async (req, res) => {...}, 'health-ready') was called directly,
but asyncHandler's signature is (logError, fn, context) — first arg is the
logger, not the handler. The fix uses boundAsyncHandler like all other routes
in the file do.

Added 8 unit tests for both /health/live and /health/ready:
- live always 200 (liveness ≠ readiness)
- ready returns 503 when config/services/docker fail
- no 'res.status is not a function' crash when dependencies fail
- all 4 check keys present in response

Also added MONITORING_PUBLIC env var (defaults true) and the new health
endpoints to PUBLIC_ROUTES so k8s probes can hit them without auth.
This commit is contained in:
Hermes
2026-06-10 20:35:27 -07:00
parent e40cb35011
commit 264de9644c
3 changed files with 285 additions and 0 deletions
+80
View File
@@ -627,6 +627,86 @@ async function createApp() {
res.json({ status: 'ok', timestamp: new Date().toISOString() });
});
// Liveness probe — "is the process alive?"
// Always returns 200 unless the Node.js event loop is completely blocked.
// Used by k8s/Docker to decide whether to RESTART the container.
// DO NOT add dependency checks here — those belong in /health/ready.
app.get('/health/live', (req, res) => {
res.json({ status: 'alive', uptime: process.uptime() });
});
// Readiness probe — "is the app ready to serve traffic?"
// Checks critical dependencies: Docker daemon, Caddy admin API, config file.
// Returns 200 with details if all OK, 503 with failed components otherwise.
// Used by k8s/Docker to decide whether to ROUTE TRAFFIC to this instance.
app.get('/health/ready', boundAsyncHandler(async (req, res) => {
const checks = {};
let allOk = true;
// Check 1: Config file readable
try {
const fs = require('fs');
if (fs.existsSync(config.CONFIG_FILE)) {
fs.readFileSync(config.CONFIG_FILE, 'utf8');
checks.configFile = { ok: true };
} else {
checks.configFile = { ok: false, error: 'Config file not found' };
allOk = false;
}
} catch (e) {
checks.configFile = { ok: false, error: e.message };
allOk = false;
}
// Check 2: Services file readable
try {
const fs = require('fs');
if (fs.existsSync(config.SERVICES_FILE)) {
fs.readFileSync(config.SERVICES_FILE, 'utf8');
checks.servicesFile = { ok: true };
} else {
checks.servicesFile = { ok: false, error: 'Services file not found' };
allOk = false;
}
} catch (e) {
checks.servicesFile = { ok: false, error: e.message };
allOk = false;
}
// Check 3: Docker daemon reachable
try {
const docker = require('dockerode')();
await docker.ping();
checks.docker = { ok: true };
} catch (e) {
checks.docker = { ok: false, error: e.message };
allOk = false;
}
// Check 4: Caddy admin API reachable
try {
const caddyUrl = config.CADDY_ADMIN_URL || 'http://localhost:2019';
const controller = new AbortController();
const timeout = setTimeout(() => controller.abort(), 3000);
const response = await fetch(`${caddyUrl}/config/`, {
signal: controller.signal
});
clearTimeout(timeout);
checks.caddy = { ok: response.ok, status: response.status };
if (!response.ok) allOk = false;
} catch (e) {
checks.caddy = { ok: false, error: e.message };
allOk = false;
}
const body = {
status: allOk ? 'ready' : 'not-ready',
timestamp: new Date().toISOString(),
checks
};
res.status(allOk ? 200 : 503).json(body);
}));
// Lightweight probe endpoint
app.get('/probe/:id', boundAsyncHandler(async (req, res) => {
const id = req.params.id;