AWS Lambda Monitoring: Cold Starts, Timeouts, and External Uptime

Four Golden Signals for Lambda

1. Error Rate

javascript
exports.handler = async (event, context) => {
try {
return await processEvent(event);
} catch (err) {
console.error(JSON.stringify({ event: 'lambda_error', fn: context.functionName, error: err.message }));
throw err;
}
};


2. Cold Start Detection

`javascript
const isColdStart = !global._initialized;
if (!isColdStart) global._initialized = true;

console.log(JSON.stringify({ cold_start: isColdStart, fn: context.functionName }));
`

3. Timeout Warning

javascript
const warningTimeout = setTimeout(() => {
console.warn('approaching_timeout: ' + context.getRemainingTimeInMillis() + 'ms left');
}, context.getRemainingTimeInMillis() - 5000);
try {
return await doWork(event);
} finally {
clearTimeout(warningTimeout);
}


4. External HTTP Monitoring

For Lambda behind API Gateway, monitor the HTTP endpoint with Vigilmon:

javascript
exports.health = async () => ({ statusCode: 200, body: JSON.stringify({ status: 'ok' }) });


Set Vigilmon to check /health every minute and alert on status != 200.

CloudWatch Alarms Worth Setting

Metric	Threshold
Errors	> 1%
Throttles	> 0
Duration	> 80% of timeout