Heartbeat and health

Doctors understand a body through its pulse, temperature, history and symptoms. You understand a running microservice the same way — through observability: logs, metrics, traces and alarms. A service without them is like a patient who can't speak.

The medical kit

Writing a useful diary: structured logs

Write logs as JSON with consistent fields so you can search them:

import { Logger } from "@aws-lambda-powertools/logger";
const logger = new Logger({ serviceName: "waiter" });

logger.info("Order placed", { orderId: order.orderId, tableNumber: order.tableNumber, total: order.total });

Include a correlation ID (for example the order ID or request ID) in every log line, and pass it along in events — so you can follow one customer's journey across many humans.

Taking the pulse: metrics

Three kinds of numbers matter:

• Technical — invocations, errors, duration, throttles, queue depth, DynamoDB throttled requests.
• Business — orders placed per minute, payments failed, average order value.
• Experience — latency percentiles (p50, p95, p99), not just averages.

import { Metrics, MetricUnit } from "@aws-lambda-powertools/metrics";
const metrics = new Metrics({ namespace: "CafeSukoon", serviceName: "waiter" });

metrics.addMetric("OrdersPlaced", MetricUnit.Count, 1);
metrics.publishStoredMetrics();

(Powertools for AWS Lambda — available for TypeScript, Python, Java and .NET — makes logging, metrics, tracing and idempotency easy and consistent.)

Following a thought: tracing

A trace shows one request's path — API Gateway → Waiter Lambda → DynamoDB → EventBridge — with the time spent in each step. Enable X-Ray tracing on API Gateway and Lambda (tracing: lambda.Tracing.ACTIVE in CDK) to find which organ is slow.

Feeling pain: alarms that matter

Alarm on symptoms customers feel, not on every twitch:

Health checks

For serverless services there's no server to ping, but you can still:

• Expose a lightweight /health route that checks critical dependencies (memory reachable, configuration loaded).
• Run synthetic canaries that place a test order every few minutes and alarm if it fails.

The doctor's dashboard

One dashboard per human: traffic, errors, latency, queue depth, dead letters, and its key business metric. One dashboard for the whole society: the journey of an order across all humans.

Written by Bipin Singh

Senior Full-Stack Engineer · AI & AWS. I design and build serverless systems on AWS — and love explaining them simply.

Work with me