Skip to main content

Observability in Serverless

Overview of tracing, metrics, logging, and alerting strategies for serverless applications, with multi-cloud considerations and recommended tooling.

Simple Explanation​

What it is​

Observability is how you see what your serverless system is doing through logs, metrics, and traces.

Why we need it​

When something breaks, you need evidence, not guesses. Observability makes that possible.

Benefits​

  • Faster diagnosis of errors and slow requests.
  • Clear visibility across services and regions.
  • Better reliability because problems are found early.

Tradeoffs​

  • More setup for dashboards and alerts.
  • Ongoing cost for log storage and tracing.

Real-world examples (architecture only)​

  • Trace shows slow database call -> Fix query.
  • Alert triggers on error spike -> Rollback.

Trace flow Alert flow

What This Lesson Covers​

  • Logging strategy and structured logs
  • Metrics that matter (latency, errors, saturation)
  • Distributed tracing and correlation IDs
  • Alerting thresholds and SLOs
  • Multi-cloud visibility patterns

Core Concepts​

  1. Logs: Details of a single event
  2. Metrics: Aggregated trends over time
  3. Traces: End-to-end request flow
  4. SLOs: Reliability targets
  5. Alerting: Automated responses to anomalies

Structured Logging (Python)​

import json
from datetime import datetime


def log_event(level, message, **data):
	print(json.dumps({
		"timestamp": datetime.utcnow().isoformat(),
		"level": level,
		"message": message,
		**data,
	}))


def handler(event, context):
	log_event("INFO", "Request received", requestId=context.aws_request_id)
	# ...

Correlation IDs​

Add a request ID to every log line so you can trace one request across services.

def get_request_id(event, context):
	return event.get("headers", {}).get("x-request-id") or context.aws_request_id

Metrics That Matter​

  • Error rate (percentage of failed requests)
  • Latency (p50, p95, p99)
  • Throttles (rate limiting)
  • Cost drivers (duration, memory, external calls)

Tracing Strategy​

  • Sample a percentage of requests
  • Always trace errors
  • Add annotations (userId, region, tier)

Multi-Cloud Observability​

Patterns:

  • Standardize log format (JSON)
  • Use consistent metric names
  • Mirror dashboards across clouds

Alerting & SLOs​

Define reliability targets and alert on breaches:

  • API read: p99 < 100ms
  • API error rate: < 1%
  • Queue delay: < 2 minutes

Project​

Design an observability plan for a serverless API.

Deliverables:

  • List the logs you will capture
  • Define 3 key metrics and alert thresholds
  • Describe how you will trace a request end-to-end

Email your work to [email protected].

References​