Release It! Design and Deploy Production-Ready Software

What the book is about

Release It! starts from an uncomfortable observation: code that passes every test can still take a company down in production, because production has real load, slow dependencies, bad networks and users who do strange things. Michael Nygard tells real outage stories, such as a single unhandled exception that grounded an airline's check-in systems, and then pulls out the pattern behind each one. The second edition, published by the Pragmatic Bookshelf, keeps the stories and patterns from the first and adds material on cloud infrastructure, containers, continuous deployment and systems that change while they run.

The ideas you will keep using

• Stability antipatterns: integration points that hang, chain reactions across a cluster, cascading failures between tiers, users who behave unexpectedly, blocked threads, self-denial attacks from your own marketing, unbalanced capacities and unbounded result sets.
• Stability patterns: timeouts on every remote call, the circuit breaker that stops calling a failing dependency, bulkheads that isolate failures to one pool, steady state so disks and caches do not fill forever, fail fast, let it crash, handshaking, back pressure, load shedding and test harnesses that misbehave on purpose.
• Design for production: health checks, configuration, logging, transparency and the operational interfaces that make a system easy to run.
• Deployment and adaptation: zero-downtime releases, versioning APIs, and evolving a system without breaking its consumers.

Is it worth it?

Yes. Many of the resilience libraries and service mesh features you configure today exist because of the vocabulary this book popularised. Reading it turns those settings from magic numbers into decisions: you will know why a retry without a timeout is dangerous, and why a circuit breaker needs a fallback. It also reads well, because every chapter is anchored in a story.

Prerequisites and time

You should have built at least one backend service that talks to a database or another service over a network. The examples are mostly in Java but the lessons are language neutral. Expect three to four weeks of evenings for a careful read.

Free or paid

It is a paid book, sold as an ebook and paperback on the Pragmatic Bookshelf site and through retailers, and it is included in an O'Reilly online learning subscription. The publisher offers free sample extracts on the book page.

What you can do after it

• Review a service and point out the missing timeouts, unbounded queries and shared pools that will hurt it under load.
• Add circuit breakers, bulkheads and back pressure with a reason for each setting.
• Write a postmortem that names the failure pattern, not just the bug.
• Design deployments that do not need a maintenance window.

Common questions

Is it outdated? The tooling examples age, the failure patterns do not. The second edition already covers cloud and containers.

Should I read it before or after Google's SRE books? Before, if you write services. It explains what makes a single system fragile; the SRE books explain how an organisation runs many of them.

Best for: backend and platform engineers who want their services to survive the real world, and anyone joining an on-call rotation.