How Session Replay Service Works: Our Free 200-Page Book

This announcement covers the first edition of How Session Replay Service Works, published by the LogNroll team in September 2026 under a CC BY 4.0 license. The book is free — we would rather it reach every engineer who works with replay than sit behind a paywall.

Why we wrote a book

Session replay has quietly become a standard feature of product analytics. Watch a real user's session, spot where they hesitated, replay the exact moment a checkout broke — the pitch is familiar, and thousands of teams now run replay in production. But here is the strange part: almost all of those teams only ever see the player. The thin slice of the system that renders a session back to a human.

Everything underneath is at least as interesting. A recorder that captures every click, input, navigation, and network response without freezing the page. A wire format compact enough to ship thousands of events per session. Transport that survives a user closing their laptop mid-upload. A receiver that ingests a firehose without dropping sessions. A bus that decouples intake from storage. An archiver that moves finished sessions to cheap object storage. A processor that turns raw events into errors, heatmaps, and scroll patterns. And, at the end, a player that reassembles the whole thing faithfully enough that watching feels like being there.

Most teams will never need to build that pipeline themselves — which is exactly why we wrote the book. When a tool is this widely used and this thinly understood, the whole category suffers: buyers compare dashboards instead of architectures, engineers trust replays they cannot reason about, and teams that do build replay start from scratch. Writing down how a real system works — trade-offs, failure modes, and all — raises the bar for everyone, including us.

What's inside

The book is 21 chapters across six parts, plus five appendices. Each part follows one stage of the pipeline, in the order a session actually travels:

The appendices are reference material rather than narrative — the protobuf schema, the storage layer, the configuration reference, a glossary of replay terms, and further reading for every part:

The story thread: Candlewood Books

Threaded through all six parts is the story of Candlewood Books, a small online bookstore that has always prided itself on knowing its customers. One Tuesday the checkout starts failing for a handful of shoppers, and the store's biggest wholesale client threatens to take its business elsewhere if the problem is not found fast. Candlewood has error logs and server metrics, but the failing users never log what they felt: the button that did nothing, the form that silently refused, the purchase path that worked for some browsers and broke for others.

As each part of the book introduces a piece of the platform, it also shows why that piece exists for Candlewood: the masking rule that keeps a card number out of the recording, the retry that saves a session when the network dips, the receiver that never drops a batch, the processor that turns a handful of broken checkouts into a signal the whole company can act on.

Free, in every format

The entire book is free — no account, no email gate, no paywall. Read it in the browser, download the PDF for offline reading, or load the EPUB into the reader you already carry. Pick a format:

The book is licensed CC BY 4.0, which means you are free to share it, quote from it, print it, and build on it — as long as you credit LogNroll. If it helps one team, send it to the next one.

Where it comes from

This is not a marketing document styled as a book. It was written from LogNroll's actual source code, and the code is the point. When the book says the recorder batches events by size and by time before sending them, that is the code path running in production today. When it walks through the protobuf wire format, you are reading the schema the receiver validates events against.

That honesty extends to the roadmap. Chapter 20 is an honest look at where the platform goes next — what is being improved, what is being reconsidered, and what we know we do not know yet. Most writing about systems happens after the fact, once the rough edges have been sanded down. We wrote this book while the system is still being built, so the rough edges are visible too. We think that makes it more useful, not less.