Opens in a new tab
Guide

Where to start with Google’s free SRE books

In brief

Choose a free Google SRE book and chapter for a current alerting, on-call or postmortem problem, then try one practice with your own team.

4 min read

Is this the right starting point?
Before you read
No prior SRE reading required; a current service or team question helps select a useful starting chapter.
What you will take away
Choose among Google’s SRE books and turn a relevant chapter into a small local practice trial with explicit assumptions.
Sources
An open reference book links to a rehearsal board with two role markers and a route that stops before an external receiver.
On this page4 sections

Start with a question from your own service. If an alert woke someone without giving them useful work to do, open the workbook chapter on alerting. If the responder knew what needed to change but could not reach the person who could approve it, start with on-call coverage. Google’s books page links to free online editions of the original Site Reliability Engineering, The Site Reliability Workbook and Building Secure & Reliable Systems. It lists a separate purchase option for the second edition of Site Reliability Engineering.

Which book to open first

Start with the original Site Reliability Engineering if you are learning the vocabulary or want to understand why teams use practices such as reliability targets and error budgets. Open The Site Reliability Workbook when you already know the concept and want to see how teams put it to work. Choose Building Secure & Reliable Systems when a design decision affects both security and reliability, such as how to control access to a recovery path people may need urgently.

Write down the decision you need to make and keep it beside the chapter. As you read, mark the assumptions the example depends on. A practice that already fits your service is worth adopting with its original attribution. Where an assumption differs, adapt that part and see what the change does.

Three places to start with work already on your desk

What brought you hereWhere to readWhat to bring
The team gets pages with no clear responseAlerting on SLOs in the workbookA recent page and the service level objective, or reliability target, it was meant to protect. Ask what would become harder if the responder waited.
The first responder needs help they cannot reachOn-Call in the workbookThe real route from page to specialist or approver, including who covers when the primary is unavailable.
Postmortems keep producing the same unfinished actionsPostmortem Culture: Learning from Failure in the original bookOne review and its action list. Check who accepted the work and what would show that it reduced the risk.

A role in the book needs a person on your rota

Imagine a team that reads about incident command, names a commander and practices coordinating two parallel investigations. The new arrangement helps people see who is handling each question. Then a drill reaches a mitigation that needs an external owner’s approval, and nobody can reach that owner. The team has learned how to coordinate the request. It still needs someone who can decide it.

Find out whether the example assumed an available approver, a permission the local role does not have or telemetry the team cannot see. You can do the same with an error budget policy. If the policy calls for pausing a release, follow that instruction far enough to see who can actually pause it. The team may be able to use most of a practice as written and still need to solve one dependency before it works.

Turn a chapter into a small local experiment
A service question selects a relevant chapter. Review its assumptions, adapt roles, access and capacity, then run a small experiment and record what transferred and what did not.
Bring a current service question to a chapter. Check what the practice assumes, then try a small local experiment.
Read diagram description

Start with a service question and read the relevant chapter. Identify what the practice needs from the people, access and capacity available locally. Try it at a small scale and record what worked and what still needs to change.

Try the part that depends on your team

Run the mitigation drill again with the primary approver marked unavailable. Give the incident lead the same request and see whether the proposed fallback reaches someone authorized to accept that specific risk. Record where the request went, whether someone took it up and any access or coverage still missing. A drill can show that the contact path works without applying a production change.

For a noisy page, the smaller experiment might be to review which recent notifications actually required immediate action. For an unfinished postmortem action, ask the owner what evidence would let them close it and what keeps them from getting it. Bring the result back to the chapter. An escalation paragraph reads differently once you know precisely where your own route stops.

Sources & context

Sources linked in this article. Read alongside the author’s analysis; a citation does not independently verify a publisher’s claims.

Report an error or outdated detail

More on Google

All Google coverage

Keep reading

Where to go next