Where to start with Google’s free SRE books
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.
SRE
Reliability engineering, from service objectives to operational judgment.
Make the canary test the change before sending more traffic
A quiet canary can still miss the operation that changed. Compare meaningful traffic, check service health and test withdrawal before expanding.
How to calculate mean time to detect from incident records
Calculate MTTD from three incident records. Keep customer reports, uncertain starts and the number of timed incidents visible.
A blameless review starts with what the responder saw
Two workers share a name. A blameless review can show why one looked like the other and test whether a change helps the next responder choose.
Write a post-incident review the next responder can use
The team documented the rollback command. The next responder still needs to know if the old application can read the new data. A review should carry that question forward.
When a shorter configuration file makes a service harder to run
Moving customer overrides tidies a file. Retiring an unused override can simplify the service. Follow the settings through investigation, migration and recovery to tell the difference.
Why the latency chart can improve while more customers time out
A faster latency line can lose the requests that timed out. Follow the counts, gauges and distributions back to the work they measured.
What an empty log search can tell you
A worker restarts mid-job. To decide whether to retry, the responder needs to know what the worker recorded and what made it into search.
An on-call rotation needs more than an evenly filled calendar
Follow a page through backup, specialist help, access and recovery to see what the whole on-call team can actually cover.
A runbook that helps the next responder choose
Use a downloadable SRE runbook template and a database-connection example to connect observations, permitted actions and recovery checks.
How a container image becomes a working service
Follow a packaged program through Kubernetes placement, health checks and resources, then trace the customer job it still needs to finish.
When a postmortem action starts helping the responder
A backup alert reaches the right person, but their access stops the response. Follow that journey to decide what the improvement still needs.