Where to start with Google’s free SRE books
A chapter-level starting point for alerting, on-call coverage and postmortems, with examples of adapting the practices to local staffing and authority.
SRE
Reliability engineering, from service objectives to operational judgment.
GuidePlatform engineering
What a canary deployment actually tests
Traffic assignment determines which behavior reaches a canary. Promotion also depends on the baseline, service objectives and state left behind by rollback.
ExplainerIncident response
Calculating mean time to detect from incident records
Worked incident records explain the MTTD formula, uncertain onset times and the effect of including failures first reported by customers.
CommentaryIncident response
Why blameless reviews investigate the conditions around a mistake
A mistaken target selection shows how an incident review can explain a responder’s choice and lead to a more useful interface improvement.
GuideIncident response
Writing a post-incident review the next responder can use
A delayed rollback connects the incident timeline to the information responders had, the compatibility question they faced and the follow-up that remains.
CommentaryPlatform engineering
What KISS means for service configuration
Feature flags and customer overrides show why a shorter configuration file can leave just as much behavior to investigate and support.
ExplainerObservability
What observability metrics keep and leave out
Counters, gauges and distributions preserve different information. A misleading latency trend shows why the counted requests matter as much as the chart.
ExplainerObservability
The path from a log event to an investigation
An unfinished worker operation connects event fields, collection loss and search permissions to the responder’s next retry decision.
Staffing an on-call rotation beyond the primary shift
Primary shifts, specialist help, absence coverage and recovery time all contribute to the service an on-call calendar promises.
GuideIncident response
SRE runbook template with a database connection example
A downloadable Markdown template and a worked database-connection investigation show how observations lead to mitigation and recovery checks.
ExplainerPlatform engineering
How containers and Kubernetes turn an image into a service
Images package processes and Kubernetes reconciles their deployment. Probes, resource settings and completed jobs reveal different stages of the service.
GuideIncident response
From a postmortem finding to a working improvement
A backup-failure alert illustrates the gap between shipping a postmortem action and giving its recipient the access needed to respond.