Transforming Production Readiness
summary
Elastic builds an SLO observability product that monitors around 900,000 SLOs across 180+ clusters and 60 regions, ingesting petabytes and ~500 TB/day, using Elasticsearch transforms and cross-cluster search so only summaries cross regions. Thousands of engineers move onto on‑call via a phased rollout (private → public beta to GA), enforced production readiness checklist with required SLOs, runbooks and routing. The system surfaces high-escalation bugs, scales horizontally, hit many painful bugs during deployment but proves battle-tested and improves ownership and incident response.
AI-written summary · about
chapters (17)
- 0:00Introduction and context
- 2:05Shift in service model
- 4:17Operational responsibility
- 7:21Architectural improvements
- 9:49Scaling the SRE organization
- 10:57Organizational change
- 12:49Service boundaries
- 14:08Production readiness
- 16:47Phased product launches
- 18:49Incident management
- 21:23Empowering teams
- 22:59Developer tools and automation
- 26:33Observability as a service
- 27:55SLOs as a common language
- 28:58Designing for scale
- 31:59Operational excellence
- 34:22Key takeaways
related talks
31mServerless Observability: a case study of SLOs
47mCultivating Production Excellence
29mCultivating Production Excellence
40mManaging to Your SLO Amidst Chaos
29mReducing Trauma in Production
25mBuilding a highly scalable system to assure Business Resilience
39mShifting Left on Production Excellence with Observability
26mIncreasing the collaboration between SREs and Developers