Building a system that never stops: New Relic at scale
summary
New Relic handles over 200 services, more than 2 PB SSD and traffic growth that made horizontal sharding with a proxy tier fail by multiplying failure points; that experiment rolls back. The successful move is a Kafka-driven async pipeline rolled out in September, which decouples teams and enables safe canary consumers on full production traffic. Operational changes include a gatekeeper deploy system, embedded SREs, heavy use of New Relic and custom Insights events, and team-owned deploys and on-call.
AI-written summary · about
chapters (9)
- 0:00Start
- 1:11The Problem With Scale
- 2:52Lesson 1
- 3:29Lesson 2
- 5:40Lesson 3
- 6:49Lesson 4
- 7:20Lesson 5
- 7:32Skin in the Game Means
- 10:13Lesson 7
related talks
44mIntroduction to application scaling
10mNew Relic Sponsored Session - Mike Panchenko, New Relic
14mObservability simplified: Empowering engineers to make better software
10mon their new analytics platform
42mPlanning To Fail
31mHow We Went From Two Major Outages to 99.98% Reliability in 6 Months
17mCloud-Native O11y: Building Scalable Enterprise Systems
15mHow Our SREs Safeguard Nanosecond Performance—at Scale—in an Environment Built to Fail