Envoy Project Intro – Matt Klein & Jose Nino, Lyft
summary
Envoy is deployed as a front-edge cluster and a sidecar per host service mesh at Lyft, supporting hundreds of services while the org grows from ~500 to ~1,000 engineers. Configuration moves from giant JSON files to an XDS-based control plane using XDS (RDS/CDS/LDS); artifacts are generated from service manifests, stored in S3, cached and served by a manager that the team plans to align with go control plane. Deploys use Git PRs, CI, chat-triggered Jenkins jobs, staged canaries, and twice-weekly Envoy upstream merges; per-hop latency is typically <=1 ms.
AI-written summary · about
chapters (10)
- 0:00Start
- 0:43My time at Lyft
- 2:27The Story
- 2:46Why is this important?
- 7:58Configuration Management - The Present
- 12:41Documentation
- 13:35Making a Change
- 14:10Deployment
- 15:16Versioning
- 24:10Configuration Management. The Present
related talks
35mEnvoy Maintainer Q+A
36mBringing the Envoy Service Mesh to Mobile
34mIntro: Envoy
11mA decade of Envoy: From Lyft's Debugging Nightmare to the AI Infra...
33mGet to know Envoy - the modern proxy for cloud-native infrastructure - Baptiste Collard
11mWelcome + Introduction
10mLightning Talk: Envoy @ Netflix
33mIntro: Envoy