How Spotify Migrated Ingress HTTP Systems to Envoy
Talkat KubeCon + CloudNativeCon North America 2019 · on Wednesday, 2019-11-20· at 10:55· Exhibit Hall AB - San Diego Convention Center, Ground Level·
summary
Spotify replaces a brittle Nginx/HAProxy perimeter with Envoy proxies and a unified control plane, migrating 130 HTTP upstreams and a web player that handles 5–10% of traffic and contributing to a perimeter that peaks >8M RPS. Migration uses Google Cloud HTTP(S) load balancers, DNS GSLB, and a custom capacity scaler to ramp traffic, a scripted config translator that covers ~75% of configs, a global panic button for quick rollback, and uncovers a few Envoy/health‑check bugs fixed during rollout.
AI-written summary · about
chapters (27)
- 0:00Introduction
- 0:12Spotify Perimeter
- 0:35Spotify optin teams
- 1:14Overview
- 1:53Spotify
- 3:07Backend
- 5:28Agenda
- 5:59Proxy
- 10:21Where are we going
- 10:56Why Envoy
- 14:18Requirements
- 16:03Stage 1 Migration
- 17:39Stage 2 Overview
- 18:43Stage 2 Strategy
- 19:20Stage 1 Setup
- 22:01Stage 2 Recap
- 28:11Bugs
- 30:16Entropy
- 31:53Mitigation
- 32:21Unified Perimeter
- 32:49Authentication Geolocation
- 33:08Quality of Infrastructure
- 33:33Expect things to go wrong
- 34:13Work iteratively
- 34:41Reduce migration surface
- 35:12Give yourself room to fail
- 35:34Outro
related talks
24mHow Spotify Migrated Ingress HTTP Systems to Envoy
31mHow We Moved Spotify To a Proxyless gRPC Service Mesh
23mBridging the gap between on-prem and cloud: a story about...
27mAdventures in Rate Limiting: Spotify’s Journey Writing a Scalable Envo...
33mGet to know Envoy - the modern proxy for cloud-native infrastructure - Baptiste Collard
28mEnvoy Project Intro
15mFrom ingress-nginx to Envoy Gateway at Zapier: The...
29mSpanning the Globe with Envoy at Stripe