Grow with Less Pains - Meshing From Monolith to Microservices
summary
Company moves a monolith to ~40 services and a service mesh that handles billions of requests per day using Linkerd + Nginx edge, Consul then SRV-based discovery backed by a Dynamo fleet DB, and Prometheus for observability. Built custom Linkerd plugins for cookie/JWT authentication and a context-sensitive experiment decider that pins users to canaries. Sidecars enforce connection pooling and circuit breaking, cutting connection churn and reducing p99 latency; the stack ships ~500k metrics every 10s and runs across EC2, ECS, Kubernetes and Lambda.
AI-written summary · about
chapters (28)
- 0:00Introduction
- 0:45About Leo
- 1:08What is a startup
- 2:08Optimize your iteration progress
- 3:17Our story
- 4:44What is a Caddy
- 6:25Architecture Evolution
- 7:40Agenda
- 8:14Edge Service
- 9:56Link ID
- 10:23URL normalization
- 11:59URL translation
- 15:11Authentication Logic
- 16:31Experiment Decider
- 19:23High Level Overview
- 23:04Microservices Overview
- 26:48Progressive Rollout
- 28:08Service vs Service
- 28:39Discovery v1
- 31:01SSR
- 31:39SRV
- 32:29Fleet EP
- 34:05Consistency
- 35:20Results
- 36:05Latency Graph
- 36:53Discovery
- 37:50Graph
- 38:09Conclusions
related talks
34mFrom Monolith to Microservices with Kubernetes and Linkerd
35mNavigating the Service Mesh Ecosystem
28mHow a Service Mesh Helped Us Build Production Cloud-Native Services
40mThe World after Microservice Migration w/ Service Mesh & Contract Testing
38mAsk Me Anything About Service Mesh
22mBuilding + operating a service mesh at a mid-size company
24mFrom Monolith to Microservices: A Journey into the Mesh
18mRe-Architecting Monoliths Into Kubernetes Microservices at Million-User Scale: Hard...