Evolution of a Microservice-Infrastructure
summary
Starts from Nginx + Docker Swarm serving ~270 microservices from ~50 teams, sees frequent dropped keepalive connections and uneven container placement; replaces Nginx with Traefik to get hot reloads, JSON access logs and real‑time metrics (keeps keepalives during redeploys), then migrates orchestration to Nomad for automatic spread, self‑healing and ACLs. Migration runs live over weeks with centralized Ansible/Ansible playbooks and one required change per service: set memory limits; result is stable routing and much better observability.
AI-written summary · about
chapters (21)
- 0:00Introduction
- 1:17About the company
- 2:49History
- 6:19Services
- 7:08Recap
- 8:20Architecture
- 10:23Deployments
- 12:15Deployment tools
- 15:35Routing requests
- 18:02Nginx configuration issues
- 19:48How can we solve this
- 20:36Alternatives
- 22:39Traffic
- 26:36Promise Profile
- 30:20Nomad
- 32:56Nomad in Live Traffic
- 34:54Benefits of Nomad
- 36:57Nomad Web Interface
- 39:18What we learned
- 40:56What do you learn
- 43:09Conclusion
related talks
52mMicroservice Migration Roadmap
22mRainbow deployments with Nomad
01h 11mCrack the monolith with Consul and Nomad
37mKnee Deep in Microservices
18mMicroservices Out In the Wild
45mContinuously delivering containerized, highly available microservices
36mHigher Order Infrastructure - Microservices on the Docker Swarm
36mService Messy: Pagerduty's Service Mesh Migration