Microservices Anti-Patterns
summary
Advice: don’t start with microservices; begin monolithic and start boring — Instagram scaled with Postgres. concrete pitfalls: one endpoint forces scaling the whole app so extract it (example: single exceptions endpoint), spikes cause 10x overprovisioning unless you use queues like Redis to amortize load, queues can fill and crash if you underprovision. manage schema/API churn with semantic versioning (run V1 and V2 side‑by‑side), run service discovery (etcd/Consul) or a router, add correlation IDs, circuit breakers, and extensive graphs.
AI-written summary · about
chapters (19)
- 0:00Intro
- 2:26Microservices
- 3:16Overzealous Services
- 3:59Solution: Start monolithic and extract
- 4:18Exception Catcher
- 6:06Schemas everywhere
- 7:03Solution: Gatekeeper
- 7:59Lock-step deployment
- 9:02Solution: Semantic Versioning
- 11:16Spiky load between services
- 11:42Solution: Amortize via queues
- 14:06Hardcoded IPs and Ports
- 14:46Solution: Service Discovery Service
- 16:11Solution: Centralised router
- 16:28Router vs Discovery Service
- 17:04Dogpiles x 100
- 17:35Solution: Correlation IDs
- 18:17Flying Blind
- 18:59Solution: Cloud Foundry
related talks
26mCreating a scalable ecosystem of Microservices
47mEmbracing Microservices
43mMicroMonolith - Top anti-patterns of distributed systems
40m10 tips for failing at microservices
36mDo Proper Monoliths before You Do Microservices!
38mMicroservices, a bittersweet symphony
45mMicroservices: A retrospective
30mHow to Fail at Building Modern Microservices