Microservices: Shackles on Your Feet

2 months ago

Microservices can complicate deployments and debugging with multiple pipelines and log streams.
Common complaints like deployment issues or scaling problems often stem from poor testing, tight coupling, or database issues, not architecture.
Microservices make sense when domains are truly independent, scaling needs differ vastly, or team size exceeds 150+ engineers.
Network calls in microservices introduce new failure modes like timeouts and retries, complicating the codebase.
Distributed transactions replace simple SQL commits with complex, error-prone async steps and event buses.
Observability tools like Jaeger, Prometheus, and centralized logs are essential before splitting into microservices.
Local development becomes cumbersome with microservices, requiring Docker and complex setups.
Modular monoliths offer a middle ground with clean boundaries within a single deployable unit.
Strangler fig pattern allows gradual migration to microservices without big-bang rewrites.
Decision to split should consider team independence, scaling needs, observability, and team bandwidth.

Hasty Briefsbeta