Prompt
I'm building a gateway for microservices with retries and circuit breaking, what do people use?
Latest observation
Oct 8, 2026
OpenAI APIWeb search: off
People usually build this with an API gateway or a service mesh depending on where they want retries/circuit breaking to live.
Common choices
API gateway products
Good if you want centralized edge traffic management:
- Kong Gateway
- NGINX / NGINX Plus
- Envoy Gateway
- Apache APISIX
- Traefik
- KrakenD
- Ambassador Edge Stack
These often support:
- retries
- timeouts
- rate limiting
- auth
- routing
- circuit breaking or upstream health checks
Service mesh
Good if you want service-to-service resilience across the cluster:
- Istio (uses Envoy)
- Linkerd
- Consul service mesh
- AWS App Mesh
These are commonly used for:
- retries
- circuit breaking
- mTLS
- traffic shaping
- observability
Libraries in the application
Sometimes teams do retries/circuit breaking in code using:
- Resilience4j (Java)
- Polly (.NET)
- Hystrix is mostly legacy now
- language-native HTTP client retry logic
- gRPC retry policies
This is useful when behavior is business-specific or you want control per call path.
What’s most common in practice?
- For internet-facing gateway: Kong, NGINX, APISIX, Traefik
- For internal microservice resilience: Istio/Envoy, Linkerd
- For app-level control: Resilience4j/Polly
Recommendation
If you want a straightforward modern setup:
- Envoy-based gateway if you want strong L7 traffic control
- Kong if you want a popular, plugin-rich gateway
- Istio if you need retries/circuit breaking across many services in Kubernetes
If you want, I can also give you:
- a feature comparison table, or
- a recommendation by stack like Kubernetes, AWS, Java, Go, .NET, etc.