FluxGate

I use reverse proxies daily and wanted to know what is inside one: streaming, balancing, health checking and rate limiting, built without a framework doing the work.

• A streaming reverse proxy that pipes requests and responses without buffering them.
• Round-robin load balancing across three backend services that skips unhealthy nodes.
• Periodic health checks that mark backends up or down and bring them back when they recover.
• Redis-backed rate limiting at 10 requests per 10 seconds per IP, with TTL-based reset.
• A metrics endpoint exposing request counts, errors, active connections and backend health.
• Proxy, backends and Redis run as containers on a private Docker network.

1. Client

   →

2. FluxGate

   rate limit, metrics

   →

3. Health-aware round robin

   →

4. Backend x3

Stream, do not buffer

Bodies are piped through, which keeps memory flat regardless of payload size.

Shared rate-limit state

Counters live in Redis, so limits would still hold if more than one proxy instance ran.

Failover without intervention

Health checks drive routing, so a dead backend drops out of rotation and returns on its own.

• Node.js
• Redis
• Docker
• Docker Compose
• HTTP

A learning project that made proxying, health checking and rate limiting concrete instead of something I only configure.