FastProxy Observability Features: Built-in pprof Profiling and Debug Endpoints

FastProxy exposes runtime profiling data through Go’s standard net/http/pprof package, automatically registering /debug/pprof endpoints for CPU, memory, goroutine, and block analysis without external dependencies.

FastProxy, from the kingson4wu/fast_proxy repository, includes production-ready observability features that allow operators to introspect running proxy instances. The codebase leverages Go’s built-in pprof tooling to expose diagnostic HTTP endpoints, enabling real-time profiling of resource consumption and concurrency behavior. These capabilities are integrated directly into the proxy binaries and require no additional configuration or third-party agents.

How FastProxy Implements Observability

Automatic pprof Handler Registration

Both the inbound and outbound proxy binaries import the Go standard library’s pprof package to enable automatic handler registration. In fast_proxy/main/inproxy/inproxy.go and fast_proxy/main/outproxy/outproxy.go, the blank import activates the side-effect registration of pprof HTTP handlers:

// fast_proxy/main/inproxy/inproxy.go
import _ "net/http/pprof"

// fast_proxy/main/outproxy/outproxy.go  
import _ "net/http/pprof"

This import ensures that the net/http/pprof init function runs, registering profiling handlers with the default serve mux. FastProxy then adapts these handlers for the fasthttp framework used throughout the project.

FastHTTP Route Configuration

The central server implementation in fast_proxy/main/common/server/server.go bridges standard net/http pprof handlers to the fasthttp router using an adaptor. The code creates a handler and maps it to the /debug/pprof path prefix:

// fast_proxy/main/common/server/server.go
pprofHandler := fasthttpadaptor.NewFastHTTPHandler(fgprof.Handler())
otherHandlers["/debug/pprof"] = pprofHandler

Incoming requests are routed based on path prefix matching. When a request path begins with /debug/pprof, the server invokes the adapted handler:

if strings.HasPrefix(path, "/debug/pprof") {
    pprofHandler(ctx)
}

This implementation exposes the full suite of Go runtime profiling data through the fasthttp server listening on the configured proxy port.

Available Observability Endpoints

FastProxy exposes the standard Go pprof endpoints under the /debug/pprof path. You can access the following runtime profiles:

• /debug/pprof/profile – CPU profiler (responds after the specified duration)
• /debug/pprof/heap – Sampling of memory allocations
• /debug/pprof/goroutine – Stack traces of all current goroutines
• /debug/pprof/block – Stack traces that led to blocking on synchronization primitives
• /debug/pprof/mutex – Stack traces of holders of contended mutexes
• /debug/pprof/allocs – Sampling of past memory allocations
• /debug/pprof/threadcreate – Stack traces that led to the creation of new OS threads
• /debug/pprof/cmdline – The command line invocation of the current program
• /debug/pprof/symbol – Symbol resolution for addresses (used by pprof tool)

Practical Usage Examples

The following commands demonstrate how to extract profiling data from a running FastProxy instance listening on localhost:8080.

Capture a 10-Second CPU Profile

go tool pprof -http=:8081 http://localhost:8080/debug/pprof/profile?seconds=10

This command downloads the profile and opens an interactive web UI at http://localhost:8081 for analysis.

Export a Heap Memory Snapshot

curl -s http://localhost:8080/debug/pprof/heap > heap.prof
go tool pprof -http=:8082 heap.prof

View Live Goroutine Stacks

curl http://localhost:8080/debug/pprof/goroutine?debug=2

Setting debug=2 returns the full goroutine stack traces in plain text, useful for detecting deadlocks or leaks.

Analyze Block Contention

go tool pprof -http=:8083 http://localhost:8080/debug/pprof/block

Source Code Reference

The observability functionality is concentrated in these specific source files within the kingson4wu/fast_proxy repository:

File	Purpose
fast_proxy/main/inproxy/inproxy.go	Blank imports net/http/pprof to activate profiling handlers for the inbound proxy.
fast_proxy/main/outproxy/outproxy.go	Blank imports net/http/pprof to activate profiling handlers for the outbound proxy.
fast_proxy/main/common/server/server.go	Configures the fasthttp route /debug/pprof and adapts pprof handlers using fasthttpadaptor.
pprof/pprof.sh	Convenience shell script for launching pprof analysis against a running instance.
pprof/wrk.sh	Load testing script that can be paired with profiling to analyze performance under stress.

Summary

• Zero-configuration profiling: FastProxy automatically exposes Go pprof endpoints via net/http/pprof imports in both inproxy and outproxy binaries.
• Fasthttp integration: The common/server/server.go file adapts standard HTTP pprof handlers to the fasthttp framework using fasthttpadaptor.
• Comprehensive runtime data: Access CPU, memory, goroutine, block, and mutex profiles through the /debug/pprof route prefix.
• Standard tooling compatible: Use go tool pprof, curl, or browser-based visualization to analyze proxy performance and resource usage.

Frequently Asked Questions

What observability protocols does FastProxy support?

FastProxy supports Go’s standard pprof profiling protocol via HTTP endpoints. It does not natively expose Prometheus metrics, OpenTelemetry traces, or structured JSON logs; observability is focused on runtime profiling through the /debug/pprof path.

How do I access profiling data in FastProxy?

Send an HTTP request to the /debug/pprof/ endpoint on the proxy’s listening port. For interactive analysis, use go tool pprof -http=:PORT http://<proxy-host>:<port>/debug/pprof/profile for CPU profiles, or curl the heap, goroutine, and block endpoints to save raw data for offline analysis.

Does FastProxy support Prometheus metrics?

No. Analysis of the kingson4wu/fast_proxy source code reveals no Prometheus client library imports or metric exposition endpoints. The repository relies exclusively on Go’s built-in pprof facilities for observability. To export metrics to Prometheus, you would need to implement a custom metrics handler or sidecar.

Can I disable observability endpoints in production?

The pprof endpoints are registered automatically through the blank import in inproxy.go and outproxy.go and the route configuration in server.go. To disable them, you would need to remove the import _ "net/http/pprof" lines and delete the /debug/pprof route registration from common/server/server.go, then recompile the binaries. There is no configuration flag to disable these endpoints at runtime.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}