httpecho

A simple HTTP debug out micro service. Because there's not enough out there.

It listens on PORT and reflects whatever it receives — any method, any path — back as JSON:

{
  "method": "POST",
  "path": "/a/b",
  "headers": { "Content-Type": ["application/json"] },
  "parameters": { "x": ["1", "2"] },
  "body": "{\"targets\":[\"24.10.22.118\"]}"
}

parameters holds the query string only, so body is always the raw bytes exactly as sent.

Every request is answered with 200, whatever the method or path, except in two cases:

• A body that cannot be read (a connection dropped mid-request, for example) answers 400 with the same application/json content type:

  { "error": "unable to read request body" }

• The router cleans the path before the handler sees it. /a/../b is echoed as /b, and a path with an empty segment such as /a//b is answered with a 307 redirect to /a/b instead of being echoed. Everything else — including trailing slashes, percent-encoding and unusual methods — reaches the handler untouched.

Configuration

Read from the environment, or from a .env file in the working directory. Real environment variables always take precedence over .env.

Variable	Required	Default	Description
PORT	no	8080	TCP port
OTEL_EXPORTER_OTLP_ENDPOINT	no	—	OTLP collector, as collector:4317 or http://collector:4317; omit to disable telemetry
OTEL_EXPORTER_OTLP_INSECURE	no	false	Set true for plaintext gRPC (typical for a local collector)
OTEL_SERVICE_NAME	no	httpecho-service	Service name on all telemetry signals
OTEL_RESOURCE_ATTRIBUTES	no	—	Comma-separated key=value pairs added to every signal (e.g. deployment.environment=production)
OTEL_TRACES_SAMPLER	no	parentbased_always_on	Sampling strategy

Telemetry

Traces, logs and metrics are exported over OTLP/gRPC to a collector, which forwards them on — Jaeger for traces, Loki for logs, Prometheus for metrics. Without OTEL_EXPORTER_OTLP_ENDPOINT nothing is exported and no exporters are created; logs still go to stderr, as they do when export is enabled.

docker run --rm -p 8080:8080 \
  -e OTEL_EXPORTER_OTLP_ENDPOINT=collector:4317 \
  -e OTEL_EXPORTER_OTLP_INSECURE=true \
  -e OTEL_SERVICE_NAME=httpecho \
  httpecho:dev

The endpoint may be a bare collector:4317 or the specification's URL form http://collector:4317. With a URL the scheme selects transport security; with a bare address OTEL_EXPORTER_OTLP_INSECURE=true does. Setting it to true always forces plaintext.

Signals are batched, not sent per request: metrics leave every 60 seconds (OTEL_METRIC_EXPORT_INTERVAL), traces and logs on their batch delays (OTEL_BSP_SCHEDULE_DELAY, OTEL_BLRP_SCHEDULE_DELAY), and everything pending is flushed on shutdown. A service that has just started looks quiet in Prometheus for up to a minute; that is the interval, not a broken pipeline.

An unreachable collector never delays startup and never affects a request. On shutdown the final flush waits up to 10 seconds for the collector before giving up; the failure is logged and the process still exits 0, since telemetry is optional.

Running

go run ./cmd/httpecho

docker build -t httpecho:dev .
docker run --rm -p 8080:8080 httpecho:dev

Usage

curl -s "http://localhost:8080/validate" \
  -H "X-Api-Key: <key>" \
  -H "X-Context: profile" | jq

curl -s -X POST "http://localhost:8080/a/b?x=1&x=2" \
  -H "Content-Type: application/json" \
  -d '{"source":["169.1.1.34"]}' | jq

Development

go vet ./...
go test -race ./...