fake-webserver

It's a Fake Webserver to generate metrics data for Prometheus load testing.

Features

• Generates configurable number of API endpoints
• Creates multiple label combinations (regions, versions) for high cardinality testing
• Simulates realistic traffic patterns with oscillating request rates
• Includes error simulation and periodic outages
• Exports Prometheus metrics at /metrics endpoint

Usage

Command-line flags

./fake-webserver [flags]

Flags:
  -num-endpoints int
        Number of API endpoints to generate (default 50)
  -num-regions int
        Number of region labels to generate (default 5)
  -num-versions int
        Number of version labels to generate (default 3)
  -oscillation-period duration
        Duration of rate oscillation period (default 5m)
  -enable-process-metrics
        Include process_* metrics (default true)
  -enable-go-metrics
        Include go_* metrics (default true)
  -allow-metrics-compression
        Allow gzip compression of metrics (default true)

Examples

Default settings (~4,860 time series):

./fake-webserver

High cardinality test (~31,200 time series):

./fake-webserver -num-endpoints=100 -num-regions=10 -num-versions=5

Extreme load test (~156,000 time series):

./fake-webserver -num-endpoints=500 -num-regions=10 -num-versions=5

Metrics Generated

• codelab_api_request_duration_seconds - Histogram of request durations (with buckets)
• codelab_api_http_requests_in_progress - Gauge of concurrent requests
• codelab_api_requests_total - Counter of total requests
• codelab_api_request_errors_total - Counter of failed requests

Each metric includes labels: method, path, status, region, version

Docker image

coldstar/fake-webserver:v2

You can simple use kubectl apply -f deploy.yaml