metrics.md

Metrics Collection (Prometheus)

Dracan includes an integrated metrics endpoint that collects and exposes performance and request metrics, which can be monitored using Prometheus. This feature enables teams to monitor HTTP request data such as request counts, response statuses, and latencies, helping track application usage patterns and performance.

What Metrics Are Gathered?

The Dracan metrics system gathers and exposes the following data:

• Request Count (http_requests_total): A counter that tracks the total number of HTTP requests made to the Dracan app, grouped by method (e.g., GET, POST) and response status.
• Request Latency (flask_http_request_duration_seconds): A histogram that captures the latency of HTTP requests, categorized by method and endpoint, to help monitor and optimize response times.
• Request and Response Sizes: Histograms tracking the size of incoming requests and outgoing responses, which can provide insights into typical payload sizes and bandwidth usage.
• System Resource Metrics: Basic runtime metrics like memory usage and garbage collection, using standard Prometheus metrics for Python applications.

Enabling Metrics Collection

Metrics collection can be controlled via environment variables:

1. ALLOW_METRICS_ENDPOINT: Enables or disables the metrics endpoint.

   • Set to "true" to enable metrics collection.
   • Set to "false" (default) to disable metrics collection. Or do not set this env at all.

2. METRICS_PORT: Specifies the port on which the metrics server should run.

   • Defaults to 9100 if not set.
   • If a different port is desired, set METRICS_PORT to the desired port number.

   Ommited when ALLOW_METRICS_ENDPOINT set to false or not set at all.

Usage Instructions

To enable and use the metrics system in Dracan:

1. Set Environment Variables:

   • To enable the metrics server on the default port: export ALLOW_METRICS_ENDPOINT=true
   • To specify a custom port (e.g., 2000):

     export ALLOW_METRICS_ENDPOINT=true
     export METRICS_PORT=2000

2. Access Metrics:

   • When metrics are enabled, access the metrics data at <dracan_ip>:METRICS_PORT/metrics.
   • For example:
     • Default Port: http://127.0.0.1:9100/metrics
     • Custom Port: http://127.0.0.1:2000/metrics (if METRICS_PORT=2000)

3. Monitoring with Prometheus:

   • Once the metrics server is running, configure your Prometheus server to scrape metrics from Dracan’s metrics endpoint.
   • Example Prometheus configuration snippet:

     scrape_configs:
       - job_name: 'dracan'
         static_configs:
           - targets: ['<dracan_ip>:9100'] # or use custom port if specified

Disabling Metrics Collection

If metrics collection is not required, it is disabled by default, or it can be explicitly disabled by setting ALLOW_METRICS_ENDPOINT=false.