Gather Prometheus HTTP metrics for Flask Gunicorn web server.
Docker compose file has two service
- flask_gunicorn_service: Python flask web server with Gunicorn
- prometheus: Prometheus for gathering metrics
After the services are brought up, access the Prometheus dashboard at 127.0.0.1:9090
As Prometheus needs to gather the metrics from all Gunicorn worker processes, it needs special configuration like a writable directory for storing db metrics from all processes, and new http server for exposing the metrics. Pass both these variables as environmental variables in .env file.
PROMETHEUS_METRICS_PORT=9200
PROMETHEUS_MULTIPROC_DIR=/opt/prometheus_metrics
Prometheus can gather the metrics for this component through the port 9200 and this needs to be configured in Prometheus configuration file as
job_name: flask_gunicorn_service
scrape_interval: 5s
static_configs:
- targets: ['flask_gunicorn_service:9200']
In the gunicorn config file gunicorn.conf.py config the Prometheus for multiprocess by starting the http server for Prometheus when server ready.
In /app/metrics/flask_prometheus_metrics.py file, add the custom metrics in the FlaskPrometheusMetrics class and function initialize_custom_metrics().
In the /app/app.py, the FlaskPrometheusMetrics is initilized at the end of the all endpoints definition, if not then the custom metrics don't track these endpoints.