README.md

monitoring uds-proxy

Exposure of performance metrics can be enabled by using uds-proxy's -prometheus-port command line argument. For example, using -prometheus-port :12345 will make metrics available at http://localhost:12345/metrics.

This directory contains Grafana provisioning configuration and prometheus configuration used by docker-compose.yml to spin up an example uds-proxy dashboard, fed with live data.

From uds-proxy source root directory, run

make monitoring_test

to bring up uds-proxy, Prometheus and Grafana for experiments. This will...

• on Mac: build and run uds-proxy locally on your host (requires Go)
• on Linux: build and run uds-proxy using Docker
• spin up Prometheus, configured to collect metrics from uds-proxy
• spin up Grafana, configured with prometheus as datasource and uds-proxy dashboard
• timeit: run 50 test requests directly and 50 via uds-proxy

Then...

• visit http://localhost:3000 and select one of the three dashboards available:
  • uds-proxy stats
  • go process stats
  • prometheus stats
• use uds-proxy-test.socket for further tests, eg. use make run_some_requests

Shut the test environment down and lose all collected data by hitting ctrl-c.

Why does make monitoring_test spin up two uds-proxy instances on Mac -- a local one and a dockerized one? Because containers on Docker for Mac will never be able to share a socket with the host. So, for testing purposes, a 2nd instance is started locally to enable test runs, while the dockerized one is sitting idle.