Galaxy Performance Tracking

Tracking performance over time and identifying slow queries in your Galaxy can be an extremely important thing to do, especially for larger Galaxies.

Most performance tracking requires sending metrics to a metrics collection server such as Graphite or StatsD. This document assumes you have already deployed a metrics server.

uWSGI

There is some built-in uWSGI support for performance logging. You can send uWSGI’s internal metrics to a carbon (Graphite) server by setting the carbon option in your galaxy.yml:

uwsgi:
   socket: ...
   carbon: 127.0.0.1:2003

Or a StatsD server via:

uwsgi:
   socket: ...
   stats-push: statsd:127.0.0.1:8125

The official documentation contains further information on uWSGI and stats servers. In the uWSGI Stats Server documentation, you can see an example of the sort of information that you will be able to collect. Note that you will need to make sure that the statsd pusher plugin is activated in your uWSGI servers.

Alternatively, you can use gxadmin to generate data ready to load in an InfluxDB database. In this case, you will need to add the stats option to your galaxy.yml:

uwsgi:
   socket: ...
   stats: 127.0.0.1:9191

And then run gxadmin like this:

gxadmin uwsgi stats_influx 127.0.0.1:9191

API / Route Timing Statistics

Galaxy provides middleware to automatically log the amount of time controllers take to execute and to send that data to a stats server. Using the stats server of your choice, you can calculate the relevant statistics to ensure that your Galaxy server is performing as expected.

The statsD configuration requires setting the following options in the galaxy section of config/galaxy.yml:

galaxy:
  #...
  statsd_host: 127.0.0.1
  statsd_port: 8125
  statsd_prefix: galaxy

Most people visualize the statistics using something like Grafana:

../../_images/grafana.png