What is a idiomatic way of tracking degradation between test runs?

We would like to run our load tests regularly, e.g. during weekends. And we want to track the results and alert on their degradation. How would you approach that?

Hi @notxcain.

We can all agree that some “golden” performance metrics to look after for creating benchmarks are well documented here.

But as you can read on that same page, in the end it might come down to your goal. I have tested different system with different goals, for i.e.:

  • 1st case: the number of transactions per second is the main metric/goal
  • 2nd case: the p95 is the main metric/goal (ignoring errors)
  • 3rd case: the p95 and the number of errors is the main metric/goal, etc…

Try to get the team together and define the goals you want to achieve.

Hey @rdt.one , thanks for the reply.

But the question is not about what to track but rather how? How do I collect p95 response times of multiple runs and track its degradation.

You can instruct k6 to pass test metrics to Influxdb and then you can create Grafana dashboards and ideally do customs configurations/alerts on the dashboard per desired metrics. Read here how to output test metrics to influxdb.

How to setup Grafana dashboards resources

1 Like