Improve k6 usage #36

New issue

Open

opened 2025-12-28 18:13:40 +00:00 by sami · 0 comments

sami commented

2025-12-28 18:13:40 +00:00

Owner

Originally created by @carpawell on GitHub (May 29, 2024).

Some improvements that can be applied:

Use checks, they are added in the summary and this is how basically we should use k6: https://k6.io/docs/using-k6/checks/
Use thresholds, they can fail a test earlier and in an automated manner (at least we may start to have some goals for the tested product): https://k6.io/docs/using-k6/thresholds/
Use non-constant executors, they allow simulating more controllable load (constant-arrival-rate is really nice IMO): https://k6.io/docs/using-k6/scenarios/#scenario-executors
Visualise summaries: https://github.com/benc-uk/k6-reporter
Visualise metrics, IMO that is the key reason we are using a loader, we should see if a service can handle the load flat or it almost dies at the end of a test: https://k6.io/docs/integrations/results-visualization/

Originally created by @carpawell on GitHub (May 29, 2024). Some improvements that can be applied: 1. Use checks, they are added in the summary and this is how basically we should use k6: https://k6.io/docs/using-k6/checks/ 2. Use thresholds, they can fail a test earlier and in an automated manner (at least we may start to have some goals for the tested product): https://k6.io/docs/using-k6/thresholds/ 3. Use non-constant executors, they allow simulating more controllable load (`constant-arrival-rate` is really nice IMO): https://k6.io/docs/using-k6/scenarios/#scenario-executors 4. Visualise summaries: https://github.com/benc-uk/k6-reporter 5. Visualise metrics, IMO that is the key reason we are using a loader, we _should_ see if a service can handle the load flat or it almost dies [at the end of a test](https://grafana.com/media/blog/ways-to-visualize-k6-results/screenshot-annotated-k6-results-analysis.png): https://k6.io/docs/integrations/results-visualization/