locust: Rename "request" to "sample", and refactor request_success/failure into one event
The word request makes no sense for a lot of non-http protocols. Users could be reporting things like the reception of async WebSocket messages, or the time to run a script. The word sample is more appropriate.
Having separate events for request_success and request_failure make no sense either, so I suggest we deprecate those (but support them for a long time) and just use a single one, sample(with a boolean parameter success).
Perhaps this should be done at the same time https://github.com/locustio/locust/issues/1413
About this issue
- Original URL
- State: closed
- Created 3 years ago
- Comments: 15 (5 by maintainers)
I don’t have a strong opinion about it. But if I were to choose today, I think I’d prefer a single event. So as long as we kept the old ones for backwards compatibility for a long time like you said, I think it’s a good idea to change it.
@DennisKrone is working on a PR to add optional metadata to the event (basically anything you like, things like customer id is a good example).
We’ve also talked about adding the User object, and I think request/response make a lot of sense as well. We just need a sensible behaviour for non-http / custom clients as well.
In the case of join both request_success/failure, I would also think is a good idea to pass the entire request/response object, there are so many cases that it has meaningful info inside it, that would be great to have them instead of the 4 attributes (request_type, name, response_time, response_length)
A metric could represent anything <- Exactly. This is why we should NOT call it a metric.
What you are suggesting is a major rewrite. Which might make sense, but what I’m talking about here is just renaming a couple of things.
As for Grafana, I too love it for visualiization - so much that I have abandoned the Locust GUI, and made this instead https://grafana.com/grafana/dashboards/10878 😃 (on a side note, I use Timescale and log every request, because I’ve always felt that you need every request for analysis later on - particularly error messages etc)