Recording of rps over time

[hrosenhorn]

It would be helpful to understand the rps over time when running a test for 24 hours or more. Either raw data for later graph visualization or built in to the UI.

[heyman]

I agree. It's important when we implement this that we think about the amount of data that will be stored (we can't store every request that is made, but rather some aggregate like the average RPS and response time every 10 seconds or so). Also the code that does this should be independent from the web UI (should run no matter if a web browser has the web UI open).

[quosa]

Have you considered using ganglia or graphite for the graphing? With ganglia we could get a lot of useful stuff from the load generator nodes (cpu/mem/net+disk io etc.) by default. Also, if the target system uses ganglia for monitoring, the load generators could be easily added as a separate cluster.

If you also bundle graphite with ganglia, you can do nifty custom reports like node a cpu compared to incoming network in node b over a small time window; aggregate network throughput over all front-end vs. load-generator nodes etc.

[heyman]

Ganglia+graphite sounds like it could be powerful. Ideally I'd like locust to have a fully fledged plugin system that would allow for such integrations to be provided in plugins. What's needed is a plugin convention (where to put them, how to structure them etc.), more internal events (and some times more data for the existing events), as well as a good way to extend the UI from the plugin.

[cgbystrom]

Consider tracking #34 for updates.