There are various options though, we've used flog in the past to generate load but really you need to match both data rates and data structure/size (e.g. lots of small messages is different to big chunky ones, etc.) https://github.com/mingrammer/flog

I did attempt to do this a while back for OSS via some pre-built images we could then trigger but that's all pretty dead now: https://github.com/fluent/fluent-bit/blob/master/.github/workflows/pr-perf-test.yaml The idea was we could get an idea of performance impact for changes in a PR

I'd be happy to help integrate or support any contribution though

It's all really bespoke though, it depends on your data, your infra and your config (e.g. lots of filters impacts things)