[statsd] Relay's statsd take up 20% CPU

[jernejstrasner]

Something in the way that we emit metrics is inefficient and fishy. I suspect we don't aggregate any metrics in-process, and as a result spend way too much CPU on emitting metrics.

Anton Ovchinnikov claimed 20% CPU increase by turning on metrics collection. More CPU usage than normal in both Relay and the statsd collector.

We also likely drop metrics as a result of that.

[iker-barriocanal]

What do we want to do with this? I think the best course of action at this point in time is to not emit many metrics, see investigation doc.

[jernejstrasner]

@jan-auer what's your take since this is metrics related?

[Dav1dde]

Not a problem at the moment, with the addition of statsdproxy this already got a lot better.