ajwerner
ajwerner
>I'm not sure it's all that rare, TBH. Say we're dealing with response times, rounded to milliseconds, and that most requests are handled within 1 second. Chances are, most requests...
I’m okay with the idea of using 0 though it introduces special cases to the code. If you know all of your counts are integer values then using 1 seems...
I haven't poked at this code really since I wrote it. https://github.com/tvondra/tdigest is almost certainly better if you're looking for a C implementation. I think that this particular issue isn't...
I guess maybe his repo won't be easy to use in another project because of its postgres ties. Should have looked first.
PRs welcome if you want to get your hands dirty.
This seems fixed in https://github.com/tokio-rs/tracing-opentelemetry by https://github.com/tokio-rs/tracing-opentelemetry/pull/43.
Exponential backoff is not what is always desirable. In cockroachdb, for example, locks are held across retries in order to avoid starvation. Backing off would induce unnecessary latency and prevent...
It seems like the real ask here is for us to make the `cgroups` library go-gettable as its own module, yeah?
Makes sense. One day if I find time I'll finish up some of the ideas here.
I actually am starting to think that all this code is not the way it should be done. Right now this is effectively duplicating code in rust that duckdb already...