Cijo Thomas
Cijo Thomas
> It looks like it is also possible to use metrics views and the feature flag spec_unstable_metrics_views to setup an SDK pipeline to transform explicit bucket histograms to exponential histograms....
https://github.com/dotnet/runtime/issues/86966 This is related I believe. This repo hasn't shipped a stable version of Logging appenders, but once it is there, I think it should allow setting LogRecord's traceid/spanid from...
https://github.com/open-telemetry/opentelemetry-dotnet/blob/main/docs/logs/extending-the-sdk/README.md#processor You can author a custom LogRecordProcessor, and override the traceid/spanids. Processors are run synchronously so they'll have access to the same context as logging statement.... If this does not...
https://github.com/open-telemetry/opentelemetry-dotnet/discussions/6047#discussioncomment-11769008 Maybe only available in main-logs branch? @CodeBlanch can confirm is nightly is published from that branch.
> avoiding the map lookup saves ~11 ns per record call on my machine OTel .NET's benchmarks show a totally diff. story. With 0 attributes, it takes 10-15ns to perform...
@yurishkuro > but I would first be interested to understand how much of an improvement they would provide https://github.com/open-telemetry/opentelemetry-specification/issues/4126#issuecomment-2207011779 Shows expected improvement based on my experiments with OTel .NET, Rust....
> Go SDK already implements so called AttributeSet `AttributeSet` is not something from the spec, but many languages have it. Yes sure, such a concept greatly reduces the overall cost,...
> Question of whether performance improvement would be moot in a world where measurements can be enriched by attribute key-values from context True! Bound ones cannot be enriched, it just...
> We would like a reference implementation to show the benefits as a prerequisite to accepting this. I'd be preparing one in OTel Rust. (mid-late Sept), but would be also...
@jack-berg https://github.com/open-telemetry/opentelemetry-specification/issues/4126#issuecomment-2207011779 has shown the potential gains with bound instruments.. Wondering if that is what TC was looking for?