Caleb Schoepp

> Concretely, what about something like . e.g. spin-llm.infer? I like this convention, but don't know how to square it with the OTEL conventions. For example `spin-redis-outbound.get` according to [semantic...

Also @lann I'm curious what your thoughts are on `#[instrument(err)]`. More specifically it seems there is a pattern in host components of returning `Result`. `err` only is able to report...

> I would suggest preferring OTEL conventions where they are clear and come up with a nice convention that we can be internally consistent about where there is no OTEL...

Okay so the new strawman argument is: - Where there are existing OTEL semantic conventions we follow those. - Where there are not existing OTEL semantic conventions we will follow...

@itowlson @lann I've modified the code to align with the big discussion here. One last sanity check might be nice. I'll give you both about a day and a half...

> @calebschoepp Is this superseded by the OpenTelemetry work? Good question. I would say it's tangentially related but not superseded by the OpenTelemetry work.

I'm re-reading through this and I feel pretty weird about passing any `RUST_LOG` value through as a `SPIN_LOG`. From my perspective setting `RUST_LOG` is purely about introspecting on the Spin...

Here is an example of what a trace might look when levels 2 through 4 are combined.

Trigger observability seems like the most tractable and immediately problem so I'm going to get started on a SIP for how we could implement it.

> Question for my own understanding: is CPU / memory utilization considered a runtime or guest metric? For example, Azure Application Insights emits a metric called [Process CPU](https://learn.microsoft.com/en-us/azure/azure-monitor/app/standard-metrics#process-cpu-performancecountersprocesscpupercentage), which shows...