Generate process metrics with semconv yaml

[braydonk]

No description provided.

[braydonk]

Hi @joaopgrassi, I decided to open my changes as a draft PR so you could check my work. This is converting the metrics from process-metrics.md into yaml that semconvgen can use. I also included the example of one generation in the markdown file. I would appreciate if you could take a look and make sure I'm on the right track before I finish off the work and open a properly-formed PR.

[joaopgrassi]

@braydonk looks pretty good to me! I think all is in the right track :)

[joaopgrassi]

Suggested change

	- id: process.cpu
	- id: attributes.process.cpu

I started adding attributes. to attribute_group to easily spot that this is not a span/metric/log convention but just rather a "bag" of attributes to be shared.

[joaopgrassi]

Suggested change

	instrument: updowncounter
	instrument: counter

It's a counter today, do we want to change?

[braydonk]

Nope we want to keep a counter, that was a copy paste mishap 😄 thanks

[jsuereth]

Looks like you have a minor YAMLlint error

[mx-psi]

I have a similar comment to the one I left in #89 (review); this PR moves the metrics to YAML and also renames the attributes. If doing each of those changes separately is hard/undesirable, I think we should at least describe this clearly in the PR title, PR description and a changelog entry indicating what changes #51 entails for this set of metrics

[braydonk]

@mx-psi I'm kind of confused, which part of this PR needs to be separated? I don't think I intentionally renamed any attributes; I tried to copy them directly from the existing document. If I missed any I will change them back.

[joaopgrassi]

@braydonk the issue is when changing to YAML, the attributes such as state are not only state but they get the metric "namespace". Ex. process.cpu.state, so it becomes a "breaking change".

There's no easy way to transform the existing metrics/attributes to YAML without this attribute name change, because in YAML the same attribute key cannot be repeated. The only way would be to define them under each metric which is very repetitive/tedious.

I suggest you take the same approach as I did in #89 (comment) and document all the existing attributes vs how they are now with these changes. Plus, the changelog entry + schema file changes (You can see how I did in my PR).

I think that should address @mx-psi concerns.

[braydonk]

Thanks @joaopgrassi I'll get that fixed soon.

[braydonk]

I've opened #330 instead. The commit history on this one was really messy with attempts to keep it up to date with main branch, so I started a new clean PR.