Add node.js runtime metrics semantic conventions

[maryliag]

Part Of #990

Changes

Adding semantic conventions for Node.js runtime metrics

Merge requirement checklist

• CONTRIBUTING.md guidelines followed.
• Change log entry added, according to the guidelines in When to add a changelog entry.
  • If your PR does not need a change log, start the PR title with [chore]
• schema-next.yaml updated with changes to existing conventions.

[maryliag]

I wasn't sure if I need to update something on the next, since there was no changes to existing semantic. This is a new one.

[trask]

cc @open-telemetry/javascript-approvers

[Netail]

I am wondering whether or not it should be called NodeJS. What if someone uses Deno or Bun?

[maryliag]

I am wondering whether or not it should be called NodeJS. What if someone uses Deno or Bun?

what about just js ?

[Netail]

I am wondering whether or not it should be called NodeJS. What if someone uses Deno or Bun?

what about just js ?

I mean I guess, but also feel like it can be confused with the language instead of the runtime environment

[maryliag]

I mean I guess, but also feel like it can be confused with the language instead of the runtime environment

if we want to go be definition, it would be runtimejs

[Netail]

I mean I guess, but also feel like it can be confused with the language instead of the runtime environment

if we want to go be definition, it would be runtimejs

jsruntime sounds maybe better? idk, might need some extra opinions

[pikalovArtemN]

For my opinion different runtimes need to have diferent preffix.
For example:
node - nodejs
bun - bun
deno - denojs
etc...
Or some attribute that specify what runtime is currently run.
Or metric idk.

Also i think for version we need to set metric like it does in the prom-client library. Attribute in all metrics with version, for my opinion, it's not a good solution.

[david-luna]

One last question from my side. I've checked that other metrics of time make use of histogram as instrument. Some examples:

• semantic-conventions/model/metrics/jvm-metrics.yaml

  Line 49 in cde003c

  - id: metric.jvm.gc.duration

• semantic-conventions/model/metrics/go-metrics.yaml

  Line 77 in cde003c

  - id: metric.go.schedule.duration

Here all the metrics related to delay and using s as unit use the gauge instrument. Probably a naive question but why is that?

[maryliag]

Here all the metrics related to delay and using s as unit use the gauge instrument. Probably a naive question but why is that?

That is a limitation of this type of metric for node, meaning, I'm using this more because is from the event loop and not because is a duration (I even use histogram on duration on other metrics).
The gist is that the values for min, max, p50, p90, p99 for these metrics come from the builtin histogram from the node runtime and is not possible to convert it to otel histogram, because we can only query the distribution not get the entire histogram object. This is why you will notice I have a separate metric for each value (min, max, etc), instead of just one called delay that would have everything.

Hope that helps clarify!

[david-luna]

Here all the metrics related to delay and using s as unit use the gauge instrument. Probably a naive question but why is that?

That is a limitation of this type of metric for node, meaning, I'm using this more because is from the event loop and not because is a duration (I even use histogram on duration on other metrics). The gist is that the values for min, max, p50, p90, p99 for these metrics come from the builtin histogram from the node runtime and is not possible to convert it to otel histogram, because we can only query the distribution not get the entire histogram object. This is why you will notice I have a separate metric for each value (min, max, etc), instead of just one called delay that would have everything.

Hope that helps clarify!

It helps a lot. Thank you! :)

[lmolkova]

That is a limitation of this type of metric for node, meaning, I'm using this more because is from the event loop and not because is a duration (I even use histogram on duration on other metrics). The gist is that the values for min, max, p50, p90, p99 for these metrics come from the builtin histogram from the node runtime and is not possible to convert it to otel histogram, because we can only query the distribution not get the entire histogram object. This is why you will notice I have a separate metric for each value (min, max, etc), instead of just one called delay that would have everything.

Could you please add a similar note to the md file itself?

@open-telemetry/javascript-approvers any last minute feedback?

[maryliag]

Could you please add a similar note to the md file itself?

Done