[connector/spanmetrics] Add option exclude_dimensions_values

[mar4uk]

Component(s)

connector/spanmetrics

Is your feature request related to a problem? Please describe.

Some data in metrics spanmetrics connector generates is not needed in specific cases. To reduce cardinality (and the price for storing those metrics) would be good to have an opportunity to exclude some dimension values

Describe the solution you'd like

Add new options to the spanmetrics config: exclude_dimensions_values

For example, the user is not interested in metrics that have span.kind != SPAN_KIND_SERVER. So the user can exclude those metrics by adding the option:

exclude_dimensions_values:
  - name: span.kind
     values: ['SPAN_KIND_INTERNAL','SPAN_KIND_CLIENT','SPAN_KIND_PRODUCER']

Describe alternatives you've considered

The workaround to exclude metrics with specific dimension values is to use filterprocessor:

  filter/drop_unneeded_span_metrics:
    error_mode: ignore
    metrics:
      datapoint:
          - 'IsMatch(metric.name, "traces.spanmetrics.calls") and IsMatch(attributes["span.kind"], "SPAN_KIND_INTERNAL|SPAN_KIND_CLIENT|SPAN_KIND_PRODUCER")'

Additional context

No response

[github-actions]

Pinging code owners:

• connector/spanmetrics: @albertteoh

See Adding Labels via Comments if you do not have permissions to add labels yourself.

[mar4uk]

I'm also interested in implementing it

[albertteoh]

I'm also interested in implementing it

Thanks for offering to implement it, @mar4uk.

Unless if there's a very good reason, I would prefer the alternative you've provided in this ticket (thanks for putting the thought into considering this alternative!) and the reason is to keep the spanmetrics connector as simple and clean as possible, following the unix philosophy of doing one thing well (or at least, striving for that 😄).

[crobert-1]

I agree with @albertteoh here. I don't see any notable benefit of another filtering implementation being introduced in the spanmetrics connector over adding the filter processor to the trace or metrics pipeline.

I'm going to close for now, but feel free to share more information in case we're missing something here!