[cmd/mdatagen] Ability to filter by resource attribute values

[dmitryax]

There are some scenarios when scrapers emit metrics for too many different resource attributes. We need to introduce a configuration interface generated and applied to every item in resource_attributes section that users can use to reduce the set of emitted resource attribute metrics.

We already have an option to disable resource attributes like this:

resource_attributes:
  process.command_line:
    enabled: false

We can add an option to filter resource metrics in the same configuration section.

So user can do the following:

resource_attributes:
  process.command_line:
    include:
      - "/home/dir/*"
  process.owned:
    exclude:
      - "root"

The exact filtering interface is still to be defined.

Originally suggested in #8188 (comment)

[github-actions]

This issue has been inactive for 60 days. It will be closed in 60 days if there is no activity. To ping code owners by adding a component label, see Adding Labels via Comments, or if you are unsure of which component this issue relates to, please ping @open-telemetry/collector-contrib-triagers. If this issue is still relevant, please ping the code owners or leave a comment explaining why it is still relevant. Otherwise, please close it.

Pinging code owners:

• cmd/mdatagen: @dmitryax

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

[dmitryax]

For the include/exclude config interface, we need to stick to the new standardized approach proposed in #25161. People are more in favor of option B, but we still have time to reconsider it.

[povilasv]

👋 Have a very raw draft PR for this functionality, would love to get some reviews if this is the approach we would like to take.

Basically it implements this kind of interface:

string.enum.resource.attr:
    description: Resource attribute with a known set of string values.
    type: string
    enum: [one, two]
    enabled: true
    exclude:
      - strict: two
    include:
      - strict: one
      - regex: one|two

1. I implemented small filter package for filter config and matching (https://github.com/open-telemetry/opentelemetry-collector-contrib/pull/29769/files#diff-73dd80db8a6a9e9d17674e3d3a868caadb9b1a01cb64c24427b6f8fde4725e3bR12)
2. In order to initialize the regexes we generate code in NewResourceAttributeBuilder (can be improved) . Example -> https://github.com/open-telemetry/opentelemetry-collector-contrib/pull/29769/files#diff-493333c8cd004749e934daeb3a36e2be37027bf7b4c8c2896c9cea13fcb93655R47-R53

NewResourceBuilder will need to now return an error, since it needs to compile the regexes.

3. I add the calls to includeFilter.Matches(val) in Set methods. https://github.com/open-telemetry/opentelemetry-collector-contrib/pull/29769/files#diff-493333c8cd004749e934daeb3a36e2be37027bf7b4c8c2896c9cea13fcb93655R92-R104

We still have the open question around what do we do with types such as int, double, bool, bytes, slice and map?

ATM we generate only for string setters

cc @dmitryax

[dmitryax]

We still have the open question around what do we do with types such as int, double, bool, bytes, slice and map?

I think those should only accept strict filtering