[FEATURE] Record flag set ID for OTEL metrics #2020

Activity

1. juanparadox

   added

   enhancementNew feature or requestNew feature or request

   Needs TriageThis issue needs to be investigated by a maintainerThis issue needs to be investigated by a maintainer

   on Aug 13, 2026
2. gargsajal9 commented on Sep 13, 2026

   

   gargsajal9

   on Sep 13, 2026

   Last edited by gargsajal9

   Contributor

   More actions

   Hi @toddbaert and @bacherfl — could you take a quick look at this proposed contract before I start implementing it?

   I traced the current evaluation telemetry paths. OpenTelemetry already defines the release-candidate feature_flag.set.id attribute, so I think we should use that instead of adding a flagd-specific name.

   My proposed behavior is:

   • add it to impression and reason/error metrics when a non-empty string ID is available;
   • use each evaluation result’s existing metadata for single and bulk evaluations;
   • for a failed lookup with an explicit flagSetId selector, use the selected ID;
   • omit the attribute when the ID is unknown; and
   • for unfiltered bulk requests spanning multiple sets, attribute each result separately instead of using a synthetic value.

   This wouldn’t export arbitrary metadata or change responses, selectors, traces, storage, metric names, or the existing cardinality limit.

   Does that behavior sound right, especially using the selector ID for failed lookups and including it on reason/error metrics?
3. juanparadox commented on Sep 30, 2026

   

   juanparadox

   on Sep 30, 2026

   MemberAuthor

   More actions

   @gargsajal9 Did you by chance start on this? Didn't want to start work on this if you were already addressing it.
4. gargsajal9 commented on Sep 30, 2026

   

   gargsajal9

   on Sep 30, 2026

   Contributor

   More actions

   Thanks for checking, @juanparadox. I haven’t started implementation — I’ve only done the setup, source tracing, focused baseline, and posted the contract proposal above, which is still waiting for maintainer confirmation. Please feel free to take it forward; I’m happy to coordinate or collaborate so we don’t duplicate work.
5. beeme1mr commented on Oct 2, 2026

   

   beeme1mr

   on Oct 2, 2026

   Member

   More actions

   Hey @juanparadox and @gargsajal9! I think `feature_flag.set.id, as @gargsajal9 recommend, would be best.

   The only real challenge is deciding when to omit the attribute due to cardinality reasons. @gargsajal9 proposal seems to address that pretty thoroughly.
6. gargsajal9 commented on Oct 2, 2026

   

   gargsajal9

   on Oct 2, 2026

   Last edited by gargsajal9

   Contributor

   More actions

   If we are onboard on the approach, I am happy to start the implementation and share the PR for review.
   @beeme1mr @juanparadox if you are ok should I start working on PR then?

   👍React with 👍2Reacted by Juan Bernal and Michael Beemer
7. erka commented on Oct 2, 2026

   

   erka

   on Oct 2, 2026

   Member

   More actions

   Just do it @gargsajal9. It’ll take less time to implement than to discuss it.

   🚀React with 🚀2Reacted by Juan Bernal and Sajal Garg
8. erka

   assigned

   gargsajal9

   on Oct 2, 2026
9. gargsajal9 commented on Oct 2, 2026

   

   gargsajal9

   on Oct 2, 2026

   Last edited by gargsajal9

   Contributor

   More actions

   Implementation is ready in #2078. It follows the feature_flag.set.id behavior discussed above
10. toddbaert

    closed this as completedin #2078on Oct 19, 2026
11. pbacterio

    added a commit that references this issue on Oct 20, 2026

    feat: add flag set ID to evaluation metrics (open-feature#2078)

    ...

    This issue will close once commit 10f2ca0 is merged into the 'main' branch.

    10f2ca0