/ok-to-test

Found and split out an unrelated regression while testing this PR end-to-end on CRC: #1555. Without it, Snapshot -> Release stalls on Application-model tenants. Not a blocker for this PR; merging order doesn't matter.

@ci-operator ,
looks good at all , this is a massive effort thanks you ,

integration-service under implementation of new component model which is one the core components in service
the implementations are made on phases this why we are now maintaining both old and new componentGroup and application, once the whole implementation we will be with only componentGroup only
I bet that would much easier.
That said if you want to continue with implementing this distributed tracing feature you should take care of both flows , for me I just saw two places it's not emitted.
I'll ask from our team colleagues to review

Codecov Report

❌ Patch coverage is 90.03559% with 28 lines in your changes missing coverage. Please review.
✅ Project coverage is 73.58%. Comparing base (5b4f7e5) to head (659b591).
⚠️ Report is 1 commits behind head on main.

@@            Coverage Diff             @@
##             main    #1543      +/-   ##
==========================================
+ Coverage   65.67%   73.58%   +7.91%     
==========================================
  Files          65       68       +3     
  Lines        8923     9200     +277     
==========================================
+ Hits         5860     6770     +910     
+ Misses       2366     1750     -616     
+ Partials      697      680      -17     

Where was this requested? Is there story related to those changes? Why and what for do we need this change?

@jsztuka here's the ADR that's being implemented, part of which is this PR: https://github.com/konflux-ci/architecture/blob/main/ADR/0062-distributed-tracing.md

Thanks for the review and for flagging the gaps. Both flows are now covered - emitBuildTimingSpans() is called on every terminal exit path in both EnsureSnapshotExists (ComponentGroup model) and EnsureSnapshotExistsApplication (Application model): success, failure, and the already-annotated early return. The dedup annotation prevents double emission on requeue.

Also folded in the spec.application ReleasePlan cache index fix (#1555) since without it the Snapshot controller crashes on Application-model tenants and the trace-context propagation chain breaks at that stage.

Update: folded the #1555 fix into this PR since Application-model trace propagation depends on it. #1555 can be closed.

/ok-to-test

🤖 Review · Started 1:40 PM UTC
Commit: ffde3b2 · View workflow run →

Resolved a rebase conflict in snapshot_adapter_test.go where main's
new ComponentGroup test cases landed in the same Describe block as the
supersede-trace test. Kept main's additions and adapted the
supersede-trace test to use the sibling-snapshot mock the new supersede
path expects.

Also added a few tests around samplerFromEnv, EmitAndMarkTimingSpans,
and the build/test emit paths to cover patch and refetch failures.

🤖 Finished Review · ✅ Success · Started 1:40 PM UTC · Completed 1:56 PM UTC
Commit: ffde3b2 · View workflow run →

🤖 Finished Review · ✅ Success · Started 7:49 PM UTC · Completed 8:03 PM UTC
Commit: 218f229 · View workflow run →

@dirgim - rebased on main; conflicts were additive only (loader/tracing imports, condition consts) so no behavior change since your last /ok-to-test. Mind re-running it on the new HEAD so CI picks up?

🤖 Finished Review · ✅ Success · Started 6:39 PM UTC · Completed 6:55 PM UTC
Commit: ec21706 · View workflow run →

/ok-to-test

🤖 Finished Retro · ✅ Success · Started 9:00 PM UTC · Completed 9:09 PM UTC
Commit: ec21706 · View workflow run →

Retro Analysis: PR #1543 — Distributed Tracing

This XXL PR (2100+ lines, 23 files) was open for 2 months with 10 review agent iterations, 5 gemini-code-assist suggestions, and human review from 4 maintainers.

What went well

• fullsend-ai-review caught a genuine HIGH security issue — the go-billy v5.8.0 downgrade that silently swallowed path-traversal errors. Author fixed immediately.
• fullsend-ai-review correctly identified a race condition in finalizer removal, which the author addressed with RetryOnConflict.
• gemini-code-assist's ParseFloat error handling suggestion was accepted and improved the code.
• CI pipeline caught a useless ctx assignment via CodeQL.
• Test coverage reached 90.04% patch coverage.

What could go better

• The most critical bugs were caught only by a human reviewer (kasemAlem): two missing emission paths in the Application model flow of buildpipeline_adapter.go. All automated reviewers missed these across 10+ iterations because they lacked context about the repo's dual-model architecture (ComponentGroup vs Application flows).
• 10 review iterations with repeated findings: The "missing-authorization" / Jira ticket finding appeared in 6 consecutive runs; the annotation-namespace finding persisted across runs after the author explained the ADR rationale. These are already tracked by existing upstream issues (#1013, #1500, #1672).
• Large PR size concerns are already tracked upstream (#2118, #2034).

Proposal filed

1. Document dual-model completeness requirement in AGENTS.md (konflux-ci/integration-service) — Adding explicit guidance that feature implementations must cover both Application and ComponentGroup flows would close the gap between human and automated review quality for this class of bug.

Proposals filed

• Document dual-model (Application + ComponentGroup) completeness requirement in AGENTS.md (in konflux-ci/integration-service)