Description
The accepted OTLP pod-log capture needs live comparison, while stale producer allowances conceal zero-headroom consequences. Measure one attributed local receiver capture before conditional promotion and reduce the allowance only at the final corpus tree. No collector model expansion or cloud operation is authorized.
Acceptance Criteria
- #1 The OTLP pod-log declaration and sole exemption expected_matches=2 land together with other exemption fields unchanged
- #2 One local receiver capture reports actual OTLP metric receipts, attribution, proposal hashes and teardown evidence; promotion follows measured constant-bound capacity
- #3 Final observed count equals expected_count and claim length with no stale or untriaged pairs, with zero-headroom consequence recorded
- #4 Integrated gates and exact-SHA CI pass and the narrower pod-log envelope finding is recorded with its evidence boundary
Definition of Done
- #1 just check (fmt-check, lint, gen-check, env-check, docs-check, test, race, hygiene, ui-check, compose-check, helm-test, lab-check, signal-fidelity)
- #2 just gen (only if a blueprint field, construct/workload config struct, or a skill under plugins/synthkit/skills/ changed)
- #3 just dump — inventory diffed against signals/
Implementation Plan
Run independent L3/L4; root R1 modelling pair, R2 gate, R3 scratch measurement and conditional promotion by equal stale swaps, R4 final bound reduction, R5 reconciliation and terminal report. Never right-size before capture disposition.
Implementation Notes
Preflight fidelity is exit 0 at observed157 expected203 claims203 stale46 untriaged0. An auxiliary export using all blueprints was the wrong recipe scope and yielded 187/203 with 30 untriaged apiserver pairs; retained diagnostically, not used for policy. Correct export follows justfile signal_fidelity_blueprints and raw comparator enumeration agrees with 157/203. L3 historical eight-key envelope origin is undetermined: old raw artifacts are absent locally and current AddLog folds by source/transport. Same-pod key changes alone do not prove an aggregation artifact; pre-fold raw resource evidence is required.
R1 paired declaration and expected_matches2 validate with exactly two underlying manifest contradictions and no untriaged gaps. R2 ordered gen-check, spdx-check and full just check exit0; ratchet observed157 expected203, claims203 stale46. Dump retains all metric, OTLP metric, log and profile inventories unchanged; one acme-datagen-analysis sampled classify_request span appears. No emission code changed. R1 CodeRabbit skipped as explicitly allowed for nonbranching one-line wiring plus JSON; documents and policy data validated instead of unit-tested. Exemption hash bf4fb9e4859f0c7be2e26d88644b3f61d4c789d39ae4c65b340891d4a2f96be9; only expected_matches changed.
R1 pair committed at 27e0288c33a3a3936fd76ccd907676251f827c28; R2 exact-SHA CI34624586338 is still running. Root read L4 final result: captured, full300s, otlp_metrics8005, otlp_logs166, metrics74, attributed32/unattributed42. Proposal55 k8s families matches the existing area membership and captured candidate names/types/transports/producers/label keys, with privacy elision only. Source receiver SHA09b71d8fb290ae36409e443fbb51cb80dc58f7fb7925e18f3a272551171018a9 matches the integrated tree. No promotion yet.
A1 root bounded-prerequisite decision: additive step7 scratch load fails duplicate area/source/substrate/permutation k8s|k3d_lab|k3s|otel-receivers before comparison. Existing README documents lab-matrix promote -merge, whose CanonicalMerge preserves established observations. Goal6 step7 clarified to use a cumulative proposal seeded from the existing document, then measure and promote its exact bytes at the same path. No retirement, raw-capture rewrite, identity change or comparator change. L4 re-dispatched only to generate the merged scratch proposal from the same capture. Confidence0.9; materiality medium because publication shape needed an explicit correction, despite using the existing authorized route. R2 CI34624586338 succeeded at27e0288c33a3a3936fd76ccd907676251f827c28, all10 jobs.
R3 scratch cumulative proposal d9a42e55da2c51053a102c845e0d7758c89bfc000e3a2c0fce639f48686885eb measured BEFORE promotion: observed189 expected203 claims203 stale46 untriaged32. All32 exact new pairs have producer otlp-native, not the promrw producer hypothesized in goal section2.5. Swap32 stale claims at unchanged203 gives observed189 claims203 stale14 untriaged0. All55 existing metric non-producer entries independently unchanged. Promotion accepted under A1; host document untouched. Right-size14 remaining stale LAST.
R3 full just check exit0. Producer coverage ratchet: observed=189 expected=203 (report-only within bound; growth fails). Cumulative corpus hash d9a42e55da2c51053a102c845e0d7758c89bfc000e3a2c0fce639f48686885eb exactly matches L4 proposal. Corpus and claims are declarative data: validated, no unit-test or CodeRabbit rerun required.
R3 CI34626745006 success at89da21b9a85ae87f8f7584b41409bc916d905592, all10 jobs. BEFORE R4 live remeasurement189/203/203 stale14 untriaged0. R4 removed exactly14 remaining stale claims; observed=expected_count=len(claims)=189, stale0 untriaged0. ZERO HEADROOM: the next new no_comparable_producer finding anywhere fails the gate; unmodelled permutation disposition does not shield counting. Future evidence work must settle existing gaps or obtain separately authorized policy decisions, never silently increase this bound.
R4 ordered just gen-check, just spdx-check, full just check all exit0. Producer coverage ratchet: observed=189 expected=189 (report-only within bound; growth fails). No corpus or emission changes after right-sizing.
R5 CI34628037101 success atef9862af2f4d02549f2b5e87f0ecac05a29a4c69, all10 jobs. All three required gate SHAs green. Conditional gen not applicable; gen-check passed. L3 historical eight-key envelope remains undetermined, not claimed solved; evidence boundary recorded above.
Final Summary
Done: paired modelling/exemption commit27e0288c33a3a3936fd76ccd907676251f827c28; one local300s OTLP receiver capture live-proved8005 metric items and166 log items; cumulative55-family k8s promotion89da21b9a85ae87f8f7584b41409bc916d905592 preserved all prior observations. Measured32 new gaps fit46 stale slots at203; final right-sizeef9862af2f4d02549f2b5e87f0ecac05a29a4c69 leaves observed=expected_count=claims=189, stale0 untriaged0, ZERO HEADROOM. otel-receivers remains unmodelled. Historical narrow envelope origin remains unknown; no cloud operations or emitter work.