You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Fix Prometheus DI layering, PR-filter wording, k3d framing, F-M6-02-01 trigger
Four review findings addressed:
- Observability port registration now respects the existing
ApplicationServiceRegistration (NoOp) vs.
TelemetryRegistration.AddBessTelemetry() (Prometheus) split, like
the three existing metrics ports. Registering the Prometheus
implementation in the Application composition root would break
the layer separation and prevent the telemetry host from
swapping in the real adapter.
- Cluster-smoke PR-filter wording now distinguishes the two
mechanisms: GitHub paths-filter in v1.1.0 stage 1, job-internal
skip sentinel from stage 2 onward (since required checks can't
use paths). Logical path list stays identical.
- "k3d-basiert" claim in the image-strategy block softened to
"Cluster-Tool als Slice-Plan-Entscheidung, Vorschlag k3d", since
the open-decisions list still lists k3d/kind/minikube as an open
choice.
- F-M6-02-01 split out of the "Isolation-Trigger" group — its
actual trigger per the source spec is tick-budget/performance
(sequential per-asset execution exceeding CycleInterval), not
isolation. F-M6-02-02 (worker-per-asset) and F-M6-02-03 (per-asset
sidecar) now listed separately with their actual triggers.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
0 commit comments