feat(notify): push the operator when runtime coverage collapses (JEF-427)#246
Merged
thejefflarson merged 1 commit intoJul 18, 2026
Merged
Conversation
… collapses (JEF-427) The breach notifier (ADR-0018) is the one sanctioned outbound push, but it fires only on breach DECISIONS — and a blind engine makes none. So the moment protector's own eBPF sensors go dark (the incident shape), the notifier stayed silent: coverage collapse was pull-only (dashboard + /metrics), nothing screamed. Adds a second, narrow event on the SAME sanctioned path: an edge-triggered runtime-coverage notice consuming JEF-421's stall edge. - Fires exactly once when a was-covering fleet goes fully dark past the stall debounce (`runtime_coverage_degraded`), and once on recovery (`runtime_coverage_restored`) — edge-triggered, not per-pass, reusing JEF-421's hysteresis so a routine DaemonSet roll never strobes it. `stamp_runtime_coverage` now returns an `Option<CoverageEdge>`; the engine loop fires the push on Some. - Same posture as the breach notice: off unless PROTECTOR_ENGINE_NOTIFY_URL is set (zero outbound calls when disabled); same bounded, in-cluster-validated, fail-safe client; a failure logged once and dropped, never touching a verdict/actuation/journal. - Counts-only, redacted by construction: event tag + feed label (our own constant) + N-of-M blind counts + a last-observed age. No node names, no topology, no secrets, no CVEs — nothing untrusted to sanitize. Tests: counts-only payloads (degraded + restored), the edge→notice mapping, the disabled-notifier no-op, and a Findings-level test that the edge fires exactly once per collapse and once per recovery (never per pass). ADR-0018 extended to record the new edge on the sanctioned egress path. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VtjoJttCvBY4dzCoE4f9vP
thejefflarson
deleted the
thejefflarson/jef-427-coverage-collapse-notifier
branch
July 18, 2026 23:53
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes JEF-427.
What
The breach notifier (ADR-0018) is protector's one sanctioned outbound push — but it fires only on breach decisions, and a blind engine makes none. So the exact moment protector's own eBPF sensors go dark (the incident shape), the notifier stayed silent; coverage collapse was pull-only (dashboard +
/metrics), nothing screamed.This adds a second, narrow event on the same sanctioned path: an edge-triggered runtime-coverage notice consuming JEF-421's stall edge.
Behavior
runtime_coverage_degraded), and once on recovery (runtime_coverage_restored) — never per pass, reusing JEF-421's hysteresis so a routine DaemonSet roll never strobes it.stamp_runtime_coveragenow returnsOption<CoverageEdge>; the pass loop fires the push onSome.PROTECTOR_ENGINE_NOTIFY_URLis set → zero outbound calls, byte-identical to today. Same bounded, in-cluster-validated, fail-safe client (reusesself.url/self.client); a failure is logged once and dropped, never touching a verdict, actuation, or the journal.Runtime, our own constant) + N-of-M blind counts + a last-observed age. No node names, no topology, no secrets, no CVEs — there is no untrusted cluster string in it to sanitize.Acceptance (all covered)
Tests
Counts-only payloads (degraded + restored), the edge→notice mapping, the disabled-notifier no-op, and a
Findings-level test that the edge fires exactly once per collapse and once per recovery. Full suite green (cargo test -p protector), clippy + fmt clean. ADR-0018 extended to record the new edge on the sanctioned egress path.🤖 Generated with Claude Code