Skip to content

Phase 10: prove guarded live k3s resilience - #117

Merged
CodeBuildder merged 1 commit into
agent/cluster-free-platform-demofrom
agent/live-platform-proof
Jul 19, 2026
Merged

Phase 10: prove guarded live k3s resilience#117
CodeBuildder merged 1 commit into
agent/cluster-free-platform-demofrom
agent/live-platform-proof

Conversation

@CodeBuildder

Copy link
Copy Markdown
Owner

Closes #116

What changed

  • adds make demo-platform-live-dry-run as a read-only preflight for the complete k3s security, chaos, agent, and SOG stack
  • turns make demo-platform-live into a guarded, observed proof instead of deterministic cluster-backed seeding
  • requires exact-context confirmation plus the explicit phrase INJECT LIVE FAULT
  • creates a two-replica HTTP target only in sentinel-live-demo, continuously measures reachability, and limits Chaos Mesh to one disposable replica
  • requires Argus/Falco evidence before fault injection, verifies the real PodChaos object, and proves a new Ready replacement plus two Ready replicas
  • publishes only verified observed + live_chaos evidence into Sentinel and emits an auditable JSON/Markdown scorecard
  • preserves the portable cluster-free judge demo and updates the Build Week/video plan

Why

The portable simulator proves integration safely, but judges also need a truthful live demonstration. Previously the k3s command still seeded deterministic evidence and could not substantiate actual detection, chaos execution, availability, or recovery claims. This phase adds that proof while preserving a narrow blast radius and human-in-the-loop authorization.

Safety boundary

  • dry-run creates nothing
  • existing namespaces are never reused
  • unsafe/system namespaces are rejected
  • fault duration is restricted to 5–30 seconds
  • exactly one of two demo replicas is targeted
  • cleanup deletes only the namespace created by this run
  • no fault runs without two explicit operator confirmations

Validation

  • bash scripts/tests/test-demo-platform-live.sh
  • bash scripts/tests/test-demo-platform.sh
  • bash -n for all changed shell scripts
  • git diff --check
  • make demo-platform-live-dry-run against context argus: all Cilium, Falco, Kyverno, Argus, Chaos Mesh, Phoenix, and Sentinel checks passed; no resources were changed
  • existing deterministic judge artifact successfully re-rendered with the updated report renderer

Stack

This PR is intentionally based on agent/cluster-free-platform-demo and should merge after the cluster-free demo PR. It leaves the default make demo-platform behavior unchanged.

@CodeBuildder
CodeBuildder marked this pull request as ready for review July 19, 2026 00:09
@CodeBuildder
CodeBuildder merged commit 5729b27 into agent/cluster-free-platform-demo Jul 19, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant