# Bookstore — Part 11 ch.07 "Chaos engineering": EXPERIMENT 3 — burn CPU
# inside catalog Pods for 60s and verify the HPA + resource limits keep the
# SLO. Tests whether resource pressure triggers the designed reaction
# (autoscale / throttle within limits) rather than a latency collapse.
#
# THE HYPOTHESIS THIS TESTS
#   "Two CPU-burning workers inside catalog Pods for 60s drive CPU toward the
#   catalog container limit (raw-manifests/10-catalog-deploy.yaml resources);
#   the HPA (raw-manifests/82-hpa-catalog.yaml) scales catalog out and the
#   catalog SLO (81-prometheusrule.yaml) is preserved — i.e. requests/limits
#   + autoscaling absorb CPU pressure as designed (Part 01 ch.03 / Part 06
#   ch.04), it does not become a user-visible outage."
#
# SELECTOR matches the REAL catalog pod label (app=catalog, ns bookstore;
# verified against 10-catalog-deploy.yaml). `mode: one` keeps the blast
# radius minimal (one Pod) — climb the blast-radius ladder (one → fixed →
# all) only as confidence grows.
#
# BOUNDED & REVERSIBLE: `duration: 60s` auto-reverts; the stressor runs in a
# Chaos-Mesh-managed process scoped to the target container's cgroup and is
# killed when the experiment ends or the object is deleted. The catalog Pods
# are NOT mutated (no injected container) and stay PSA-restricted-compliant;
# the cpu limit is the natural ceiling on the damage. `kubectl delete -f`
# aborts early.
#
# !!! CRD-INTRINSIC DRY-RUN (identical precedent to 18-/51-/70-/83-/argocd) !!!
#   `StressChaos` is a Chaos Mesh CRD (chaos-mesh.org/v1alpha1). WITHOUT Chaos
#   Mesh installed a client dry-run prints:
#     no matches for kind "StressChaos" in version "chaos-mesh.org/v1alpha1"
#   EXPECTED, schema-correct — Chaos Mesh CRDs/controller must be installed
#   first. Whole-dir dry-run prints this for CRD-backed files only and
#   continues. Schema verified against Chaos Mesh chaos-mesh.org/v1alpha1
#   StressChaos.
#
# Requires: Chaos Mesh installed (own ns); the Bookstore running with the
#   catalog HPA (82-) + metrics (80-/81-) so the reaction is observable.
# Apply:
#   kubectl apply -f examples/bookstore/chaos/30-stresschaos-catalog-cpu.yaml
#   # observe: kubectl get hpa catalog -n bookstore -w   (replicas should rise)
#   kubectl delete -f examples/bookstore/chaos/30-stresschaos-catalog-cpu.yaml
apiVersion: chaos-mesh.org/v1alpha1
kind: StressChaos
metadata:
  name: catalog-cpu-stress
  namespace: bookstore
  labels:
    app.kubernetes.io/part-of: bookstore
spec:
  mode: one                       # SMALLEST blast radius: one catalog Pod
  selector:
    namespaces:
      - bookstore
    labelSelectors:
      app: catalog                # REAL catalog pod label (10-)
  stressors:
    cpu:
      workers: 2                  # 2 CPU-burn workers (bounded by the
                                  # container cpu LIMIT — Part 01 ch.03)
      load: 100
  duration: "60s"                 # BOUNDED — auto-reverts after 60s