perf(ci): stop the integration and e2e jobs waiting on rook-ceph (#435)

charts/platform/rook-ceph-cluster/templates/cephcluster.yaml

@@ -44,3 +44,17 @@ spec:
     daemonHealth:
       mon:
         interval: 45s
+      # Bootstrap is a serial chain — mon, then mgr, then OSD, then the object
+      # store — and the operator only advances it when a status check observes
+      # the previous daemon up. At the 60s default that polling dominates.
+      status:
+        interval: {{ .Values.probes.statusInterval }}
+    # These daemons come up in a couple of seconds, but Rook's default probe
+    # waits 10s and then polls every 10s, so each one rounds up to ~20s of
+    # nothing. Finer granularity, same total budget — see values.yaml.
+    startupProbe:
+{{- range $daemon, $probe := .Values.probes.startup }}
+      {{ $daemon }}:
+        probe:
+{{ toYaml $probe | indent 10 }}
+{{- end }}

charts/platform/rook-ceph-cluster/templates/cephobjectstore.yaml

@@ -14,6 +14,15 @@ spec:
       size: 1
       requireSafeReplicaSize: false
   preservePoolsOnDelete: false
+  # Same reasoning as the daemon probes in cephcluster.yaml — the RGW is the
+  # last link in the bootstrap chain and the one the S3 suites wait on.
+  healthCheck:
+    startupProbe:
+      probe:
+{{ toYaml .Values.probes.rgw.startup | indent 8 }}
+    readinessProbe:
+      probe:
+{{ toYaml .Values.probes.rgw.readiness | indent 8 }}
   gateway:
     port: {{ .Values.objectStore.gateway.port }}
     instances: {{ .Values.objectStore.gateway.instances }}

charts/platform/rook-ceph-cluster/values.yaml

@@ -11,6 +11,21 @@ mgr:
 dashboard:
   enabled: false
 
+# Bootstrap-latency tuning. Rook's stock probes are sized for real hardware:
+# 10s of initial delay then a 10s poll, which on a dev cluster is almost all of
+# each daemon's startup time. failureThreshold is scaled by the same factor the
+# period shrinks by, so every daemon keeps the total budget it had by default
+# (mon/mgr 10s + 6x10s; osd 10s + 720x10s; rgw 10s + 33x10s).
+probes:
+  statusInterval: 10s
+  startup:
+    mon: { initialDelaySeconds: 2, periodSeconds: 2, failureThreshold: 35 }
+    mgr: { initialDelaySeconds: 2, periodSeconds: 2, failureThreshold: 35 }
+    osd: { initialDelaySeconds: 2, periodSeconds: 2, failureThreshold: 3600 }
+  rgw:
+    startup: { initialDelaySeconds: 2, periodSeconds: 2, failureThreshold: 170 }
+    readiness: { initialDelaySeconds: 2, periodSeconds: 2 }
+
 storage:
   # Rook does NOT auto-select loop devices via useAllDevices (even with
   # allowLoopDevices on the operator) — they must be named explicitly. This is