From f32066b842f20f0277eadbf0eab68000298ca3cb Mon Sep 17 00:00:00 2001
From: Daniel Simionato <daniel@giantswarm.io>
Date: Thu, 2 May 2024 18:12:33 +0200
Subject: [PATCH] Add `cluster_control_plane_unhealthy` inhibition.

---
 CHANGELOG.md                                             | 4 ++++
 .../alerting-rules/inhibit.management-cluster.rules.yml  | 9 +++++++++
 2 files changed, 13 insertions(+)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 396269241..b53cdeb4a 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -15,6 +15,10 @@ Add opsrecipe to `CoreDNSMaxHPAReplicasReached`
 
 - Remove cilium entry from KAAS SLOs.
 
+### Added
+
+- Add `cluster_control_plane_unhealthy` inhibition.
+
 ## [3.13.1] - 2024-04-30
 
 ### Removed
diff --git a/helm/prometheus-rules/templates/alerting-rules/inhibit.management-cluster.rules.yml b/helm/prometheus-rules/templates/alerting-rules/inhibit.management-cluster.rules.yml
index d23d0f156..12cf66a5f 100644
--- a/helm/prometheus-rules/templates/alerting-rules/inhibit.management-cluster.rules.yml
+++ b/helm/prometheus-rules/templates/alerting-rules/inhibit.management-cluster.rules.yml
@@ -95,4 +95,13 @@ spec:
         instance_state_not_running: "true"
         team: phoenix
         topic: status
+    - alert: InhibitionControlPlaneUnhealthy
+      annotations:
+        description: '{{`Control plane of cluster {{ $labels.cluster_id }} is not healthy.`}}'
+      expr: capi_kubeadmcontrolplane_status_condition{type="ControlPlaneComponentsHealthy", status="False"} == 1 or capi_kubeadmcontrolplane_status_condition{type="EtcdClusterHealthy", status="False"} == 1 or capi_kubeadmcontrolplane_status_condition{type="Available", status="False"} == 1
+      labels:
+        area: kaas
+        cluster_control_plane_unhealthy: "true"
+        team: phoenix
+        topic: status
 {{- end }}