From c3d9f2a1a5d909ce27bcbd3fc63a8d39bc536b02 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Herv=C3=A9=20Nicol?= <hervenicol@users.noreply.github.com>
Date: Tue, 22 Oct 2024 09:55:26 +0200
Subject: [PATCH] Add InhibitionClusterWithoutWorkerNodes for CAPA (#1397)

Co-authored-by: Herve Nicol <12008875+hervenicol@users.noreply.github.com>
Co-authored-by: Quentin Bisson <quentin@giantswarm.io>
---
 CHANGELOG.md                                  |  4 ++
 .../alerting-rules/capa.inhibition.rules.yml  | 34 ++++++++++++++
 .../capa.inhibition.rules.test.yml            | 47 +++++++++++++++++++
 3 files changed, 85 insertions(+)
 create mode 100644 helm/prometheus-rules/templates/kaas/phoenix/alerting-rules/capa.inhibition.rules.yml
 create mode 100644 test/tests/providers/capi/capa-mimir/kaas/phoenix/alerting-rules/capa.inhibition.rules.test.yml

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 9d8c2850..6d5540b3 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -7,6 +7,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Added
+
+- Added InhibitionClusterWithoutWorkerNodes for CAPA
+
 ### Changed
 
 - Modify `KyvernoWebhookHasNoAvailableReplicas` to check specifically for Kyverno resource webhook.
diff --git a/helm/prometheus-rules/templates/kaas/phoenix/alerting-rules/capa.inhibition.rules.yml b/helm/prometheus-rules/templates/kaas/phoenix/alerting-rules/capa.inhibition.rules.yml
new file mode 100644
index 00000000..b5767549
--- /dev/null
+++ b/helm/prometheus-rules/templates/kaas/phoenix/alerting-rules/capa.inhibition.rules.yml
@@ -0,0 +1,34 @@
+{{- if eq .Values.managementCluster.provider.kind "capa" }}
+apiVersion: monitoring.coreos.com/v1
+kind: PrometheusRule
+metadata:
+  creationTimestamp: null
+  labels:
+    {{- include "labels.common" . | nindent 4 }}
+    cluster_type: "management_cluster"
+  name: capa.inhibitions.rules
+  namespace: {{ .Values.namespace  }}
+spec:
+  groups:
+  - name: capa.inhibitions
+    rules:
+    - alert: InhibitionClusterWithoutWorkerNodes
+      annotations:
+        description: '{{`Cluster ({{ $labels.cluster_id }}) has no worker nodes.`}}'
+      expr: |-
+        label_replace(
+            capi_cluster_status_condition{type="ControlPlaneReady", status="True"},
+                "cluster_id",
+                "$1",
+                "name",
+                "(.*)"
+            ) == 1
+        unless on (cluster_id) (
+            sum(capi_machinepool_spec_replicas{} > 0) by (cluster_id)
+        )
+      labels:
+        area: kaas
+        has_worker_nodes: "false"
+        team: phoenix
+        topic: status
+{{- end }}
diff --git a/test/tests/providers/capi/capa-mimir/kaas/phoenix/alerting-rules/capa.inhibition.rules.test.yml b/test/tests/providers/capi/capa-mimir/kaas/phoenix/alerting-rules/capa.inhibition.rules.test.yml
new file mode 100644
index 00000000..0615f71b
--- /dev/null
+++ b/test/tests/providers/capi/capa-mimir/kaas/phoenix/alerting-rules/capa.inhibition.rules.test.yml
@@ -0,0 +1,47 @@
+---
+rule_files:
+- capa.inhibition.rules.yml
+
+tests:
+  # Tests for `InhibitionClusterWithoutWorkerNodes` inhibition alert
+  - interval: 1m
+    input_series:
+      - series: 'capi_cluster_status_condition{cluster_id="golem", cluster_type="management_cluster", name="golem", pipeline="testing", status="True", type="ControlPlaneReady"}'
+        values: "1+0x300"
+      - series: 'capi_machinepool_spec_replicas{cluster_id="golem", cluster_name="golem", cluster_type="management_cluster", customer="giantswarm", installation="golem", organization="giantswarm", pipeline="testing", provider="capa"}'
+        values: "_x60 0x60 3x60"
+    alert_rule_test:
+      - alertname: InhibitionClusterWithoutWorkerNodes
+        eval_time: 30m
+        exp_alerts:
+          - exp_labels:
+              area: kaas
+              cluster_id: "golem"
+              cluster_type: "management_cluster"
+              has_worker_nodes: "false"
+              name: "golem"
+              pipeline: "testing"
+              status: "True"
+              team: "phoenix"
+              topic: "status"
+              type: "ControlPlaneReady"
+            exp_annotations:
+              description: "Cluster (golem) has no worker nodes."
+      - alertname: InhibitionClusterWithoutWorkerNodes
+        eval_time: 90m
+        exp_alerts:
+          - exp_labels:
+              area: kaas
+              cluster_id: "golem"
+              cluster_type: "management_cluster"
+              has_worker_nodes: "false"
+              name: "golem"
+              pipeline: "testing"
+              status: "True"
+              team: "phoenix"
+              topic: "status"
+              type: "ControlPlaneReady"
+            exp_annotations:
+              description: "Cluster (golem) has no worker nodes."
+      - alertname: InhibitionClusterWithoutWorkerNodes
+        eval_time: 150m