Use ready replicas for KyvernoWebhookHasNoAvailableReplicas (#1224)

Co-authored-by: Quentin Bisson <[email protected]>
giantswarm · Jun 10, 2024 · 90fb609 · 90fb609
1 parent 479a2e8
commit 90fb609
Show file tree

Hide file tree

Showing 3 changed files with 3 additions and 2 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -25,6 +25,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Move the management cluster certificate alerts into the shared alerts because it is provider independent
 - Review and fix phoenix alerts towards Mimir and multi-provider MCs.
 - Moves cluster-autoscaler and vpa alerts to turtles.
+- Use `ready` replicas for Kyverno webhooks alert.
 - Moves ownership of alerts for shared components to turtles.
 
 ### Fixed

diff --git a/helm/prometheus-rules/templates/platform/shield/alerting-rules/kyverno.all.rules.yml b/helm/prometheus-rules/templates/platform/shield/alerting-rules/kyverno.all.rules.yml
@@ -14,7 +14,7 @@ spec:
       annotations:
         description: 'Kyverno has no available replicas but webhooks are present.'
         opsrecipe: kyverno-webhooks/
-      expr: sum(kube_validatingwebhookconfiguration_info{validatingwebhookconfiguration=~"kyverno-.*"}) by (cluster_id, installation, pipeline, provider) > 0 and sum(kube_deployment_status_replicas{deployment=~"kyverno|kyverno-admission-controller"}) by (cluster_id, installation, pipeline, provider) == 0
+      expr: sum(kube_validatingwebhookconfiguration_info{validatingwebhookconfiguration=~"kyverno-.*"}) by (cluster_id, installation, pipeline, provider) > 0 and sum(kube_deployment_status_replicas_ready{deployment=~"kyverno|kyverno-admission-controller"}) by (cluster_id, installation, pipeline, provider) == 0
       for: 15m
       labels:
         area: platform

diff --git a/test/tests/providers/global/platform/shield/alerting-rules/kyverno.all.rules.test.yml b/test/tests/providers/global/platform/shield/alerting-rules/kyverno.all.rules.test.yml
@@ -12,7 +12,7 @@ tests:
       - series: 'kube_validatingwebhookconfiguration_info{app="kube-state-metrics", cluster_id="gremlin", cluster_type="management_cluster", container="kube-state-metrics", customer="giantswarm", endpoint="http", installation="gremlin", instance="10.0.135.241:8080", job="kube-state-metrics", node="master-00000y", organization="giantswarm", pipeline="testing", pod="prometheus-operator-app-kube-state-metrics-d7f4ff68d-qn6sb", prometheus="kube-system/prometheus-agent", prometheus_replica="prometheus-prometheus-agent-0", provider="aws", region="germanywestcentral", service="prometheus-operator-app-kube-state-metrics", service_priority="highest", validatingwebhookconfiguration="kyverno-resource-validating-webhook-cfg"}'
         values: "1+0x20"
       # Kyverno deployment status replicas
-      - series: 'kube_deployment_status_replicas{app="kube-state-metrics", cluster_id="gremlin", cluster_type="management_cluster", container="kube-state-metrics", customer="giantswarm", deployment="kyverno", endpoint="http", installation="gremlin", instance="10.0.135.241:8080", job="kube-state-metrics", namespace="kyverno", node="master-00000y", organization="giantswarm", pipeline="testing", pod="prometheus-operator-app-kube-state-metrics-d7f4ff68d-qn6sb", prometheus="kube-system/prometheus-agent", prometheus_replica="prometheus-prometheus-agent-0", provider="aws", region="germanywestcentral", service="prometheus-operator-app-kube-state-metrics", service_priority="highest"}'
+      - series: 'kube_deployment_status_replicas_ready{app="kube-state-metrics", cluster_id="gremlin", cluster_type="management_cluster", container="kube-state-metrics", customer="giantswarm", deployment="kyverno", endpoint="http", installation="gremlin", instance="10.0.135.241:8080", job="kube-state-metrics", namespace="kyverno", node="master-00000y", organization="giantswarm", pipeline="testing", pod="prometheus-operator-app-kube-state-metrics-d7f4ff68d-qn6sb", prometheus="kube-system/prometheus-agent", prometheus_replica="prometheus-prometheus-agent-0", provider="aws", region="germanywestcentral", service="prometheus-operator-app-kube-state-metrics", service_priority="highest"}'
         values: "0+0x20"
       # Kyverno deployment spec replicas
       - series: 'kube_deployment_spec_replicas{app="kube-state-metrics", cluster_id="gremlin", cluster_type="management_cluster", container="kube-state-metrics", customer="giantswarm", deployment="kyverno", endpoint="http", installation="gremlin", instance="10.0.135.241:8080", job="kube-state-metrics", namespace="kyverno", node="master-00000y", organization="giantswarm", pipeline="testing", pod="prometheus-operator-app-kube-state-metrics-d7f4ff68d-qn6sb", prometheus="kube-system/prometheus-agent", prometheus_replica="prometheus-prometheus-agent-0", provider="aws", region="germanywestcentral", service="prometheus-operator-app-kube-state-metrics", service_priority="highest"}'