Merge branch 'main' into get-rid-of-shared-alerts

Signed-off-by: QuentinBisson <[email protected]>
giantswarm · Jun 11, 2024 · 61b77ec · 61b77ec
2 parents 380d5a0 + ff29140
commit 61b77ec
Show file tree

Hide file tree

Showing 4 changed files with 12 additions and 7 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -29,10 +29,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 - Fixed usage of yq, and jq in check-opsrecipes.sh
 - Fetch jq with make install-tools
-- Fixed and improve the check-opsrecipes.sh script so support <directory>/_index.md based ops-recipes.
+- Fixed and improve the check-opsrecipes.sh script to support <directory>/_index.md based ops-recipes.
 - Fixed cabbage alerts for multi-provider MCs.
 - Fixed all area alert labels.
 - Fixed `cert-exporter` alerts to page on all providers.
+- Fix `ManagementClusterDexAppMissing` use of absent for mimir.
 
 ### Removed
 

diff --git a/helm/prometheus-rules/templates/kaas/bigmac/alerting-rules/dex.rules.yml b/helm/prometheus-rules/templates/kaas/bigmac/alerting-rules/dex.rules.yml
@@ -41,7 +41,11 @@ spec:
       annotations:
         description: '{{`dex-operator did not register a dex-app in giantswarm namespace.`}}'
         opsrecipe: dex-operator/
+      {{- if .Values.mimir.enabled }}
+      expr: absent(dex_operator_idp_secret_expiry_time{app_namespace="giantswarm", cluster_type="management_cluster", cluster_id="{{ .Values.managementCluster.name }}", installation="{{ .Values.managementCluster.name }}", provider="{{ .Values.managementCluster.provider.kind }}", pipeline="{{ .Values.managementCluster.pipeline }}"})
+      {{- else }}
       expr: absent(dex_operator_idp_secret_expiry_time{app_namespace="giantswarm", cluster_type="management_cluster"}) == 1
+      {{- end }}
       for: 30m
       labels:
         area: kaas

diff --git a/helm/prometheus-rules/templates/kaas/turtles/alerting-rules/management-cluster.rules.yml b/helm/prometheus-rules/templates/kaas/turtles/alerting-rules/management-cluster.rules.yml
@@ -17,7 +17,7 @@ spec:
       annotations:
         description: '{{`Management cluster {{ $labels.cluster_id }} has less than 3 nodes.`}}'
         opsrecipe: management-cluster-less-than-three-workers/
-      expr: sum(kubelet_node_name{cluster_type="management_cluster"} * on (cluster_id, node) kube_node_role{role="worker", cluster_type="management_cluster"}) by (cluster_id) < 3
+      expr: sum(kubelet_node_name{cluster_type="management_cluster"} * on (cluster_id, node) kube_node_role{role="worker", cluster_type="management_cluster"}) by (cluster_id, installation, pipeline, provider) < 3
       for: 1h
       labels:
         area: kaas
@@ -28,7 +28,7 @@ spec:
     - alert: ManagementClusterMissingNodes
       annotations:
         description: '{{`Management cluster {{ $labels.cluster_id }} has less than 4 minimum nodes.`}}'
-      expr: sum(kube_node_status_condition{cluster_type="management_cluster", condition="Ready", status="true"}) by (cluster_id) < 4
+      expr: sum(kube_node_status_condition{cluster_type="management_cluster", condition="Ready", status="true"}) by (cluster_id, installation, pipeline, provider) < 4
       for: 15m
       labels:
         area: kaas
@@ -62,7 +62,7 @@ spec:
     - alert: ManagementClusterPodLimitAlmostReached
       annotations:
         description: '{{`Cluster {{ $labels.cluster_id }} is almost exceeding its pod limit.`}}'
-      expr: (sum(kube_pod_info{cluster_type="management_cluster"}) by (cluster_id) / sum(kube_node_status_capacity{resource="pods", cluster_type="management_cluster"}) by (cluster_id)) > 0.8
+      expr: (sum(kube_pod_info{cluster_type="management_cluster"}) by (cluster_id, installation, pipeline, provider) / sum(kube_node_status_capacity{resource="pods", cluster_type="management_cluster"}) by (cluster_id, installation, pipeline, provider)) > 0.8
       for: 5m
       labels:
         area: kaas

diff --git a/test/conf/promtool_ignore b/test/conf/promtool_ignore
@@ -1,7 +1,4 @@
 kaas/bigmac/alerting-rules/cert-manager.rules.yml
-kaas/bigmac/alerting-rules/certificate.all.rules.yml
-kaas/bigmac/alerting-rules/certificate.management-cluster.rules.yml
-kaas/bigmac/alerting-rules/certificate.workload-cluster.rules.yml
 kaas/bigmac/alerting-rules/dex.rules.yml
 kaas/bigmac/alerting-rules/teleport.rules.yml
 kaas/phoenix/alerting-rules/aws-load-balancer-controller.rules.yml
@@ -27,6 +24,9 @@ kaas/turtles/alerting-rules/capi-machinedeployment.rules.yml
 kaas/turtles/alerting-rules/capi-machinepool.rules.yml
 kaas/turtles/alerting-rules/capi-machineset.rules.yml
 kaas/turtles/alerting-rules/capi.management-cluster.rules.yml
+kaas/turtles/alerting-rules/certificate.all.rules.yml
+kaas/turtles/alerting-rules/certificate.management-cluster.rules.yml
+kaas/turtles/alerting-rules/certificate.workload-cluster.rules.yml
 kaas/turtles/alerting-rules/cluster-autoscaler.rules.yml
 kaas/turtles/alerting-rules/docker.rules.yml
 kaas/turtles/alerting-rules/etcd.management-cluster.rules.yml