Add cluster-info recording rule

giantswarm · Oct 12, 2023 · 1470252 · 1470252
1 parent 3fb9947
commit 1470252
Show file tree

Hide file tree

Showing 4 changed files with 45 additions and 0 deletions.
diff --git 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
+
+- Add recording rules to list all clusters.
+
 ## [2.138.1] - 2023-10-11
 
 ### Fixed

diff --git a/helm/prometheus-rules/templates/recording-rules/grafana-cloud.rules.yml b/helm/prometheus-rules/templates/recording-rules/grafana-cloud.rules.yml
@@ -58,6 +58,40 @@ spec:
       record: aggregation:giantswarm:app_upgrade_available
   - name: clusters.grafana-cloud.recording
     rules:
+    # This recording rule is used to list all clusters. The last expression is used to list vintage MCs
+    - expr: |-
+        avg by (cluster_id, customer, installation) (
+          label_replace(
+            capi_cluster_info,
+            "cluster_id",
+            "$1",
+            "name",
+            "(.*)"
+          )
+          or cluster_operator_cluster_status
+          or cluster_service_cluster_info
+          or
+            label_replace(
+              label_replace(
+                label_replace(
+                  vector(1),
+                  "cluster_id",
+                  {{ .Values.managementCluster.name | quote }},
+                  "",
+                  ""
+                ),
+                "customer",
+                {{ .Values.managementCluster.customer | quote }},
+                "",
+                ""
+              ),
+              "installation",
+              {{ .Values.managementCluster.name | quote }},
+              "",
+              ""
+            )
+        )
+      record: aggregation:giantswarm:cluster
     - expr: sum(label_replace(azure_operator_cluster_release{release_version!=""}, "cluster_id", "$1", "exported_cluster_id", "(.*)")) by (release_version, cluster_id) or sum(cluster_service_cluster_info) by (release_version, cluster_id) / 2 or sum(cluster_operator_cluster_status{release_version!=""}) by (release_version, cluster_id)
       record: aggregation:giantswarm:cluster_release_version
     - expr: avg_over_time(cluster_operator_cluster_create_transition[1w]) or avg_over_time(azure_operator_cluster_create_transition[1w])

diff --git a/helm/prometheus-rules/values.schema.json b/helm/prometheus-rules/values.schema.json
@@ -5,6 +5,9 @@
         "managementCluster": {
             "type": "object",
             "properties": {
+                "customer": {
+                    "type": "string"
+                },
                 "name": {
                     "type": "string"
                 },
@@ -14,6 +17,9 @@
                 "provider": {
                     "type": "object",
                     "properties": {
+                        "flavor": {
+                            "type": "string"
+                        },
                         "kind": {
                             "type": "string"
                         }

diff --git a/helm/prometheus-rules/values.yaml b/helm/prometheus-rules/values.yaml
@@ -5,6 +5,7 @@ project:
   branch: '[[ .Branch ]]'
   commit: '[[ .SHA ]]'
 managementCluster:
+  customer: ""
   name: ""
   pipeline: ""
   provider: