Update telemeter and regenerate rules

rhobs · Oct 16, 2024 · 9efffa7 · 9efffa7
1 parent 0262e4d
commit 9efffa7
Show file tree

Hide file tree

Showing 2 changed files with 32 additions and 2 deletions.
diff --git a/jsonnetfile.lock.json b/jsonnetfile.lock.json
@@ -293,8 +293,8 @@
           "subdir": "jsonnet/telemeter"
         }
       },
-      "version": "a9417a6062c3a31ed78c06ea3a0613a52f2029b2",
-      "sum": "IFJkQRq4+QFi0/7Rc5OVdqyV6KXNXeBp8hyFAcdEna8="
+      "version": "0453f452b7ca677e3be10faa95bea714733b4830",
+      "sum": "0m1kvO0SH4YoldGek69nhBGMTPxYz3gYcyNhfrymAOE="
     },
     {
       "source": {

diff --git a/resources/services/observatorium-metrics-template.yaml b/resources/services/observatorium-metrics-template.yaml
@@ -2027,6 +2027,11 @@ objects:
       - "interval": "4m"
         "name": "telemeter-telemeter.rules"
         "rules":
+        - "expr": |
+            count_over_time(vector(1)[1h:5m])
+          "labels":
+            "tenant_id": "FB870BF3-9F3A-44FF-9BF7-D7A047A52F43"
+          "record": "steps:count1h"
         - "expr": |
             count by (name,reason) (cluster_operator_conditions{condition="Degraded"} == 1)
           "labels":
@@ -2157,6 +2162,31 @@ objects:
           "labels":
             "tenant_id": "FB870BF3-9F3A-44FF-9BF7-D7A047A52F43"
           "record": "cluster:capacity_effective_cpu_cores"
+        - "expr": |
+            # self managed OpenShift cluster
+            max by (_id, managed_cluster_id) (acm_managed_cluster_info{product="OpenShift"}) * on(managed_cluster_id) group_left() (
+                # On one side, the acm_managed_cluster_info metric has the managed_cluster_id label identifiying the managed cluster and the _id label identifying the hub cluster.
+                # On the other side, the cluster:capacity_effective_cpu_cores metric has the _id label which identifying the managed cluster.
+                # To join the 2 metrics, we need to add a managed_cluster_id label with the same value as _id to the cluster:capacity_effective_cpu_cores metric.
+                label_replace(
+                  max by(_id) (cluster:capacity_effective_cpu_cores), "managed_cluster_id", "$1", "_id", "(.*)"
+                )
+              ) * 2 or
+            # managed OpenShift cluster and non-OpenShift clusters
+            max by (_id, managed_cluster_id) (acm_managed_cluster_worker_cores:max)
+          "labels":
+            "tenant_id": "FB870BF3-9F3A-44FF-9BF7-D7A047A52F43"
+          "record": "acm_capacity_effective_cpu_cores"
+        - "expr": |
+            max by(_id)(sum_over_time(hostedcluster:hypershift_cluster_vcpus:max[1h:5m])) / scalar(count_over_time(vector(1)[1h:5m]))
+          "labels":
+            "tenant_id": "FB870BF3-9F3A-44FF-9BF7-D7A047A52F43"
+          "record": "hostedcluster:hypershift_cluster_vcpus:vcpu_hours"
+        - "expr": |
+            hostedcluster:hypershift_cluster_vcpus:vcpu_hours or on (_id) cluster:usage:workload:capacity_virtual_cpu_hours
+          "labels":
+            "tenant_id": "FB870BF3-9F3A-44FF-9BF7-D7A047A52F43"
+          "record": "rosa:cluster:vcpu_hours"
   kind: ConfigMap
   metadata:
     annotations: