Merge pull request #286 from kakkoyun/add_example_recording_rules

Add first Telemeter recording rules
openshift · Dec 19, 2019 · 04e8962 · 04e8962
2 parents 24d6d00 + b415e9a
commit 04e8962
Show file tree

Hide file tree

Showing 4 changed files with 37 additions and 6 deletions.
diff --git a/docs/data-collection.md b/docs/data-collection.md
@@ -151,10 +151,10 @@ For the OpenShift 4 Developer Preview we will be sending back these exact attrib
   '{__name__="noobaa_total_usage"}',
   // console_url is the url of the console running on the cluster.
   '{__name__="console_url"}',
-  // cluster:network_attachment_definition_instances:max" gives max no of instance 
+  // cluster:network_attachment_definition_instances:max" gives max no of instance
   //in the cluster that are annotated with k8s.v1.cni.cncf.io/networks, labelled by networks.
   '{__name__="cluster:network_attachment_definition_instances:max"}',
-  // cluster:network_attachment_definition_enabled_instance_up  informs (1 or 0) if the cluster has 
+  // cluster:network_attachment_definition_enabled_instance_up  informs (1 or 0) if the cluster has
   //at least max of one instance with  k8s.v1.cni.cncf.io/networks annotation, labelled by networks (any or sriov).
   '{__name__="cluster:network_attachment_definition_enabled_instance_up:max"}',
   // insightsclient_request_send tracks the number of metrics sends.

diff --git a/jsonnet/telemeter/metrics.jsonnet b/jsonnet/telemeter/metrics.jsonnet
@@ -143,10 +143,10 @@
   '{__name__="noobaa_total_usage"}',
   // console_url is the url of the console running on the cluster.
   '{__name__="console_url"}',
-  // cluster:network_attachment_definition_instances:max" gives max no of instance 
+  // cluster:network_attachment_definition_instances:max" gives max no of instance
   //in the cluster that are annotated with k8s.v1.cni.cncf.io/networks, labelled by networks.
   '{__name__="cluster:network_attachment_definition_instances:max"}',
-  // cluster:network_attachment_definition_enabled_instance_up  informs (1 or 0) if the cluster has 
+  // cluster:network_attachment_definition_enabled_instance_up  informs (1 or 0) if the cluster has
   //at least max of one instance with  k8s.v1.cni.cncf.io/networks annotation, labelled by networks (any or sriov).
   '{__name__="cluster:network_attachment_definition_enabled_instance_up:max"}',
   // insightsclient_request_send tracks the number of metrics sends.

diff --git a/jsonnet/telemeter/prometheus/kubernetes.libsonnet b/jsonnet/telemeter/prometheus/kubernetes.libsonnet
@@ -42,8 +42,8 @@ local k = import 'ksonnet/ksonnet.beta.3/k.libsonnet';
       serviceAccount.mixin.metadata.withNamespace($._config.namespace) +
       serviceAccount.mixin.metadata.withAnnotations({
         // TODO: Remove observatorium-thanos-querier once we have a separate clusterRole
-        'serviceaccounts.openshift.io/oauth-redirectreference.observatorium-thanos-querier':'{"kind":"OAuthRedirectReference","apiVersion":"v1","reference":{"kind":"Route","name":"observatorium-thanos-querier"}}',
-        'serviceaccounts.openshift.io/oauth-redirectreference.observatorium-thanos-querier-cache':'{"kind":"OAuthRedirectReference","apiVersion":"v1","reference":{"kind":"Route","name":"observatorium-thanos-querier-cache"}}',
+        'serviceaccounts.openshift.io/oauth-redirectreference.observatorium-thanos-querier': '{"kind":"OAuthRedirectReference","apiVersion":"v1","reference":{"kind":"Route","name":"observatorium-thanos-querier"}}',
+        'serviceaccounts.openshift.io/oauth-redirectreference.observatorium-thanos-querier-cache': '{"kind":"OAuthRedirectReference","apiVersion":"v1","reference":{"kind":"Route","name":"observatorium-thanos-querier-cache"}}',
         'serviceaccounts.openshift.io/oauth-redirectreference.prometheus-k8s': '{"kind":"OAuthRedirectReference","apiVersion":"v1","reference":{"kind":"Route","name":"prometheus-telemeter"}}',
         'serviceaccounts.openshift.io/oauth-redirectreference.observatorium-jaeger': '{"kind":"OAuthRedirectReference","apiVersion":"v1","reference":{"kind":"Route","name":"observatorium-jaeger"}}',
       }),

diff --git a/jsonnet/telemeter/rules.libsonnet b/jsonnet/telemeter/rules.libsonnet
@@ -0,0 +1,31 @@
+{
+  prometheus+:: {
+    recordingrules+: {
+      groups+: [
+        {
+          name: 'telemeter.rules',
+          rules: [
+            {
+              record: 'name_reason:cluster_operator_degraded:count',
+              expr: |||
+                count by (name,reason) (cluster_operator_conditions{condition="Degraded"} == 1)
+              |||,
+            },
+            {
+              record: 'name_reason:cluster_operator_unavailable:count',
+              expr: |||
+                count by (name,reason) (cluster_operator_conditions{condition="Available"} == 0)
+              |||,
+            },
+            {
+              record: 'id_code:apiserver_request_error_rate_sum:max',
+              expr: |||
+                sort_desc(max by (_id,code) (code:apiserver_request_count:rate:sum{code=~"(4|5)\\d\\d"}) > 0.5)
+              |||,
+            },
+          ],
+        },
+      ],
+    },
+  },
+}