MON-4115: expose label metrics for jobs and cronjobs

Adds `AdditionalLabelsAllowList` to KSM config.

Signed-off-by: Pranshu Srivastava <[email protected]>

Author: rexagod

CHANGELOG.md

@@ -1,5 +1,9 @@
 # Note: This CHANGELOG is only for the monitoring team to track all monitoring related changes. Please see OpenShift release notes for official changes.
 
+## 4.21
+
+- [#2553](https://github.com/openshift/cluster-monitoring-operator/pull/2553) Expose an `additionalLabelsAllowList` field in CMO's KubeStateMetrics configuration to allow users to specify which additional label metrics from Kubernetes objects should be exposed by KSM.
+
 ## 4.20
 
 - [#2595](https://github.com/openshift/cluster-monitoring-operator/pull/2595) Multi-tenant support for KSM's CRS feature-set downstream.

Documentation/api.md

@@ -174,6 +174,7 @@ The `KubeStateMetricsConfig` resource defines settings for the `kube-state-metri
 | resources | *[v1.ResourceRequirements](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.33/#resourcerequirements-v1-core) | Defines resource requests and limits for the KubeStateMetrics container. |
 | tolerations | [][v1.Toleration](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.33/#toleration-v1-core) | Defines tolerations for the pods. |
 | topologySpreadConstraints | []v1.TopologySpreadConstraint | Defines a pod's topology spread constraints. |
+| additionalLabelsAllowList | *string | Defines label-metrics' allow list for resources in addition to the default one. Currently, this is only supported for `jobs` and `cronjobs`, due to cardinality concerns. This follows the format: resource1=[label1,label2,labelN...],...,resourceN=[...], which is then validated and appended to the default labels' allow list. |
 
 [Back to TOC](#table-of-contents)
 

Documentation/openshiftdocs/modules/kubestatemetricsconfig.adoc

@@ -26,6 +26,8 @@ Appears in: link:clustermonitoringconfiguration.adoc[ClusterMonitoringConfigurat
 
 |topologySpreadConstraints|[]v1.TopologySpreadConstraint|Defines a pod's topology spread constraints.
 
+|additionalLabelsAllowList|*string|Defines label-metrics' allow list for resources in addition to the default one. Currently, this is only supported for `jobs` and `cronjobs`, due to cardinality concerns. This follows the format: resource1=[label1,label2,labelN...],...,resourceN=[...], which is then validated and appended to the default labels' allow list.
+
 |===
 
 link:../index.adoc[Back to TOC]

pkg/manifests/manifests.go

@@ -48,6 +48,7 @@ import (
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/util/sets"
 	auditv1 "k8s.io/apiserver/pkg/apis/audit/v1"
+	"k8s.io/klog/v2"
 	apiregistrationv1 "k8s.io/kube-aggregator/pkg/apis/apiregistration/v1"
 	"k8s.io/utils/ptr"
 	k8syaml "sigs.k8s.io/yaml"
@@ -760,6 +761,34 @@ func (f *Factory) KubeStateMetricsDeployment() (*appsv1.Deployment, error) {
 			if f.config.ClusterMonitoringConfiguration.KubeStateMetricsConfig.Resources != nil {
 				d.Spec.Template.Spec.Containers[i].Resources = *f.config.ClusterMonitoringConfiguration.KubeStateMetricsConfig.Resources
 			}
+			additionalAllowList := f.config.ClusterMonitoringConfiguration.KubeStateMetricsConfig.AdditionalLabelsAllowList
+			if additionalAllowList != nil && *additionalAllowList != "" {
+				err = validateLabelsAllowListFormat(*additionalAllowList)
+				if err != nil {
+					return nil, fmt.Errorf("error parsing allowlist: %w", err)
+				}
+				for i = range container.Args {
+					if strings.HasPrefix(container.Args[i], "--metric-labels-allowlist=") {
+						allowedResources := sets.New[string]("jobs", "cronjobs")
+						gotResourcesKeyValues := strings.Split(*additionalAllowList, ",")
+						acceptedResources := make([]string, 0, len(gotResourcesKeyValues))
+						for _, keyValue := range gotResourcesKeyValues {
+							parts := strings.SplitN(keyValue, "=", 2)
+							if len(parts) != 2 {
+								return nil, fmt.Errorf("error parsing allowlist: %w", ErrConfigValidation)
+							}
+							if allowedResources.Has(parts[0]) {
+								acceptedResources = append(acceptedResources, keyValue)
+							} else {
+								klog.V(4).Infof("ignoring unsupported resource %q in additional labels allowlist", parts[0])
+							}
+						}
+						if len(acceptedResources) > 0 {
+							container.Args[i] += "," + strings.Join(acceptedResources, ",")
+						}
+					}
+				}
+			}
 		}
 	}
 
@@ -3614,3 +3643,65 @@ func hashStringMap(m map[string]string) string {
 	}
 	return hashByteMap(byteMap)
 }
+
+func validateLabelsAllowListFormat(value string) error {
+	var errLabelsAllowListFormat = errors.New("invalid format, should be: resource1=[label1,label2,labelN...],...,resourceN=[...]")
+
+	// Taken from text/scanner EOF constant.
+	const EOF = -1
+	var (
+		m            = map[string][]string{}
+		previous     rune
+		next         rune
+		firstWordPos int
+		name         string
+	)
+	firstWordPos = 0
+
+	for i, v := range value {
+		if i+1 == len(value) {
+			next = EOF
+		} else {
+			next = []rune(value)[i+1]
+		}
+		if i-1 >= 0 {
+			previous = []rune(value)[i-1]
+		} else {
+			previous = v
+		}
+
+		switch v {
+		case '=':
+			if previous == ',' || next != '[' {
+				return errLabelsAllowListFormat
+			}
+			name = strings.TrimSpace(string([]rune(value)[firstWordPos:i]))
+			m[name] = []string{}
+			firstWordPos = i + 1
+		case '[':
+			if previous != '=' {
+				return errLabelsAllowListFormat
+			}
+			firstWordPos = i + 1
+		case ']':
+			// if after metric group, has char not comma or end.
+			if next != EOF && next != ',' {
+				return errLabelsAllowListFormat
+			}
+			if previous != '[' {
+				m[name] = append(m[name], strings.TrimSpace(string(([]rune(value)[firstWordPos:i]))))
+			}
+			firstWordPos = i + 1
+		case ',':
+			// if starts or ends with comma
+			if previous == v || next == EOF || next == ']' {
+				return errLabelsAllowListFormat
+			}
+			if previous != ']' {
+				m[name] = append(m[name], strings.TrimSpace(string(([]rune(value)[firstWordPos:i]))))
+			}
+			firstWordPos = i + 1
+		}
+	}
+	return nil
+}

pkg/manifests/types.go

@@ -175,6 +175,11 @@ type KubeStateMetricsConfig struct {
 	Tolerations []v1.Toleration `json:"tolerations,omitempty"`
 	// Defines a pod's topology spread constraints.
 	TopologySpreadConstraints []v1.TopologySpreadConstraint `json:"topologySpreadConstraints,omitempty"`
+	// Defines label-metrics' allow list for resources in addition to the default one.
+	// Currently, this is only supported for `jobs` and `cronjobs`, due to cardinality concerns.
+	// This follows the format: resource1=[label1,label2,labelN...],...,resourceN=[...],
+	// which is then validated and appended to the default labels' allow list.
+	AdditionalLabelsAllowList *string `json:"additionalLabelsAllowList,omitempty"`
 }
 
 // The `PrometheusK8sConfig` resource defines settings for the Prometheus