MON-3487: incorporate MetricsCollectionProfiles

Put the collection-profiles feature-set behind a dedicated feature-gate.

Signed-off-by: Pranshu Srivastava <rexagod@gmail.com>

go.mod

@@ -10,7 +10,7 @@ require (
 	github.com/ghodss/yaml v1.0.0
 	github.com/go-kit/log v0.2.1
 	github.com/imdario/mergo v0.3.16
-	github.com/openshift/api v0.0.0-20240320105957-b09bc66e1760
+	github.com/openshift/api v0.0.0-20240327143419-1e5b8d673d96
 	github.com/openshift/client-go v0.0.0-20231218140158-47f6d749b9d9
 	github.com/openshift/library-go v0.0.0-20240216151214-738f3fa4ccf8
 	github.com/prometheus-operator/prometheus-operator/pkg/apis/monitoring v0.71.0

go.sum

@@ -390,8 +390,8 @@ github.com/opencontainers/go-digest v1.0.0 h1:apOUWs51W5PlhuyGyz9FCeeBIOUDA/6nW8
 github.com/opencontainers/go-digest v1.0.0/go.mod h1:0JzlMkj0TRzQZfJkVvzbP0HBR3IKzErnv2BNG4W4MAM=
 github.com/opencontainers/image-spec v1.0.2 h1:9yCKha/T5XdGtO0q9Q9a6T5NUCsTn/DrBg0D7ufOcFM=
 github.com/opencontainers/image-spec v1.0.2/go.mod h1:BtxoFyWECRxE4U/7sNtV5W15zMzWCbyJoFRP3s7yZA0=
-github.com/openshift/api v0.0.0-20240320105957-b09bc66e1760 h1:UzSlJWW6VKseRrpUO05xh2AyGyZ+OSbuCJjst5STVdM=
-github.com/openshift/api v0.0.0-20240320105957-b09bc66e1760/go.mod h1:CxgbWAlvu2iQB0UmKTtRu1YfepRg1/vJ64n2DlIEVz4=
+github.com/openshift/api v0.0.0-20240327143419-1e5b8d673d96 h1:H6N843FYq/qnQkAgU8qNxSI6GkKxGkUasDFjMfb0D2s=
+github.com/openshift/api v0.0.0-20240327143419-1e5b8d673d96/go.mod h1:CxgbWAlvu2iQB0UmKTtRu1YfepRg1/vJ64n2DlIEVz4=
 github.com/openshift/client-go v0.0.0-20231218140158-47f6d749b9d9 h1:kjgW3luAkf9NWu+8u+jqNNbexDG+CY82/INw8hGbG14=
 github.com/openshift/client-go v0.0.0-20231218140158-47f6d749b9d9/go.mod h1:kKmxYRXTMutfF7XzGppFdbLhNGX1brXkRsZx5ID8c7U=
 github.com/openshift/library-go v0.0.0-20240216151214-738f3fa4ccf8 h1:dKtHGYiOwl0DKZEWBW4MFWFS6IYW02AVD1WSuUAVwEo=

pkg/client/client.go

@@ -595,15 +595,6 @@ func (c *Client) GetConsoleConfig(ctx context.Context, name string) (*configv1.C
 	return c.oscclient.ConfigV1().Consoles().Get(ctx, name, metav1.GetOptions{})
 }
 
-func (c *Client) TechPreviewEnabled(ctx context.Context) (bool, error) {
-	fg, err := c.oscclient.ConfigV1().FeatureGates().Get(ctx, "cluster", metav1.GetOptions{})
-	if err != nil {
-		return false, err
-	}
-
-	return fg.Spec.FeatureSet == configv1.TechPreviewNoUpgrade, nil
-}
-
 func (c *Client) GetConfigmap(ctx context.Context, namespace, name string) (*v1.ConfigMap, error) {
 	return c.kclient.CoreV1().ConfigMaps(namespace).Get(ctx, name, metav1.GetOptions{})
 }

pkg/manifests/config.go

@@ -49,9 +49,9 @@ const (
 )
 
 type Config struct {
-	Images      *Images `json:"-"`
-	RemoteWrite bool    `json:"-"`
-	TechPreview bool    `json:"-"`
+	Images                               *Images `json:"-"`
+	RemoteWrite                          bool    `json:"-"`
+	CollectionProfilesFeatureGateEnabled bool    `json:"-"`
 
 	ClusterMonitoringConfiguration *ClusterMonitoringConfiguration `json:"-"`
 	UserWorkloadConfiguration      *UserWorkloadConfiguration      `json:"-"`
@@ -190,8 +190,8 @@ func (cps CollectionProfiles) String() string {
 	return sb.String()
 }
 
-func NewConfig(content io.Reader, tp bool) (*Config, error) {
-	c := Config{TechPreview: tp}
+func NewConfig(content io.Reader, collectionProfilesFeatureGateEnabled bool) (*Config, error) {
+	c := Config{CollectionProfilesFeatureGateEnabled: collectionProfilesFeatureGateEnabled}
 	cmc := defaultClusterMonitoringConfiguration()
 	err := k8syaml.NewYAMLOrJSONDecoder(content, 4096).Decode(&cmc)
 	if err != nil {
@@ -438,12 +438,12 @@ func (c *Config) LoadEnforcedBodySizeLimit(pcr PodCapacityReader, ctx context.Co
 }
 
 func (c *Config) Precheck() error {
-	if c.ClusterMonitoringConfiguration.PrometheusK8sConfig.CollectionProfile != FullCollectionProfile && !c.TechPreview {
-		return fmt.Errorf("collectionProfiles is a TechPreview feature, to be able to use a profile different from the default (\"full\") please enable TechPreview: %w", ErrConfigValidation)
+	if c.ClusterMonitoringConfiguration.PrometheusK8sConfig.CollectionProfile != FullCollectionProfile && !c.CollectionProfilesFeatureGateEnabled {
+		return fmt.Errorf("%w: collectionProfiles is currently a TechPreview feature behind the \"MetricsCollectionProfiles\" feature-gate, to be able to use a profile different from the default (\"full\") please enable it first", ErrConfigValidation)
 	}
 
 	// Validate the configured collection profile iff tech preview is enabled, even if the default profile is set.
-	if c.TechPreview {
+	if c.CollectionProfilesFeatureGateEnabled {
 		for _, profile := range SupportedCollectionProfiles {
 			var v float64
 			if profile == c.ClusterMonitoringConfiguration.PrometheusK8sConfig.CollectionProfile {
@@ -484,12 +484,12 @@ func calculateBodySizeLimit(podCapacity int) string {
 // structure that facilitates programmatical checks of that configuration. The
 // content of the data structure might change if TechPreview is enabled (tp), as
 // some features are only meant for TechPreview.
-func NewConfigFromString(content string, tp bool) (*Config, error) {
+func NewConfigFromString(content string, collectionProfilesFeatureGateEnabled bool) (*Config, error) {
 	if content == "" {
 		return NewDefaultConfig(), nil
 	}
 
-	return NewConfig(bytes.NewBuffer([]byte(content)), tp)
+	return NewConfig(bytes.NewBuffer([]byte(content)), collectionProfilesFeatureGateEnabled)
 }
 
 func NewDefaultConfig() *Config {

pkg/manifests/manifests.go

@@ -727,7 +727,7 @@ func (f *Factory) KubeStateMetricsClusterRole() (*rbacv1.ClusterRole, error) {
 }
 
 func (f *Factory) KubeStateMetricsServiceMonitors() ([]*monv1.ServiceMonitor, error) {
-	return serviceMonitors(f.config.TechPreview, f.KubeStateMetricsServiceMonitor, f.KubeStateMetricsMinimalServiceMonitor)
+	return serviceMonitors(f.config.CollectionProfilesFeatureGateEnabled, f.KubeStateMetricsServiceMonitor, f.KubeStateMetricsMinimalServiceMonitor)
 }
 
 func (f *Factory) KubeStateMetricsServiceMonitor() (*monv1.ServiceMonitor, error) {
@@ -847,7 +847,7 @@ func (f *Factory) OpenShiftStateMetricsRBACProxySecret() (*v1.Secret, error) {
 }
 
 func (f *Factory) NodeExporterServiceMonitors() ([]*monv1.ServiceMonitor, error) {
-	return serviceMonitors(f.config.TechPreview, f.NodeExporterServiceMonitor, f.NodeExporterMinimalServiceMonitor)
+	return serviceMonitors(f.config.CollectionProfilesFeatureGateEnabled, f.NodeExporterServiceMonitor, f.NodeExporterMinimalServiceMonitor)
 }
 
 func (f *Factory) NodeExporterServiceMonitor() (*monv1.ServiceMonitor, error) {
@@ -1964,7 +1964,7 @@ func (f *Factory) PrometheusAdapterService() (*v1.Service, error) {
 }
 
 func (f *Factory) PrometheusAdapterServiceMonitors() ([]*monv1.ServiceMonitor, error) {
-	return serviceMonitors(f.config.TechPreview, f.PrometheusAdapterServiceMonitor, f.PrometheusAdapterMinimalServiceMonitor)
+	return serviceMonitors(f.config.CollectionProfilesFeatureGateEnabled, f.PrometheusAdapterServiceMonitor, f.PrometheusAdapterMinimalServiceMonitor)
 }
 
 func (f *Factory) PrometheusAdapterServiceMonitor() (*monv1.ServiceMonitor, error) {
@@ -2627,7 +2627,7 @@ func (f *Factory) ControlPlanePrometheusRule() (*monv1.PrometheusRule, error) {
 }
 
 func (f *Factory) ControlPlaneKubeletServiceMonitors() ([]*monv1.ServiceMonitor, error) {
-	return serviceMonitors(f.config.TechPreview, f.ControlPlaneKubeletServiceMonitor, f.ControlPlaneKubeletMinimalServiceMonitor)
+	return serviceMonitors(f.config.CollectionProfilesFeatureGateEnabled, f.ControlPlaneKubeletServiceMonitor, f.ControlPlaneKubeletMinimalServiceMonitor)
 }
 
 func (f *Factory) ControlPlaneKubeletServiceMonitor() (*monv1.ServiceMonitor, error) {

pkg/metrics/metrics.go

@@ -5,6 +5,12 @@ import (
 	"k8s.io/component-base/metrics/legacyregistry"
 )
 
+// The various stability levels used here are analogous to the ones used in the Kubernetes project. They are:
+// INTERNAL metrics have no stability guarantees, as such, labels may be arbitrarily added/removed and the metric may be deleted at any time.
+// ALPHA metrics have no stability guarantees, as such, labels may be arbitrarily added/removed, while metrics undergo a deprecation cycle before deletion.
+// BETA metrics have some stability guarantees, as such, labels may be added but not removed, while metrics undergo a deprecation cycle before deletion.
+// STABLE metrics are guaranteed not be mutated.
+
 // ReconcileAttempts is a counter that indicates the number of attempts to reconcile the operator configuration.
 var ReconcileAttempts = metrics.NewCounter(&metrics.CounterOpts{
 	Name:           "cluster_monitoring_operator_reconcile_attempts_total",
@@ -25,7 +31,7 @@ var ReconcileStatus = metrics.NewGauge(&metrics.GaugeOpts{
 var CollectionProfile = metrics.NewGaugeVec(&metrics.GaugeOpts{
 	Name:           "cluster_monitoring_operator_collection_profile",
 	Help:           "Information about collection profiles. Set to 1 for the configured collection profile, else 0.",
-	StabilityLevel: metrics.ALPHA,
+	StabilityLevel: metrics.BETA,
 }, []string{"profile"})
 
 var DeprecatedConfig = metrics.NewGaugeVec(&metrics.GaugeOpts{

pkg/operator/operator.go

@@ -168,6 +168,7 @@ type Operator struct {
 	remoteWrite               bool
 	userWorkloadEnabled       bool
 	metricsServerEnabled      bool
+	collectionProfilesEnabled bool
 
 	lastKnowInfrastructureConfig *InfrastructureConfig
 	lastKnowProxyConfig          *ProxyConfig
@@ -244,6 +245,7 @@ func New(
 		remoteWrite:               remoteWrite,
 		userWorkloadEnabled:       false,
 		metricsServerEnabled:      false,
+		collectionProfilesEnabled: false,
 		namespace:                 namespace,
 		namespaceUserWorkload:     namespaceUserWorkload,
 		client:                    c,
@@ -435,6 +437,7 @@ func New(
 			return nil, err
 		}
 		o.metricsServerEnabled = featureGates.Enabled(configv1.FeatureGateMetricsServer)
+		o.collectionProfilesEnabled = featureGates.Enabled(configv1.FeatureGateMetricsCollectionProfiles)
 	case <-time.After(1 * time.Minute):
 		return nil, fmt.Errorf("timed out waiting for FeatureGate detection")
 	}
@@ -924,7 +927,7 @@ func (o *Operator) loadUserWorkloadConfig(ctx context.Context) (*manifests.UserW
 	return uwc, nil
 }
 
-func (o *Operator) loadConfig(key string, tp bool) (*manifests.Config, error) {
+func (o *Operator) loadConfig(key string) (*manifests.Config, error) {
 	obj, found, err := o.cmapInf.GetStore().GetByKey(key)
 	if err != nil {
 		return nil, fmt.Errorf("an error occurred when retrieving the Cluster Monitoring ConfigMap: %w", err)
@@ -942,7 +945,7 @@ func (o *Operator) loadConfig(key string, tp bool) (*manifests.Config, error) {
 		return nil, errors.New("the Cluster Monitoring ConfigMap doesn't contain a 'config.yaml' key")
 	}
 
-	cParsed, err := manifests.NewConfigFromString(configContent, tp)
+	cParsed, err := manifests.NewConfigFromString(configContent, o.collectionProfilesEnabled)
 	if err != nil {
 		return nil, fmt.Errorf("the Cluster Monitoring ConfigMap could not be parsed: %w", err)
 	}
@@ -951,12 +954,7 @@ func (o *Operator) loadConfig(key string, tp bool) (*manifests.Config, error) {
 }
 
 func (o *Operator) Config(ctx context.Context, key string) (*manifests.Config, error) {
-	tp, err := o.client.TechPreviewEnabled(ctx)
-	if err != nil {
-		return nil, err
-	}
-
-	c, err := o.loadConfig(key, tp)
+	c, err := o.loadConfig(key)
 	if err != nil {
 		return nil, err
 	}