pkg/manifests: Expose retention size settings for Platform Prometheus

Introduce a new "retentionSize" field under the prometheusK8s key
When neither retention nor retentionSize are defined, a default of 15d for time retention applies.
When "retentionSize" is defined and "retention" isn't, only size retention applies.
When "retentionSize" isn't defined and "retention" is, only time retention applies
When both time and size retention fields are defined, both apply.

Signed-off-by: Jayapriya Pai <[email protected]>

Author: slashpai

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.
 
+## 4.11
+
+- [#1579](https://github.com/openshift/cluster-monitoring-operator/pull/1579) Expose retention size settings for Platform Prometheus
+
 ## 4.10
 
 - [#1509](https://github.com/openshift/cluster-monitoring-operator/pull/1509) add NLB usage metrics for network edge

pkg/manifests/config.go

@@ -170,6 +170,7 @@ type RemoteWriteSpec struct {
 type PrometheusK8sConfig struct {
 	LogLevel            string                               `json:"logLevel"`
 	Retention           string                               `json:"retention"`
+	RetentionSize       string                               `json:"retentionSize"`
 	NodeSelector        map[string]string                    `json:"nodeSelector"`
 	Tolerations         []v1.Toleration                      `json:"tolerations"`
 	Resources           *v1.ResourceRequirements             `json:"resources"`

pkg/manifests/manifests.go

@@ -1411,6 +1411,10 @@ func (f *Factory) PrometheusK8sTrustedCABundle() (*v1.ConfigMap, error) {
 	return cm, nil
 }
 
+func isUnsetRetentionConfig(retentionSize string) bool {
+	return retentionSize == "" || retentionSize == "0"
+}
+
 func (f *Factory) PrometheusK8s(grpcTLS *v1.Secret, trustedCABundleCM *v1.ConfigMap) (*monv1.Prometheus, error) {
 	p, err := f.NewPrometheus(f.assets.MustNewAssetReader(PrometheusK8s))
 	if err != nil {
@@ -1425,6 +1429,10 @@ func (f *Factory) PrometheusK8s(grpcTLS *v1.Secret, trustedCABundleCM *v1.Config
 		p.Spec.Retention = f.config.ClusterMonitoringConfiguration.PrometheusK8sConfig.Retention
 	}
 
+	if !isUnsetRetentionConfig(f.config.ClusterMonitoringConfiguration.PrometheusK8sConfig.RetentionSize) {
+		p.Spec.RetentionSize = f.config.ClusterMonitoringConfiguration.PrometheusK8sConfig.RetentionSize
+	}
+
 	p.Spec.Image = &f.config.Images.Prometheus
 
 	if f.consoleConfig != nil {

pkg/manifests/manifests_test.go

@@ -1034,6 +1034,7 @@ func TestPrometheusK8sRemoteWrite(t *testing.T) {
 func TestPrometheusK8sConfiguration(t *testing.T) {
 	c, err := NewConfigFromString(`prometheusK8s:
   retention: 25h
+  retentionSize: 10Gi
   nodeSelector:
     type: master
   tolerations:
@@ -1083,6 +1084,10 @@ ingress:
 		t.Fatal("Retention is not configured correctly")
 	}
 
+	if p.Spec.RetentionSize != "10Gi" {
+		t.Fatal("RetentionSize is not configured correctly")
+	}
+
 	if *p.Spec.Image != "docker.io/openshift/origin-prometheus:latest" {
 		t.Fatal("Prometheus image is not configured correctly")
 	}

test/e2e/config_test.go

@@ -151,6 +151,7 @@ func TestClusterMonitorPrometheusK8Config(t *testing.T) {
 	data := fmt.Sprintf(`prometheusK8s:
   logLevel: debug
   retention: 10h
+  retentionSize: 15GB
   queryLogFile: /tmp/test.log
   tolerations:
     - operator: "Exists"
@@ -189,6 +190,7 @@ func TestClusterMonitorPrometheusK8Config(t *testing.T) {
 					expectMatchingRequests(podName, containerName, mem, cpu),
 					expectContainerArg("--log.level=debug", containerName),
 					expectContainerArg("--storage.tsdb.retention.time=10h", containerName),
+					expectContainerArg("--storage.tsdb.retention.size=15GB", containerName),
 				},
 			),
 		},