Merge pull request #584 from ingvagabund/release-4.19

Backporting 4.20/master fixes to 4.19/KDO 5.2.0 release

Author: ingvagabund

README.md

@@ -189,7 +189,7 @@ The profile exposes the following customization:
 - `devActualUtilizationProfile`: Enable load-aware descheduling.
 - `devDeviationThresholds`: Have the thresholds be based on the average utilization.
 
-By default, this profile will enable load-aware descheduling based on the `PrometheusCPUCombinedProfile` Prometheus query.
+By default, this profile will enable load-aware descheduling based on the `PrometheusCPUCombined` Prometheus query.
 By default, the thresholds will be dynamic (based on the distance from the average utilization) and asymmetric (all the nodes below the average will be considered as underutilized to help rebalancing overutilized outliers) tolerating low deviations (10%).
 
 ### EvictPodsWithPVC
@@ -227,7 +227,7 @@ The operator provides the following profiles:
 - `PrometheusCPUPSIPressureByUtilization`: `avg by (instance) ( rate(node_pressure_cpu_waiting_seconds_total[1m])) and (1 - avg by (instance) (rate(node_cpu_seconds_total{mode="idle"}[1m]))) > 0.7 or avg by (instance) ( rate(node_pressure_cpu_waiting_seconds_total[1m])) * 0` (`node_pressure_cpu_waiting_seconds_total` is reported in OpenShift only for nodes configured with psi=1 kernel argument; the query is filtering out PSI pressure on nodes with average CPU utilization < 0.7 to filter out false positives pressure spikes due to self imposed CPU throttling)
 - `PrometheusMemoryPSIPressure`: `rate(node_pressure_memory_waiting_seconds_total[1m])` (`node_pressure_memory_waiting_seconds_total` is reported in OpenShift only for nodes configured with psi=1 kernel argument)
 - `PrometheusIOPSIPressure`: `rate(node_pressure_io_waiting_seconds_total[1m])` (`node_pressure_memory_waiting_seconds_total` is reported in OpenShift only for nodes configured with psi=1 kernel argument)
-- `PrometheusCPUCombinedProfile`: `descheduler:combined_utilization_and_pressure:avg1m` (`descheduler:combined_utilization_and_pressure:avg1m` uses a combination of CPU utilization and CPU PSI pressure based on a recording rule; CPU PSI pressure is reported in OpenShift only for nodes configured with psi=1 kernel argument)
+- `PrometheusCPUCombined`: `descheduler:combined_utilization_and_pressure:avg1m` (`descheduler:combined_utilization_and_pressure:avg1m` uses a combination of CPU utilization and CPU PSI pressure based on a recording rule; CPU PSI pressure is reported in OpenShift only for nodes configured with psi=1 kernel argument)
 
 ```yaml
 apiVersion: operator.openshift.io/v1

pkg/operator/target_config_reconciler.go

@@ -155,8 +155,8 @@ func (c TargetConfigReconciler) sync() error {
 		return err
 	}
 
-	if descheduler.Spec.DeschedulingIntervalSeconds == nil {
-		return fmt.Errorf("descheduler should have an interval set")
+	if descheduler.Spec.DeschedulingIntervalSeconds == nil || *descheduler.Spec.DeschedulingIntervalSeconds <= 0 {
+		return fmt.Errorf("descheduler should have an interval set and it should be greater than 0")
 	}
 
 	specAnnotations := map[string]string{
@@ -1301,19 +1301,19 @@ func (c *TargetConfigReconciler) manageConfigMap(descheduler *deschedulerv1.Kube
 		}
 	}
 
-	if descheduler.Spec.ProfileCustomizations != nil && descheduler.Spec.ProfileCustomizations.DevActualUtilizationProfile != "" {
+	if c.isPrometheusAsMetricsProviderForProfiles(descheduler) {
 		// detect the prometheus server url
 		route, err := c.routeRouteLister.Routes("openshift-monitoring").Get("prometheus-k8s")
 		if err != nil {
-			return nil, false, fmt.Errorf("unable to get openshift-monitoring/prometheus-k8s route: %v", err)
+			return nil, true, fmt.Errorf("unable to get openshift-monitoring/prometheus-k8s route: %v", err)
 		}
 		if len(route.Status.Ingress) == 0 {
-			return nil, false, fmt.Errorf("No ingress found in openshift-monitoring/prometheus-k8s route")
+			return nil, true, fmt.Errorf("No ingress found in openshift-monitoring/prometheus-k8s route")
 		}
 		if route.Status.Ingress[0].Host == "" {
-			return nil, false, fmt.Errorf("Host for status.ingress[0] in openshift-monitoring/prometheus-k8s route is empty")
+			return nil, true, fmt.Errorf("Host for status.ingress[0] in openshift-monitoring/prometheus-k8s route is empty")
 		}
-		err = c.checkNamepsaceMonitoringLabel()
+		err = c.checkNamespaceMonitoringLabel()
 		if err != nil {
 			return nil, false, err
 		}
@@ -1640,7 +1640,7 @@ func (c *TargetConfigReconciler) eventHandler() cache.ResourceEventHandler {
 	}
 }
 
-func (c *TargetConfigReconciler) checkNamepsaceMonitoringLabel() error {
+func (c *TargetConfigReconciler) checkNamespaceMonitoringLabel() error {
 	operatorNamespace, err := c.namespaceLister.Get(operatorclient.OperatorNamespace)
 	if err != nil {
 		klog.ErrorS(err, "error fetching operator namespace")
@@ -1714,3 +1714,14 @@ func (c *TargetConfigReconciler) isSoftTainterNeeded(descheduler *deschedulerv1.
 	}
 	return leftoverSoftTaints, nil
 }
+
+// isPrometheusAsMetricsProviderForProfiles returns true when at least a profile that by default relies on PrometheusMetrics is in use
+// or the user is explicitly configuring DevActualUtilizationProfile profile customization
+func (c *TargetConfigReconciler) isPrometheusAsMetricsProviderForProfiles(descheduler *deschedulerv1.KubeDescheduler) bool {
+	if descheduler != nil &&
+		(slices.Contains(descheduler.Spec.Profiles, deschedulerv1.RelieveAndMigrate) ||
+			(descheduler.Spec.ProfileCustomizations != nil && descheduler.Spec.ProfileCustomizations.DevActualUtilizationProfile != "")) {
+		return true
+	}
+	return false
+}