alert-mgmt-11: platform rule update (Drop/Restore)

internal/managementrouter/health_get_test.go

@@ -116,6 +116,10 @@ func (s *healthStubManagementClient) ListRules(ctx context.Context, prOptions ma
 	return nil, nil
 }
 
+func (s *healthStubManagementClient) GetRuleById(ctx context.Context, alertRuleId string) (monitoringv1.Rule, error) {
+	return monitoringv1.Rule{}, nil
+}
+
 func (s *healthStubManagementClient) CreateUserDefinedAlertRule(ctx context.Context, alertRule monitoringv1.Rule, prOptions management.PrometheusRuleOptions) (string, error) {
 	return "", nil
 }
@@ -128,6 +132,18 @@ func (s *healthStubManagementClient) DeleteUserDefinedAlertRuleById(ctx context.
 	return nil
 }
 
+func (s *healthStubManagementClient) UpdatePlatformAlertRule(ctx context.Context, alertRuleId string, alertRule monitoringv1.Rule) error {
+	return nil
+}
+
+func (s *healthStubManagementClient) DropPlatformAlertRule(ctx context.Context, alertRuleId string) error {
+	return nil
+}
+
+func (s *healthStubManagementClient) RestorePlatformAlertRule(ctx context.Context, alertRuleId string) error {
+	return nil
+}
+
 func (s *healthStubManagementClient) GetAlerts(ctx context.Context, req k8s.GetAlertsRequest) ([]k8s.PrometheusAlert, error) {
 	return nil, nil
 }

internal/managementrouter/rules_get_test.go

@@ -145,6 +145,10 @@ func (s *stubManagementClient) ListRules(ctx context.Context, prOptions manageme
 	return nil, nil
 }
 
+func (s *stubManagementClient) GetRuleById(ctx context.Context, alertRuleId string) (monitoringv1.Rule, error) {
+	return monitoringv1.Rule{}, nil
+}
+
 func (s *stubManagementClient) CreateUserDefinedAlertRule(ctx context.Context, alertRule monitoringv1.Rule, prOptions management.PrometheusRuleOptions) (string, error) {
 	return "", nil
 }
@@ -157,6 +161,18 @@ func (s *stubManagementClient) DeleteUserDefinedAlertRuleById(ctx context.Contex
 	return nil
 }
 
+func (s *stubManagementClient) UpdatePlatformAlertRule(ctx context.Context, alertRuleId string, alertRule monitoringv1.Rule) error {
+	return nil
+}
+
+func (s *stubManagementClient) DropPlatformAlertRule(ctx context.Context, alertRuleId string) error {
+	return nil
+}
+
+func (s *stubManagementClient) RestorePlatformAlertRule(ctx context.Context, alertRuleId string) error {
+	return nil
+}
+
 func (s *stubManagementClient) GetAlerts(ctx context.Context, req k8s.GetAlertsRequest) ([]k8s.PrometheusAlert, error) {
 	return nil, nil
 }

pkg/management/delete_user_defined_alert_rule_by_id.go

@@ -63,7 +63,7 @@ func (c *client) deletePlatformAlertRuleById(ctx context.Context, relabeled moni
 		return fmt.Errorf("failed to get AlertingRule %s: %w", arName, err)
 	}
 	if !found || ar == nil {
-		return &NotFoundError{Resource: "AlertingRule", Id: arName}
+		return c.deleteUserAlertRuleById(ctx, namespace, name, alertRuleId)
 	}
 	// Common preconditions for platform delete
 	if err := validatePlatformDeletePreconditions(ar); err != nil {
@@ -84,9 +84,20 @@ func (c *client) deletePlatformAlertRuleById(ctx context.Context, relabeled moni
 			AdditionalInfo: fmt.Sprintf("alert %q not found in AlertingRule %s", originalRule.Alert, arName),
 		}
 	}
-	ar.Spec.Groups = newGroups
-	if err := c.k8sClient.AlertingRules().Update(ctx, *ar); err != nil {
-		return fmt.Errorf("failed to update AlertingRule %s: %w", ar.Name, err)
+
+	if len(newGroups) == 0 {
+		if err := c.k8sClient.AlertingRules().Delete(ctx, ar.Name); err != nil {
+			return fmt.Errorf("failed to delete AlertingRule %s: %w", ar.Name, err)
+		}
+	} else {
+		ar.Spec.Groups = newGroups
+		if err := c.k8sClient.AlertingRules().Update(ctx, *ar); err != nil {
+			return fmt.Errorf("failed to update AlertingRule %s: %w", ar.Name, err)
+		}
+	}
+
+	if err := c.deleteAssociatedARC(ctx, k8s.ClusterMonitoringNamespace, name, alertRuleId); err != nil {
+		return fmt.Errorf("failed to clean up ARC for platform rule %s: %w", alertRuleId, err)
 	}
 	return nil
 }
@@ -100,6 +111,9 @@ func (c *client) deleteUserAlertRuleById(ctx context.Context, namespace, name, a
 	if !found {
 		return &NotFoundError{Resource: "PrometheusRule", Id: fmt.Sprintf("%s/%s", namespace, name)}
 	}
+	if gitOpsManaged, _ := k8s.IsExternallyManagedObject(pr); gitOpsManaged {
+		return notAllowedGitOpsRemove()
+	}
 
 	updated := false
 	var newGroups []monitoringv1.RuleGroup
@@ -120,14 +134,39 @@ func (c *client) deleteUserAlertRuleById(ctx context.Context, namespace, name, a
 		if err := c.k8sClient.PrometheusRules().Delete(ctx, pr.Namespace, pr.Name); err != nil {
 			return fmt.Errorf("failed to delete PrometheusRule %s/%s: %w", pr.Namespace, pr.Name, err)
 		}
-		return nil
+		return c.cleanupARCForDeletedRule(ctx, namespace, name, alertRuleId)
 	}
 
 	pr.Spec.Groups = newGroups
 	if err := c.k8sClient.PrometheusRules().Update(ctx, *pr); err != nil {
 		return fmt.Errorf("failed to update PrometheusRule %s/%s: %w", pr.Namespace, pr.Name, err)
 	}
-	return nil
+	return c.cleanupARCForDeletedRule(ctx, namespace, name, alertRuleId)
+}
+
+// cleanupARCForDeletedRule attempts to remove any associated ARC after a rule is deleted.
+// It determines the ARC namespace from the rule's namespace and silently skips if no ARC exists.
+func (c *client) cleanupARCForDeletedRule(ctx context.Context, namespace, name, alertRuleId string) error {
+	nn := types.NamespacedName{Namespace: namespace, Name: name}
+	arcNamespace, err := c.arcNamespaceForRule(nn)
+	if err != nil {
+		return nil
+	}
+	return c.deleteAssociatedARC(ctx, arcNamespace, name, alertRuleId)
+}
+
+// deleteAssociatedARC removes the AlertRelabelConfig associated with an alert rule, if it exists.
+// This is best-effort: if the ARC does not exist or is GitOps-managed, it is silently skipped.
+func (c *client) deleteAssociatedARC(ctx context.Context, namespace, prName, alertRuleId string) error {
+	arcName := k8s.GetAlertRelabelConfigName(prName, alertRuleId)
+	arc, found, err := c.k8sClient.AlertRelabelConfigs().Get(ctx, namespace, arcName)
+	if err != nil || !found {
+		return nil
+	}
+	if gitOpsManaged, _ := k8s.IsExternallyManagedObject(arc); gitOpsManaged {
+		return nil
+	}
+	return c.k8sClient.AlertRelabelConfigs().Delete(ctx, namespace, arcName)
 }
 
 // removeAlertFromAlertingRuleGroups removes all instances of an alert by alert name across groups.