cronjob to cleanup resolved alarms (#474)

config/rbac/role.yaml

@@ -119,6 +119,18 @@ rules:
   - subjectaccessreviews
   verbs:
   - create
+- apiGroups:
+  - batch
+  resources:
+  - cronjobs
+  verbs:
+  - create
+  - delete
+  - get
+  - list
+  - patch
+  - update
+  - watch
 - apiGroups:
   - cluster.open-cluster-management.io
   resources:

internal/controllers/inventory_controller.go

@@ -73,6 +73,7 @@ import (
 //+kubebuilder:rbac:urls="/internal/v1/caas-alerts/alertmanager",verbs=create;post
 //+kubebuilder:rbac:urls="/o2ims-infrastructureCluster/v1/nodeClusterTypes",verbs=get;list
 //+kubebuilder:rbac:urls="/o2ims-infrastructureCluster/v1/nodeClusters",verbs=get;list
+//+kubebuilder:rbac:groups="batch",resources=cronjobs,verbs=get;list;watch;create;update;patch;delete
 
 // Reconciler reconciles a Inventory object
 type Reconciler struct {
@@ -1140,12 +1141,15 @@ func (t *reconcilerTask) createAlarmServerClusterRole(ctx context.Context) error
 				},
 				Resources: []string{
 					"secrets",
+					"configmaps",
 				},
 				Verbs: []string{
 					"get",
 					"list",
 					"watch",
 					"update",
+					"create",
+					"delete",
 				},
 			},
 			{
@@ -1161,6 +1165,36 @@ func (t *reconcilerTask) createAlarmServerClusterRole(ctx context.Context) error
 					"watch",
 				},
 			},
+			{
+				APIGroups: []string{
+					"apps",
+				},
+				Resources: []string{
+					"deployments",
+				},
+				Verbs: []string{
+					"get",
+					"list",
+					"watch",
+				},
+			},
+			{
+				APIGroups: []string{
+					"batch",
+				},
+				Resources: []string{
+					"cronjobs",
+				},
+				Verbs: []string{
+					"get",
+					"list",
+					"watch",
+					"create",
+					"delete",
+					"update",
+					"patch",
+				},
+			},
 			{
 				NonResourceURLs: []string{
 					"/o2ims-infrastructureCluster/v1/nodeClusterTypes",
@@ -1405,6 +1439,30 @@ func (t *reconcilerTask) deployServer(ctx context.Context, serverName string) (u
 		}...)
 	}
 
+	// Common evn for server deployments
+	envVars = append(envVars,
+		corev1.EnvVar{
+			Name: "POD_NAMESPACE",
+			ValueFrom: &corev1.EnvVarSource{
+				FieldRef: &corev1.ObjectFieldSelector{
+					FieldPath: "metadata.namespace",
+				},
+			},
+		},
+	)
+
+	// Server specific env var
+	if serverName == utils.InventoryAlarmServerName {
+		postgresImage := os.Getenv(utils.PostgresImageName)
+		if postgresImage == "" {
+			return "", fmt.Errorf("missing %s environment variable value", utils.PostgresImageName)
+		}
+		envVars = append(envVars, corev1.EnvVar{
+			Name:  utils.PostgresImageName,
+			Value: postgresImage,
+		})
+	}
+
 	// Build the deployment's spec.
 	deploymentSpec := appsv1.DeploymentSpec{
 		Replicas: k8sptr.To(int32(1)),

internal/service/alarms/api/server.go

@@ -9,6 +9,8 @@ import (
 	"sync"
 	"time"
 
+	"github.com/openshift-kni/oran-o2ims/internal/service/alarms/internal/serviceconfig"
+
 	"github.com/google/uuid"
 	"github.com/jackc/pgerrcode"
 	"github.com/jackc/pgx/v5/pgconn"
@@ -27,8 +29,8 @@ import (
 )
 
 const (
-	DefaultRetentionPeriod = 10
-	minRetentionPeriod     = 1
+	DefaultRetentionPeriod = 1 // Default retention of resolved alarms in days
+	minRetentionPeriod     = 1 // Minimum retention of resolved alarms in days
 )
 
 // AlarmsServerConfig defines the configuration attributes for the alarms server
@@ -51,6 +53,8 @@ type AlarmsServer struct {
 	NotificationProvider notifier.NotificationProvider
 	// Notifier to notify subscribers with new events
 	Notifier *notifier.Notifier
+	// ServiceConfig config needed to manage ServiceConfig
+	ServiceConfig serviceconfig.Config
 }
 
 // AlarmsServer implements StrictServerInterface. This ensures that we've conformed to the `StrictServerInterface` with a compile-time check
@@ -446,8 +450,12 @@ func (a *AlarmsServer) PatchAlarmServiceConfiguration(ctx context.Context, reque
 		return nil, fmt.Errorf("failed to patch Alarm Service Configuration: %w", err)
 	}
 
-	slog.Debug("Alarm Service Configuration patched", "retentionPeriod", patched.RetentionPeriod, "extensions", patched.Extensions)
+	// Update Cronjob
+	if err := a.ServiceConfig.EnsureCleanupCronJob(ctx, patched); err != nil {
+		return nil, fmt.Errorf("failed to start cleanup cronjob during AlarmServiceConfiguration patch: %w", err)
+	}
 
+	slog.Debug("Alarm Service Configuration patched", "retentionPeriod", patched.RetentionPeriod, "extensions", patched.Extensions)
 	return api.PatchAlarmServiceConfiguration200JSONResponse(models.ConvertServiceConfigurationToAPI(*patched)), nil
 }
 
@@ -486,8 +494,12 @@ func (a *AlarmsServer) UpdateAlarmServiceConfiguration(ctx context.Context, requ
 		return nil, fmt.Errorf("failed to update Alarm Service Configuration: %w", err)
 	}
 
-	slog.Debug("Alarm Service Configuration updated", "retentionPeriod", updated.RetentionPeriod, "extensions", updated.Extensions)
+	// Update Cronjob
+	if err := a.ServiceConfig.EnsureCleanupCronJob(ctx, updated); err != nil {
+		return nil, fmt.Errorf("failed to start cleanup cronjob during AlarmServiceConfiguration update: %w", err)
+	}
 
+	slog.Debug("Alarm Service Configuration updated", "retentionPeriod", updated.RetentionPeriod, "extensions", updated.Extensions)
 	return api.UpdateAlarmServiceConfiguration200JSONResponse(models.ConvertServiceConfigurationToAPI(*updated)), nil
 
 }