feat(helm): option to set the odigos-gateway collector name (#4239) (#4242)

Fixes: CORE-624

This PR adds a new option to helm to specify the name of the
`odigos-gateway` deployment (which will also show up in pods name)
Reasons is that some users prefer to have this name with the name of the
team, so if there are any issues, the pod name is enough to know which
team to ping.

feature, dependency update, or breaking change??

```release-note
new helm values option - set a custom name for the gateway collector deployment
```

Author: blumamir

api/config/crd/bases/odigos.io_collectorsgroups.yaml

@@ -52,6 +52,11 @@ spec:
                   This can be used to resolve conflicting ports when a collector is using the host network.
                 format: int32
                 type: integer
+              deploymentName:
+                description: |-
+                  Deployment name for the collectors group deployment.
+                  Only relevant for cluster gateway collector.
+                type: string
               enableDataCompression:
                 description: |-
                   Deprecated - use OtlpExporterConfiguration instead.

api/generated/odigos/applyconfiguration/odigos/v1alpha1/collectorsgroupspec.go

@@ -11,9 +11,11 @@ const (
 	OdigosClusterCollectorCollectorGroupName = OdigosClusterCollectorDeploymentName
 	OdigosClusterCollectorConfigMapKey       = "collector-conf"
 
-	OdigosClusterCollectorServiceAccountName = "odigos-gateway"
-	OdigosClusterCollectorRoleName           = "odigos-gateway"
-	OdigosClusterCollectorRoleBindingName    = "odigos-gateway"
+	OdigosClusterCollectorServiceAccountName = OdigosClusterCollectorDeploymentName
+	OdigosClusterCollectorRoleName           = OdigosClusterCollectorDeploymentName
+	OdigosClusterCollectorRoleBindingName    = OdigosClusterCollectorDeploymentName
+
+	OdigosClusterCollectorHpaName = OdigosClusterCollectorDeploymentName
 
 	OdigosClusterCollectorContainerName = "gateway"
 

api/k8sconsts/clustercollector.go

@@ -198,6 +198,10 @@ type CollectorsGroupSpec struct {
 	// This is a hard requirement: the pod will be scheduled ONLY on nodes that match all labels.
 	// If no matching nodes exist, the pod will remain Pending.
 	NodeSelector *map[string]string `json:"nodeSelector,omitempty"`
+
+	// Deployment name for the collectors group deployment.
+	// Only relevant for cluster gateway collector.
+	DeploymentName string `json:"deploymentName,omitempty"`
 }
 
 // CollectorsGroupStatus defines the observed state of Collector

api/odigos/v1alpha1/collectorsgroup_types.go

@@ -11,8 +11,8 @@ import (
 
 type CollectorsGroupReconciler struct {
 	client.Client
-	Scheme           *runtime.Scheme
-	OdigosVersion    string
+	Scheme        *runtime.Scheme
+	OdigosVersion string
 }
 
 func (r *CollectorsGroupReconciler) Reconcile(ctx context.Context, req ctrl.Request) (ctrl.Result, error) {

autoscaler/controllers/clustercollector/collectorsgroup_controller.go

@@ -83,6 +83,15 @@ func addSelfTelemetryPipeline(c *config.Config, ownTelemetryPort int32, destinat
 			},
 		},
 	}
+	c.Processors["resource/odigos-collector-role"] = config.GenericMap{
+		"attributes": []config.GenericMap{
+			{
+				"key":    "odigos.collector.role",
+				"value":  string(k8sconsts.CollectorsRoleClusterGateway),
+				"action": "upsert",
+			},
+		},
+	}
 	// odigostrafficmetrics processor should be the last processor in the pipeline
 	// as it helps to calculate the size of the data being exported.
 	// In case of performance impact caused by this processor, we should modify this config to reduce the sampling ratio.
@@ -98,7 +107,7 @@ func addSelfTelemetryPipeline(c *config.Config, ownTelemetryPort int32, destinat
 	}
 	c.Service.Pipelines["metrics/otelcol"] = config.Pipeline{
 		Receivers:  []string{"prometheus/self-metrics"},
-		Processors: []string{"resource/pod-name"},
+		Processors: []string{"resource/pod-name", "resource/odigos-collector-role"},
 		Exporters:  []string{"otlp/odigos-own-telemetry-ui"},
 	}
 

autoscaler/controllers/clustercollector/configmap.go

@@ -112,9 +112,10 @@ func TestAddSelfTelemetryPipeline(t *testing.T) {
 			}
 			assert.NotEmpty(t, c.Receivers["prometheus/self-metrics"])
 			assert.NotEmpty(t, c.Processors["resource/pod-name"])
+			assert.NotEmpty(t, c.Processors["resource/odigos-collector-role"])
 			assert.NotEmpty(t, c.Service.Pipelines["metrics/otelcol"])
 			assert.Equal(t, []string{"prometheus/self-metrics"}, c.Service.Pipelines["metrics/otelcol"].Receivers)
-			assert.Equal(t, []string{"resource/pod-name"}, c.Service.Pipelines["metrics/otelcol"].Processors)
+			assert.Equal(t, []string{"resource/pod-name", "resource/odigos-collector-role"}, c.Service.Pipelines["metrics/otelcol"].Processors)
 			assert.Equal(t, []string{"otlp/odigos-own-telemetry-ui"}, c.Service.Pipelines["metrics/otelcol"].Exporters)
 			pullExporter := c.Service.Telemetry.Metrics.Readers[0]["pull"].(config.GenericMap)["exporter"].(config.GenericMap)
 			port := pullExporter["prometheus"].(config.GenericMap)["port"]

autoscaler/controllers/clustercollector/configmap_test.go

@@ -59,11 +59,16 @@ func syncDeployment(enabledDests *odigosv1.DestinationList, gateway *odigosv1.Co
 	}
 
 	existingDeployment := &appsv1.Deployment{}
-	getError := c.Get(ctx, client.ObjectKey{Name: gateway.Name, Namespace: gateway.Namespace}, existingDeployment)
+	getError := c.Get(ctx, client.ObjectKey{Name: desiredDeployment.Name, Namespace: desiredDeployment.Namespace}, existingDeployment)
 	if getError != nil && !apierrors.IsNotFound(getError) {
 		return nil, errors.Join(getError, errors.New("failed to get gateway deployment"))
 	}
 
+	err = deleteOldDeployments(ctx, c, gateway.Namespace, desiredDeployment.Name)
+	if err != nil {
+		return nil, errors.Join(err, errors.New("failed to delete old deployments"))
+	}
+
 	if apierrors.IsNotFound(getError) {
 		logger.V(0).Info("Creating new gateway deployment")
 		err := c.Create(ctx, desiredDeployment)
@@ -81,6 +86,32 @@ func syncDeployment(enabledDests *odigosv1.DestinationList, gateway *odigosv1.Co
 	}
 }
 
+// users can set the deploymentName of the gateway collector to a custom value.
+// if that happens, the old deployments stays around, so this function takes care of deleting them.
+func deleteOldDeployments(ctx context.Context, c client.Client, namespace string, deploymentName string) error {
+	var deployments appsv1.DeploymentList
+	err := c.List(ctx, &deployments, client.InNamespace(namespace), client.MatchingLabels(ClusterCollectorGateway))
+	if err != nil {
+		return err
+	}
+
+	if len(deployments.Items) == 1 && deployments.Items[0].Name == deploymentName {
+		return nil
+	}
+
+	logger := log.FromContext(ctx)
+	for _, deployment := range deployments.Items {
+		if deployment.Name != deploymentName {
+			logger.V(0).Info("Deleting old gateway deployment pre odigos cluster collector deployment rename", "deployment", deployment.Name)
+			err := c.Delete(ctx, &deployment)
+			if err != nil {
+				return err
+			}
+		}
+	}
+	return nil
+}
+
 func patchDeployment(existing *appsv1.Deployment, desired *appsv1.Deployment, ctx context.Context, c client.Client) (*appsv1.Deployment, error) {
 	logger := log.FromContext(ctx)
 	res, err := controllerutil.CreateOrPatch(ctx, c, existing, func() error {
@@ -132,9 +163,11 @@ func getDesiredDeployment(ctx context.Context, c client.Client, enabledDests *od
 		})
 	}
 
+	deploymentName := commonconfig.GetDeploymentName(gateway)
+
 	desiredDeployment := &appsv1.Deployment{
 		ObjectMeta: v1.ObjectMeta{
-			Name:      k8sconsts.OdigosClusterCollectorDeploymentName,
+			Name:      deploymentName,
 			Namespace: gateway.Namespace,
 			Labels:    ClusterCollectorGateway,
 		},
@@ -152,7 +185,7 @@ func getDesiredDeployment(ctx context.Context, c client.Client, enabledDests *od
 				},
 				Spec: corev1.PodSpec{
 					NodeSelector:       *nodeSelector,
-					ServiceAccountName: k8sconsts.OdigosClusterCollectorDeploymentName,
+					ServiceAccountName: k8sconsts.OdigosClusterCollectorServiceAccountName,
 					SecurityContext: &corev1.PodSecurityContext{
 						RunAsNonRoot: boolPtr(true),
 						RunAsUser:    int64Ptr(65534), // nobody user

autoscaler/controllers/clustercollector/deployment.go

@@ -11,8 +11,8 @@ import (
 
 type DestinationReconciler struct {
 	client.Client
-	Scheme           *runtime.Scheme
-	OdigosVersion    string
+	Scheme        *runtime.Scheme
+	OdigosVersion string
 }
 
 func (r *DestinationReconciler) Reconcile(ctx context.Context, req ctrl.Request) (ctrl.Result, error) {

autoscaler/controllers/clustercollector/destination_controller.go

@@ -30,7 +30,7 @@ func (r *ClusterCollectorDeploymentReconciler) Reconcile(ctx context.Context, re
 	var gatewayCollectorGroup odigosv1.CollectorsGroup
 	if err := r.Get(ctx, types.NamespacedName{
 		Namespace: dep.Namespace,
-		Name:      k8sconsts.OdigosClusterCollectorDeploymentName,
+		Name:      k8sconsts.OdigosClusterCollectorCollectorGroupName,
 	}, &gatewayCollectorGroup); err != nil {
 		return ctrl.Result{}, client.IgnoreNotFound(err)
 	}