[Cherry-pick][Telemetry] KubeRay version and CRD (#2024) (#2032)

Author: kevin85421

docs/development/release.md

@@ -65,6 +65,8 @@ Manual testing can be time-consuming, and to relieve the workload, we plan to ad
 
 * Merge a PR into the release branch updating Helm chart versions, Helm chart image tags, and kustomize manifest image tags. For `v0.5.0-rc0`, we did this in [PR #1001](https://github.com/ray-project/kuberay/pull/1001)
 
+* Merge a PR into the release branch to update the `KUBERAY_VERSION` in `constant.go` for telemetry purposes.
+
 * Release `rc0` images using the [release-image-build](https://github.com/ray-project/kuberay/actions/workflows/image-release.yaml) workflow on GitHub actions.
 You will be prompted for a commit reference and an image tag. The commit reference should be the SHA of the tip of the release branch. The image tag should be `vX.Y.Z-rc.0`.
 

ray-operator/controllers/ray/common/pod.go

@@ -629,6 +629,13 @@ func setContainerEnvVars(pod *corev1.Pod, rayNodeType rayv1.RayNodeType, rayStar
 		usageEnv := corev1.EnvVar{Name: utils.RAY_USAGE_STATS_KUBERAY_IN_USE, Value: "1"}
 		container.Env = append(container.Env, usageEnv)
 	}
+	if rayNodeType == rayv1.HeadNode {
+		extraTagsEnv := corev1.EnvVar{
+			Name:  utils.RAY_USAGE_STATS_EXTRA_TAGS,
+			Value: fmt.Sprintf("kuberay_version=%s;kuberay_crd=%s", utils.KUBERAY_VERSION, string(creatorCRDType)),
+		}
+		container.Env = append(container.Env, extraTagsEnv)
+	}
 	if !utils.EnvVarExists(utils.REDIS_PASSWORD, container.Env) {
 		// setting the REDIS_PASSWORD env var from the params
 		redisPasswordEnv := corev1.EnvVar{Name: utils.REDIS_PASSWORD}

ray-operator/controllers/ray/common/pod_test.go

@@ -316,6 +316,7 @@ func TestBuildPod(t *testing.T) {
 	checkContainerEnv(t, rayContainer, utils.RAY_CLUSTER_NAME, fmt.Sprintf("metadata.labels['%s']", utils.RayClusterLabelKey))
 	checkContainerEnv(t, rayContainer, utils.RAY_DASHBOARD_ENABLE_K8S_DISK_USAGE, "1")
 	checkContainerEnv(t, rayContainer, utils.RAY_NODE_TYPE_NAME, fmt.Sprintf("metadata.labels['%s']", utils.RayNodeGroupLabelKey))
+	checkContainerEnv(t, rayContainer, utils.RAY_USAGE_STATS_EXTRA_TAGS, fmt.Sprintf("kuberay_version=%s;kuberay_crd=%s", utils.KUBERAY_VERSION, utils.RayClusterCRD))
 	headRayStartCommandEnv := getEnvVar(rayContainer, utils.KUBERAY_GEN_RAY_START_CMD)
 	assert.True(t, strings.Contains(headRayStartCommandEnv.Value, "ray start"))
 

ray-operator/controllers/ray/utils/constant.go

@@ -102,6 +102,7 @@ const (
 	RAY_GCS_SERVER_REQUEST_TIMEOUT_SECONDS  = "RAY_gcs_server_request_timeout_seconds"
 	RAY_SERVE_KV_TIMEOUT_S                  = "RAY_SERVE_KV_TIMEOUT_S"
 	RAY_USAGE_STATS_KUBERAY_IN_USE          = "RAY_USAGE_STATS_KUBERAY_IN_USE"
+	RAY_USAGE_STATS_EXTRA_TAGS              = "RAY_USAGE_STATS_EXTRA_TAGS"
 	RAYCLUSTER_DEFAULT_REQUEUE_SECONDS_ENV  = "RAYCLUSTER_DEFAULT_REQUEUE_SECONDS_ENV"
 	RAYCLUSTER_DEFAULT_REQUEUE_SECONDS      = 300
 	KUBERAY_GEN_RAY_START_CMD               = "KUBERAY_GEN_RAY_START_CMD"
@@ -167,6 +168,9 @@ const (
 
 	// RayNodeHeadGroupLabelValue is the value for the RayNodeGroupLabelKey label on a head node
 	RayNodeHeadGroupLabelValue = "headgroup"
+
+	// Telemetry
+	KUBERAY_VERSION = "nightly"
 )
 
 type ServiceType string

ray-operator/controllers/ray/utils/util.go

@@ -38,7 +38,6 @@ const (
 	RayClusterCRD CRDType = "RayCluster"
 	RayJobCRD     CRDType = "RayJob"
 	RayServiceCRD CRDType = "RayService"
-	UnknownCRD    CRDType = "Unknown"
 )
 
 var crdMap = map[string]CRDType{
@@ -51,7 +50,7 @@ func GetCRDType(key string) CRDType {
 	if crdType, exists := crdMap[key]; exists {
 		return crdType
 	}
-	return UnknownCRD
+	return RayClusterCRD
 }
 
 // GetClusterDomainName returns cluster's domain name

ray-operator/main.go

@@ -37,11 +37,8 @@ import (
 )
 
 var (
-	_version_   = "0.2"
-	_buildTime_ = ""
-	_commitId_  = ""
-	scheme      = runtime.NewScheme()
-	setupLog    = ctrl.Log.WithName("setup")
+	scheme   = runtime.NewScheme()
+	setupLog = ctrl.Log.WithName("setup")
 	// TODO: include the kuberay version in the user-agent. E.g. ray-operator/v1.1.0
 	userAgent = "kuberay-operator"
 )
@@ -57,7 +54,6 @@ func init() {
 }
 
 func main() {
-	var version bool
 	var metricsAddr string
 	var enableLeaderElection bool
 	var leaderElectionNamespace string
@@ -70,7 +66,6 @@ func main() {
 	var configFile string
 
 	// TODO: remove flag-based config once Configuration API graduates to v1.
-	flag.BoolVar(&version, "version", false, "Show the version information.")
 	flag.StringVar(&metricsAddr, "metrics-addr", configapi.DefaultMetricsAddr, "The address the metric endpoint binds to.")
 	flag.StringVar(&probeAddr, "health-probe-bind-address", configapi.DefaultProbeAddr, "The address the probe endpoint binds to.")
 	flag.BoolVar(&enableLeaderElection, "enable-leader-election", configapi.DefaultEnableLeaderElection,
@@ -100,12 +95,6 @@ func main() {
 	}
 	opts.BindFlags(flag.CommandLine)
 	flag.Parse()
-	if version {
-		fmt.Printf("Version:\t%s\n", _version_)
-		fmt.Printf("Commit ID:\t%s\n", _commitId_)
-		fmt.Printf("Build time:\t%s\n", _buildTime_)
-		os.Exit(0)
-	}
 
 	var config configapi.Configuration
 	if configFile != "" {
@@ -161,7 +150,6 @@ func main() {
 		ctrl.SetLogger(k8szap.New(k8szap.UseFlagOptions(&opts)))
 	}
 
-	setupLog.Info("the operator", "version:", os.Getenv("OPERATOR_VERSION"))
 	if ray.ForcedClusterUpgrade {
 		setupLog.Info("Feature flag forced-cluster-upgrade is enabled.")
 	}