[Feature][RayJob] Remove wget dependency in Sidecar GCS wait

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

@@ -125,14 +125,28 @@ func BuildJobSubmitCommand(rayJobInstance *rayv1.RayJob, submissionMode rayv1.Jo
 	jobFollowCommand := []string{"ray", "job", "logs", "--address", address, "--follow", jobId}
 
 	if submissionMode == rayv1.SidecarMode {
+		rayVersion := ""
+		if rayJobInstance.Spec.RayClusterSpec != nil {
+			rayVersion = rayJobInstance.Spec.RayClusterSpec.RayVersion
+		}
+
 		// Wait until Ray Dashboard GCS is healthy before proceeding.
-		// Use the same Ray Dashboard GCS health check command as the readiness probe
-		rayDashboardGCSHealthCommand := fmt.Sprintf(
-			utils.BaseWgetHealthCommand,
-			utils.DefaultReadinessProbeFailureThreshold,
-			port,
-			utils.RayDashboardGCSHealthPath,
-		)
+		var rayDashboardGCSHealthCommand string
+		if supportsUnifiedHealthCheck(rayVersion) {
+			rayDashboardGCSHealthCommand = fmt.Sprintf(
+				utils.BasePythonHealthCommand,
+				port,
+				utils.RayDashboardGCSHealthPath,
+				utils.DefaultReadinessProbeFailureThreshold,
+			)
+		} else {
+			rayDashboardGCSHealthCommand = fmt.Sprintf(
+				utils.BaseWgetHealthCommand,
+				utils.DefaultReadinessProbeFailureThreshold,
+				port,
+				utils.RayDashboardGCSHealthPath,
+			)
+		}
 
 		waitLoop := []string{
 			"until", rayDashboardGCSHealthCommand, ">/dev/null", "2>&1", ";",

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

@@ -138,6 +138,73 @@ func TestBuildJobSubmitCommandWithSidecarMode(t *testing.T) {
 	assert.Equal(t, expected, command)
 }
 
+func TestBuildJobSubmitCommandWithSidecarModeVersionSwitch(t *testing.T) {
+	tests := []struct {
+		name            string
+		rayVersion      string
+		usePythonHealth bool
+	}{
+		{
+			name:            "uses python health command for ray >= 2.53",
+			rayVersion:      "2.53.0",
+			usePythonHealth: true,
+		},
+		{
+			name:            "uses wget health command for ray < 2.53",
+			rayVersion:      "2.52.0",
+			usePythonHealth: false,
+		},
+		{
+			name:            "uses wget health command when rayVersion is invalid",
+			rayVersion:      "invalid-version",
+			usePythonHealth: false,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			testRayJob := rayJobTemplate()
+			testRayJob.Spec.RayClusterSpec.RayVersion = tt.rayVersion
+			if tt.rayVersion == "invalid-version" {
+				// Avoid metadata-json version parsing failure; this test only checks health command selection.
+				testRayJob.Spec.Metadata = nil
+			}
+			testRayJob.Spec.RayClusterSpec.HeadGroupSpec.Template.Spec.Containers = []corev1.Container{
+				{
+					Ports: []corev1.ContainerPort{
+						{
+							Name:          utils.DashboardPortName,
+							ContainerPort: utils.DefaultDashboardPort,
+						},
+					},
+				},
+			}
+
+			command, err := BuildJobSubmitCommand(testRayJob, rayv1.SidecarMode)
+			require.NoError(t, err)
+			require.GreaterOrEqual(t, len(command), 2)
+			assert.Equal(t, "until", command[0])
+
+			expected := fmt.Sprintf(
+				utils.BaseWgetHealthCommand,
+				utils.DefaultReadinessProbeFailureThreshold,
+				utils.DefaultDashboardPort,
+				utils.RayDashboardGCSHealthPath,
+			)
+			if tt.usePythonHealth {
+				expected = fmt.Sprintf(
+					utils.BasePythonHealthCommand,
+					utils.DefaultDashboardPort,
+					utils.RayDashboardGCSHealthPath,
+					utils.DefaultReadinessProbeFailureThreshold,
+				)
+			}
+
+			assert.Equal(t, expected, command[1])
+		})
+	}
+}
+
 func TestBuildJobSubmitCommandWithK8sJobModeAndYAML(t *testing.T) {
 	rayJobWithYAML := &rayv1.RayJob{
 		Spec: rayv1.RayJobSpec{

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

@@ -244,8 +244,13 @@ const (
 	RayAgentRayletHealthPath  = "api/local_raylet_healthz"
 	RayDashboardGCSHealthPath = "api/gcs_healthz"
 	RayServeProxyHealthPath   = "-/healthz"
-	BaseWgetHealthCommand     = "wget --tries 1 -T %d -q -O- http://localhost:%d/%s | grep success"
-	RayNodeHealthPath         = "/api/healthz"
+	// BaseWgetHealthCommand checks a single health URL; args: timeout_sec, port, path (no leading slash).
+	// This is used for Ray versions that rely on exec probes and assume common CLI tools exist in the image.
+	BaseWgetHealthCommand = `wget -q -T %d -O- http://localhost:%d/%s | grep -q success`
+	// BasePythonHealthCommand checks a single health URL; args: port, path (no leading slash), timeout_sec.
+	// This is used when wget is not available (e.g. slim Ray images).
+	BasePythonHealthCommand = `python3 -c "import urllib.request; r=urllib.request.urlopen('http://localhost:%d/%s', timeout=%d); exit(0 if b'success' in r.read() else 1)"`
+	RayNodeHealthPath       = "/api/healthz"
 
 	// Finalizers for RayJob
 	RayJobStopJobFinalizer = "ray.io/rayjob-finalizer"