Make vLLM port number explicit in InferenceServerConfig

Signed-off-by: Jun Duan <jun.duan.phd@outlook.com>

Author: waltforme

api/fma/v1alpha1/inferenceserverconfig_types.go

@@ -33,7 +33,11 @@ type InferenceServerConfigSpec struct {
 
 // ModelServerConfig defines the configuration for a model server
 type ModelServerConfig struct {
-	// Options are the vLLM startup options
+	// Port is the port on which the vLLM server will listen
+	// Particularly, management of vLLM instances' sleep state is done through this port
+	Port int32 `json:"port"`
+
+	// Options are the vLLM startup options, excluding Port
 	// +optional
 	Options string `json:"options,omitempty"`
 

config/crd/fma.llm-d.ai_inferenceserverconfigs.yaml

@@ -66,8 +66,16 @@ spec:
                       type: string
                     type: object
                   options:
-                    description: Options are the vLLM startup options
+                    description: Options are the vLLM startup options, excluding Port
                     type: string
+                  port:
+                    description: |-
+                      Port is the port on which the vLLM server will listen
+                      Particularly, management of vLLM instances' sleep state is done through this port
+                    format: int32
+                    type: integer
+                required:
+                - port
                 type: object
             required:
             - launcherConfigName