Add SchedulingConfig for Enhanced GPU and Pod Affinity Scheduling

k8s/apis/k8s/v1alpha1/vineyardd_types.go

@@ -101,6 +101,19 @@ type VolumeConfig struct {
 	MountPath string `json:"mountPath,omitempty"`
 }
 
+// SchedulingConfig holds all configurations about scheduling
+type SchedulingConfig struct {
+	// GPUType is a request to schedule this pod onto a specific GPU node. If it is non-empty,
+	// the scheduler simply schedules this pod onto that node, assuming that it fits resource
+	// requirements.
+	// +optional
+	GPUType string `json:"gpuType,omitempty"`
+	// If specified, the pod's scheduling constraints
+	// value should be the model name, model.aibrix.ai/name: deepseek-coder-7b-instruct
+	// +optional
+	AffinityWorkload string `json:"affinityWorkload,omitempty"`
+}
+
 // VineyardConfig holds all configuration about vineyard container
 type VineyardConfig struct {
 	// represent the vineyardd's image
@@ -228,6 +241,11 @@ type VineyarddSpec struct {
 	// +kubebuilder:default:={pvcName: "", mountPath: ""}
 	Volume VolumeConfig `json:"volume,omitempty"`
 
+	// the configuration of scheduling
+	// +kubebuilder:validation:Optional
+	// +kubebuilder:default:={}
+	Scheduler SchedulingConfig `json:"scheduler,omitempty"`
+
 	// SecurityContext holds the security context settings for the vineyardd container.
 	// +kubebuilder:validation:Optional
 	// +kubebuilder:default:={}

k8s/cmd/commands/manager/manager.go

@@ -297,7 +297,7 @@ func startScheduler(mgr manager.Manager, schedulerConfigFile string) {
 }
 
 // checkKubernetesVersionForScheduler checks the kubernetes version for the scheduler
-// the scheduler only supports kubernetes version 1.19-1.24
+// the scheduler only supports kubernetes version 1.19-1.30
 func checkKubernetesVersionForScheduler() (bool, error) {
 	config := util.GetKubernetesConfig()
 
@@ -321,8 +321,8 @@ func checkKubernetesVersionForScheduler() (bool, error) {
 		return false, err
 	}
 
-	if major != 1 || minor < 19 || minor > 24 {
-		return false, fmt.Errorf("the scheduler only supports kubernetes version 1.19-1.24")
+	if major != 1 || minor < 19 || minor > 30 {
+		return false, fmt.Errorf("the scheduler only supports kubernetes version 1.19-1.30")
 	}
 
 	return true, nil

k8s/config/crd/bases/k8s.v6d.io_backups.yaml

@@ -3,7 +3,7 @@ apiVersion: apiextensions.k8s.io/v1
 kind: CustomResourceDefinition
 metadata:
   annotations:
-    controller-gen.kubebuilder.io/version: v0.8.0
+    controller-gen.kubebuilder.io/version: v0.11.0
   creationTimestamp: null
   name: backups.k8s.v6d.io
 spec:
@@ -622,9 +622,3 @@ spec:
     storage: true
     subresources:
       status: {}
-status:
-  acceptedNames:
-    kind: ""
-    plural: ""
-  conditions: []
-  storedVersions: []

k8s/config/crd/bases/k8s.v6d.io_csidrivers.yaml

@@ -3,7 +3,7 @@ apiVersion: apiextensions.k8s.io/v1
 kind: CustomResourceDefinition
 metadata:
   annotations:
-    controller-gen.kubebuilder.io/version: v0.8.0
+    controller-gen.kubebuilder.io/version: v0.11.0
   creationTimestamp: null
   name: csidrivers.k8s.v6d.io
 spec:
@@ -95,9 +95,3 @@ spec:
     storage: true
     subresources:
       status: {}
-status:
-  acceptedNames:
-    kind: ""
-    plural: ""
-  conditions: []
-  storedVersions: []

k8s/config/crd/bases/k8s.v6d.io_globalobjects.yaml

@@ -3,7 +3,7 @@ apiVersion: apiextensions.k8s.io/v1
 kind: CustomResourceDefinition
 metadata:
   annotations:
-    controller-gen.kubebuilder.io/version: v0.8.0
+    controller-gen.kubebuilder.io/version: v0.11.0
   creationTimestamp: null
   name: globalobjects.k8s.v6d.io
 spec:
@@ -75,9 +75,3 @@ spec:
     storage: true
     subresources:
       status: {}
-status:
-  acceptedNames:
-    kind: ""
-    plural: ""
-  conditions: []
-  storedVersions: []

k8s/config/crd/bases/k8s.v6d.io_localobjects.yaml

@@ -3,7 +3,7 @@ apiVersion: apiextensions.k8s.io/v1
 kind: CustomResourceDefinition
 metadata:
   annotations:
-    controller-gen.kubebuilder.io/version: v0.8.0
+    controller-gen.kubebuilder.io/version: v0.11.0
   creationTimestamp: null
   name: localobjects.k8s.v6d.io
 spec:
@@ -82,9 +82,3 @@ spec:
     storage: true
     subresources:
       status: {}
-status:
-  acceptedNames:
-    kind: ""
-    plural: ""
-  conditions: []
-  storedVersions: []

k8s/config/crd/bases/k8s.v6d.io_operations.yaml

@@ -3,7 +3,7 @@ apiVersion: apiextensions.k8s.io/v1
 kind: CustomResourceDefinition
 metadata:
   annotations:
-    controller-gen.kubebuilder.io/version: v0.8.0
+    controller-gen.kubebuilder.io/version: v0.11.0
   creationTimestamp: null
   name: operations.k8s.v6d.io
 spec:
@@ -59,9 +59,3 @@ spec:
     storage: true
     subresources:
       status: {}
-status:
-  acceptedNames:
-    kind: ""
-    plural: ""
-  conditions: []
-  storedVersions: []

k8s/config/crd/bases/k8s.v6d.io_recovers.yaml

@@ -3,7 +3,7 @@ apiVersion: apiextensions.k8s.io/v1
 kind: CustomResourceDefinition
 metadata:
   annotations:
-    controller-gen.kubebuilder.io/version: v0.8.0
+    controller-gen.kubebuilder.io/version: v0.11.0
   creationTimestamp: null
   name: recovers.k8s.v6d.io
 spec:
@@ -53,9 +53,3 @@ spec:
     storage: true
     subresources:
       status: {}
-status:
-  acceptedNames:
-    kind: ""
-    plural: ""
-  conditions: []
-  storedVersions: []

k8s/config/crd/bases/k8s.v6d.io_sidecars.yaml

@@ -3,7 +3,7 @@ apiVersion: apiextensions.k8s.io/v1
 kind: CustomResourceDefinition
 metadata:
   annotations:
-    controller-gen.kubebuilder.io/version: v0.8.0
+    controller-gen.kubebuilder.io/version: v0.11.0
   creationTimestamp: null
   name: sidecars.k8s.v6d.io
 spec:
@@ -1524,9 +1524,3 @@ spec:
     storage: true
     subresources:
       status: {}
-status:
-  acceptedNames:
-    kind: ""
-    plural: ""
-  conditions: []
-  storedVersions: []

k8s/config/crd/bases/k8s.v6d.io_vineyardds.yaml

@@ -3,7 +3,7 @@ apiVersion: apiextensions.k8s.io/v1
 kind: CustomResourceDefinition
 metadata:
   annotations:
-    controller-gen.kubebuilder.io/version: v0.8.0
+    controller-gen.kubebuilder.io/version: v0.11.0
   creationTimestamp: null
   name: vineyardds.k8s.v6d.io
 spec:
@@ -80,6 +80,13 @@ spec:
               replicas:
                 default: 3
                 type: integer
+              scheduler:
+                properties:
+                  affinityWorkload:
+                    type: string
+                  gpuType:
+                    type: string
+                type: object
               securityContext:
                 properties:
                   allowPrivilegeEscalation:
@@ -1570,9 +1577,3 @@ spec:
     storage: true
     subresources:
       status: {}
-status:
-  acceptedNames:
-    kind: ""
-    plural: ""
-  conditions: []
-  storedVersions: []

k8s/pkg/templates/vineyardd/deployment.yaml

@@ -205,6 +205,28 @@ spec:
       - name: log
         emptyDir: {}
       affinity:
+        {{- if .Spec.Scheduler.GPUType }}
+        nodeAffinity:
+          requiredDuringSchedulingIgnoredDuringExecution:
+            nodeSelectorTerms:
+              - matchExpressions:
+                  # extract the key to configuration
+                  - key: machine.cluster.vke.volcengine.com/gpu-name
+                    operator: In
+                    values:
+                      - {{ .Spec.Scheduler.GPUType }}
+        {{- end }}
+        {{- if .Spec.Scheduler.AffinityWorkload }}
+        podAffinity:
+          requiredDuringSchedulingIgnoredDuringExecution:
+            - labelSelector:
+                matchExpressions:
+                  - key: "model.aibrix.ai/name"
+                    operator: In
+                    values:
+                      - {{ .Spec.Scheduler.AffinityWorkload }}
+              topologyKey: "kubernetes.io/hostname"
+        {{- end }}
         podAntiAffinity:
           requiredDuringSchedulingIgnoredDuringExecution:
             - labelSelector: