fix: node discovery bug

Author: Code2Life

charts/tensor-fusion/Chart.yaml

@@ -15,10 +15,10 @@ type: application
 # This is the chart version. This version number should be incremented each time you make changes
 # to the chart and its templates, including the app version.
 # Versions are expected to follow Semantic Versioning (https://semver.org/)
-version: 1.1.0
+version: 1.1.1
 
 # This is the version number of the application being deployed. This version number should be
 # incremented each time you make changes to the application. Versions are not expected to
 # follow Semantic Versioning. They should reflect the version the application is using.
 # It is recommended to use it with quotes.
-appVersion: "1.12.0"
+appVersion: "1.12.1"

charts/tensor-fusion/templates/gpu-public-gpu-info.yaml

@@ -1,7 +1,7 @@
 apiVersion: v1
 kind: ConfigMap
 metadata:
-  name: {{ include "tensor-fusion.fullname" . }}-public-gpu-info
+  name: {{ .Release.Name }}-public-gpu-info
   namespace: {{ include "tensor-fusion.namespace" . }}
   labels:
     {{- include "tensor-fusion.labels" . | nindent 4 }}
@@ -22,81 +22,96 @@ data:
     
     # Turing Architecture Series
     - model: T4
+      fullModelName: "Tesla T4"
       vendor: NVIDIA
       costPerHour: 0.53
       fp16TFlops: 65
 
     # Ampere Architecture Series
     - model: A100_SXM4
+      fullModelName: "A100 SXM4"
       vendor: NVIDIA
       costPerHour: 1.89
       fp16TFlops: 312
     
     - model: A100_PCIe
+      fullModelName: "A100 PCIe"
       vendor: NVIDIA
       costPerHour: 1.64
       fp16TFlops: 312
 
     - model: A10
+      fullModelName: "A10"
       vendor: NVIDIA
       costPerHour: 0.9
       fp16TFlops: 125
 
     # A10G has less CUDA core than A10, but with RT cores for rendering case
     - model: A10G
+      fullModelName: "A10G"
       vendor: NVIDIA
       costPerHour: 0.75 # from lambda labs
       fp16TFlops: 125
 
     - model: A40
+      fullModelName: "A40"
       vendor: NVIDIA
       costPerHour: 0.44
       fp16TFlops: 125
     
     - model: RTX3090
+      fullModelName: "RTX3090"
       vendor: NVIDIA
       costPerHour: 0.43
       fp16TFlops: 143
 
     # Ada Lovelace Architecture Series  
     - model: L4
+      fullModelName: "L4"
       vendor: NVIDIA
       costPerHour: 0.43
       fp16TFlops: 121
 
     - model: L40
+      fullModelName: "L40"
       vendor: NVIDIA
       costPerHour: 0.86 # should be a bit cheaper than L40s
       fp16TFlops: 362
 
     - model: L40s
+      fullModelName: "L40s"
       vendor: NVIDIA
       costPerHour: 0.86
       fp16TFlops: 362
 
     - model: RTX4090
+      fullModelName: "RTX4090"
       vendor: NVIDIA
       costPerHour: 0.69
       fp16TFlops: 330
 
     # Hopper Architecture Series
     - model: H100_SXM4
+      fullModelName: "H100 SXM4"
       vendor: NVIDIA
       costPerHour: 2.99
       fp16TFlops: 989
     
     - model: H100_PCIe
+      fullModelName: "H100 PCIe"
       vendor: NVIDIA
       costPerHour: 2.39
       fp16TFlops: 835
    
     # Blackwell Architecture Series
     - model: B200_SXM4
+      fullModelName: "B200 SXM4"
       vendor: NVIDIA
       costPerHour: 10.99 # unknown price,on-request
       fp16TFlops: 2250
 
     - model: RTX5090
+      fullModelName: "RTX5090"
       vendor: NVIDIA
       costPerHour: 2.99
       fp16TFlops: 838

charts/tensor-fusion/templates/rbac.yaml

@@ -67,10 +67,21 @@ rules:
   - batch
   resources:
   - cronjobs
+  verbs:
+  - get
+  - list
+  - watch
+- apiGroups:
+  - batch
+  resources:
   - jobs
   verbs:
+  - create
+  - delete
   - get
   - list
+  - patch
+  - update
   - watch
 - apiGroups:
   - tensor-fusion.ai

config/rbac/role.yaml

@@ -68,10 +68,21 @@ rules:
   - batch
   resources:
   - cronjobs
+  verbs:
+  - get
+  - list
+  - watch
+- apiGroups:
+  - batch
+  resources:
   - jobs
   verbs:
+  - create
+  - delete
   - get
   - list
+  - patch
+  - update
   - watch
 - apiGroups:
   - tensor-fusion.ai

internal/controller/gpupool_controller.go

@@ -202,6 +202,7 @@ func (r *GPUPoolReconciler) startNodeDiscovery(
 	if err != nil {
 		return fmt.Errorf("unmarshal pod template: %w", err)
 	}
+
 	selector := labels.NewSelector()
 	poolReq, err := labels.NewRequirement(fmt.Sprintf(constants.GPUNodePoolIdentifierLabelFormat, pool.Name), selection.DoubleEquals, []string{"true"})
 	if err != nil {

internal/controller/node_controller.go

@@ -128,4 +128,6 @@ func (r *NodeReconciler) SetupWithManager(mgr ctrl.Manager) error {
 		For(&corev1.Node{}, builder.WithPredicates(p)).
 		Named("node").
 		Complete(r)
+	// When Pool changed, all nodes should re-generated, delete not matched ones
+	//
 }

internal/controller/tensorfusioncluster_controller.go

@@ -51,7 +51,7 @@ type TensorFusionClusterReconciler struct {
 // +kubebuilder:rbac:groups=tensor-fusion.ai,resources=tensorfusionclusters/finalizers,verbs=update
 // +kubebuilder:rbac:groups=core,resources=events,verbs=create;patch
 // +kubebuilder:rbac:groups=apps,resources=deployments;namespaces;configmaps;secrets,verbs=get;list;watch;create;update;patch;delete
-// +kubebuilder:rbac:groups=batch,resources=jobs,verbs=get;list;watch
+// +kubebuilder:rbac:groups=batch,resources=jobs,verbs=get;list;watch;create;update;patch;delete
 // +kubebuilder:rbac:groups=batch,resources=cronjobs,verbs=get;list;watch
 // +kubebuilder:rbac:groups="",resources=secrets,verbs=get;list;watch
 // +kubebuilder:rbac:groups=apps,resources=daemonsets,verbs=get;list;watch;create;update;patch;delete