chore: merge code unit test issues

Author: Code2Life

internal/controller/tensorfusionworkload_controller.go

@@ -393,7 +393,7 @@ func (r *TensorFusionWorkloadReconciler) scaleUpWorkers(ctx context.Context, wor
 			return ctrl.Result{RequeueAfter: constants.PendingRequeueDuration}, nil
 		}
 
-		pod, err := r.tryStartWorker(ctx, workerGenerator, gpus, workload, hash)
+		_, err = r.tryStartWorker(ctx, workerGenerator, gpus, workload, hash)
 		if err != nil {
 			// Try to release all allocated GPUs if pod creation fails
 			gpus := lo.Map(gpus, func(gpu *tfv1.GPU, _ int) types.NamespacedName {

internal/gpuallocator/gpuallocator.go

@@ -138,7 +138,7 @@ func (s *GpuAllocator) Alloc(
 }
 
 // Dealloc a request from gpu to release available resources on it.
-func (s *GpuAllocator) Dealloc(ctx context.Context, workloadNameNamespace tfv1.NameNamespace, request tfv1.Resource, gpu *tfv1.GPU) error {
+func (s *GpuAllocator) Dealloc(ctx context.Context, workloadNameNamespace tfv1.NameNamespace, request tfv1.Resource, gpus []types.NamespacedName) error {
 	log := log.FromContext(ctx)
 	s.storeMutex.Lock()
 	defer s.storeMutex.Unlock()

internal/gpuallocator/gpuallocator_test.go

@@ -176,7 +176,7 @@ var _ = Describe("GPU Allocator", func() {
 			allocatedVram := allocatedGPU.Status.Available.Vram.DeepCopy()
 
 			// Now deallocate
-			err = allocator.Dealloc(ctx, workloadNameNs, request, allocatedGPU)
+			err = allocator.Dealloc(ctx, workloadNameNs, request, []types.NamespacedName{client.ObjectKeyFromObject(gpus[0])})
 			Expect(err).NotTo(HaveOccurred())
 
 			allocator.syncToK8s(ctx)
@@ -201,7 +201,7 @@ var _ = Describe("GPU Allocator", func() {
 			}
 
 			// Allocate 2 GPUs
-			allocatedGPUs, err := allocator.Alloc(ctx, "test-pool", request, 2, "")
+			allocatedGPUs, err := allocator.Alloc(ctx, "test-pool", tfv1.NameNamespace{Namespace: "default", Name: "test-workload"}, request, 2, "")
 			Expect(err).NotTo(HaveOccurred())
 			Expect(allocatedGPUs).To(HaveLen(2))
 
@@ -234,7 +234,7 @@ var _ = Describe("GPU Allocator", func() {
 				return client.ObjectKeyFromObject(gpu)
 			})
 			// Now deallocate all GPUs including the non-existent one
-			err = allocator.Dealloc(ctx, request, gpusToDeallocKeys)
+			err = allocator.Dealloc(ctx, tfv1.NameNamespace{Namespace: "default", Name: "test-workload"}, request, gpusToDeallocKeys)
 			Expect(err).NotTo(HaveOccurred())
 
 			// Verify resources were restored for existing GPUs

internal/metrics/recorder.go

@@ -178,7 +178,7 @@ func (mr *MetricsRecorder) RecordMetrics(writer io.Writer) {
 		enc.AddTag("worker_name", metrics.WorkerName)
 		enc.AddTag("workload_name", metrics.WorkloadName)
 
-		enc.AddField("gpu_count", metricsProto.MustNewValue(metrics.GPUCount))
+		enc.AddField("gpu_count", metricsProto.MustNewValue(int64(metrics.GPUCount)))
 		enc.AddField("tflops_limit", metricsProto.MustNewValue(metrics.TflopsLimit))
 		enc.AddField("tflops_request", metricsProto.MustNewValue(metrics.TflopsRequest))
 		enc.AddField("raw_cost", metricsProto.MustNewValue(metrics.RawCost))
@@ -205,7 +205,7 @@ func (mr *MetricsRecorder) RecordMetrics(writer io.Writer) {
 		enc.AddField("allocated_tflops_percent", metricsProto.MustNewValue(metrics.AllocatedTflopsPercent))
 		enc.AddField("allocated_vram_bytes", metricsProto.MustNewValue(metrics.AllocatedVramBytes))
 		enc.AddField("allocated_vram_percent", metricsProto.MustNewValue(metrics.AllocatedVramPercent))
-		enc.AddField("gpu_count", metricsProto.MustNewValue(len(metrics.GPUModels)))
+		enc.AddField("gpu_count", metricsProto.MustNewValue(int64(len(metrics.GPUModels))))
 		enc.AddField("raw_cost", metricsProto.MustNewValue(metrics.RawCost))
 		enc.EndLine(now)
 	}