latest

Signed-off-by: Braulio Dumba <Braulio.Dumba@ibm.com>

Author: dumb0002

internal/engines/scalefromzero/engine.go

@@ -40,9 +40,6 @@ import (
 	poolutil "github.com/llm-d-incubation/workload-variant-autoscaler/internal/utils/pool"
 )
 
-// NOTE: This is a placeholder for the scale-from-zero engine implementation.
-// The actual logic for the scale-from-zero engine should be implemented here.
-
 type Engine struct {
 	client        client.Client
 	executor      executor.Executor
@@ -194,32 +191,34 @@ func (e *Engine) processInactiveVariant(ctx context.Context, va wvav1alpha1.Vari
 		return err
 	}
 
-	// result := results["all_metrics"]
-	// for _, value := range result.Values {
-	// 	// Check for pending requests using queue size metrics
-	// 	metricName := value.Labels["__name__"]
-	// 	if metricName == "inference_pool_average_queue_size" && value.Value > 0 {
-	// 		// ctrl.Log.V(logging.DEBUG).Info(
-	// 		// 	"Target worklod has pending request, not scaling up", "metricName", metricName,
-	// 		// 	"metric", value.Labels, "value", value.Value)
-
-	// 		ctrl.Log.Info(
-	// 			"Target worklod has pending request, not scaling up", "metricName", metricName,
-	// 			"metric", value.Labels, "value", value.Value)
-	// 		break
-	// 	}
-	// // }
-
-	// // Scale up
+	result := results["all_metrics"]
+	for _, value := range result.Values {
+		// Check for pending requests using queue size metrics
+		metricName := value.Labels["__name__"]
+		if metricName == "inference_pool_average_queue_size" && value.Value > 0 {
+			// ctrl.Log.V(logging.DEBUG).Info(
+			// 	"Target worklod has pending request, not scaling up", "metricName", metricName,
+			// 	"metric", value.Labels, "value", value.Value)
+
+			ctrl.Log.Info(
+				"Target worklod has pending request, not scaling up", "metricName", metricName,
+				"metric", value.Labels, "value", value.Value)
+			break
+		}
+	}
+
+	// **** Compare the modelname in the VA and EPP collected metrics ****
+	// TO DO: Compute scalefromzero decision and implement direct actuati
+
+	// // Scale up from zero to one
+	// ADD A NOTE HERE:
 	// err = e.Actuator.ScaleTargetObject(ctx, unstructuredObj, int32(1))
 	// if err != nil {
 	// 	ctrl.Log.Error(err, "Error scaling up Target Workload", "EPP source", pool.EndpointPicker.ServiceName)
 	// 	return err
 	// }
 
 	ctrl.Log.Info("Target workload was succesfully scaling up", "EPP source", pool.EndpointPicker.ServiceName)
-	// **** Compare the modelname in the VA and EPP collected metrics ****
-	// TO DO: Compute scalefromzero decision and implement direct actuation
 
 	return nil
 }