diff --git a/CHANGELOG.md b/CHANGELOG.md
index 15289cd44a2..072f1d7b55d 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -27,7 +27,8 @@
 
 ### Improvements
 
-- Graphite Scaler: use the latest datapoint returned, not the earliest (https://github.com/kedacore/keda/pull/2365)
+- Graphite Scaler: use the latest datapoint returned, not the earliest ([#2365](https://github.com/kedacore/keda/pull/2365))
+- Kubernetes Workload Scaler: ignore terminated pods ([#2384](https://github.com/kedacore/keda/pull/2384))
 
 - TODO ([#XXX](https://github.com/kedacore/keda/pull/XXX))
 
diff --git a/pkg/scalers/kubernetes_workload_scaler.go b/pkg/scalers/kubernetes_workload_scaler.go
index 153683ee32e..b7185b96ead 100644
--- a/pkg/scalers/kubernetes_workload_scaler.go
+++ b/pkg/scalers/kubernetes_workload_scaler.go
@@ -27,6 +27,11 @@ const (
 	valueKey                     = "value"
 )
 
+var phasesCountedAsTerminated = []corev1.PodPhase{
+	corev1.PodSucceeded,
+	corev1.PodFailed,
+}
+
 type kubernetesWorkloadMetadata struct {
 	podSelector labels.Selector
 	namespace   string
@@ -125,5 +130,19 @@ func (s *kubernetesWorkloadScaler) getMetricValue(ctx context.Context) (int, err
 		return 0, err
 	}
 
-	return len(podList.Items), nil
+	count := 0
+	for _, pod := range podList.Items {
+		count += getCountValue(pod)
+	}
+
+	return count, nil
+}
+
+func getCountValue(pod corev1.Pod) int {
+	for _, ignore := range phasesCountedAsTerminated {
+		if pod.Status.Phase == ignore {
+			return 0
+		}
+	}
+	return 1
 }
diff --git a/pkg/scalers/kubernetes_workload_scaler_test.go b/pkg/scalers/kubernetes_workload_scaler_test.go
index e59361489f8..c99feb21628 100644
--- a/pkg/scalers/kubernetes_workload_scaler_test.go
+++ b/pkg/scalers/kubernetes_workload_scaler_test.go
@@ -3,6 +3,7 @@ package scalers
 import (
 	"context"
 	"fmt"
+	"strings"
 	"testing"
 	"time"
 
@@ -138,3 +139,58 @@ func createPodlist(count int) *v1.PodList {
 	}
 	return list
 }
+
+func TestWorkloadPhase(t *testing.T) {
+	phases := map[v1.PodPhase]bool{
+		v1.PodRunning: true,
+		// succeeded and failed clearly count as terminated
+		v1.PodSucceeded: false,
+		v1.PodFailed:    false,
+		// unknown could be for example a temporarily unresponsive node; count the pod
+		v1.PodUnknown: true,
+		// count pre-Running to avoid an additional delay on top of the poll interval
+		v1.PodPending: true,
+	}
+	for phase, active := range phases {
+		list := &v1.PodList{}
+		pod := &v1.Pod{
+			ObjectMeta: metav1.ObjectMeta{
+				Name:        strings.ToLower(fmt.Sprintf("phase-%s", phase)),
+				Namespace:   "default",
+				Annotations: map[string]string{},
+				Labels: map[string]string{
+					"app": "testphases",
+				},
+			},
+			Status: v1.PodStatus{
+				Phase: phase,
+			},
+		}
+		list.Items = append(list.Items, *pod)
+		s, err := NewKubernetesWorkloadScaler(
+			fake.NewClientBuilder().WithRuntimeObjects(list).Build(),
+			&ScalerConfig{
+				TriggerMetadata: map[string]string{
+					"podSelector": "app=testphases",
+					"value":       "1",
+				},
+				AuthParams:        map[string]string{},
+				GlobalHTTPTimeout: 1000 * time.Millisecond,
+				Namespace:         "default",
+			},
+		)
+		if err != nil {
+			t.Errorf("Failed to create test scaler -- %v", err)
+		}
+		isActive, err := s.IsActive(context.TODO())
+		if err != nil {
+			t.Errorf("Failed to count active -- %v", err)
+		}
+		if active && !isActive {
+			t.Errorf("Expected active for phase %s but got inactive", phase)
+		}
+		if !active && isActive {
+			t.Errorf("Expected inactive for phase %s but got active", phase)
+		}
+	}
+}