Fix bug about unintentional scale out during updating deployment.

shibataka000 · Joseph Burnett · commit d2e35c7a6a7b · 2020-03-26T12:30:24.000+01:00
During rolling update with maxSurge=1 and maxUnavailable=0,
len(metrics) is greater than currentReplcas
and it may cause unintentional scale out.
diff --git a/pkg/controller/podautoscaler/replica_calculator.go b/pkg/controller/podautoscaler/replica_calculator.go
@@ -134,9 +134,15 @@ func (c *ReplicaCalculator) GetResourceReplicas(currentReplicas int32, targetUti
 		return currentReplicas, utilization, rawUtilization, timestamp, nil
 	}
 
+	newReplicas := int32(math.Ceil(newUsageRatio * float64(len(metrics))))
+	if (newUsageRatio < 1.0 && newReplicas > currentReplicas) || (newUsageRatio > 1.0 && newReplicas < currentReplicas) {
+		// return the current replicas if the change of metrics length would cause a change in scale direction
+		return currentReplicas, utilization, rawUtilization, timestamp, nil
+	}
+
 	// return the result, where the number of replicas considered is
 	// however many replicas factored into our calculation
-	return int32(math.Ceil(newUsageRatio * float64(len(metrics)))), utilization, rawUtilization, timestamp, nil
+	return newReplicas, utilization, rawUtilization, timestamp, nil
 }
 
 // GetRawResourceReplicas calculates the desired replica count based on a target resource utilization (as a raw milli-value)
diff --git a/pkg/controller/podautoscaler/replica_calculator_test.go b/pkg/controller/podautoscaler/replica_calculator_test.go
@@ -1247,6 +1247,24 @@ func TestReplicaCalcMissingMetricsUnreadyScaleDown(t *testing.T) {
 	tc.runTest(t)
 }
 
+func TestReplicaCalcDuringRollingUpdateWithMaxSurge(t *testing.T) {
+	tc := replicaCalcTestCase{
+		currentReplicas:  2,
+		expectedReplicas: 2,
+		podPhase:         []v1.PodPhase{v1.PodRunning, v1.PodRunning, v1.PodRunning},
+		resource: &resourceInfo{
+			name:     v1.ResourceCPU,
+			requests: []resource.Quantity{resource.MustParse("1.0"), resource.MustParse("1.0"), resource.MustParse("1.0")},
+			levels:   []int64{100, 100},
+
+			targetUtilization:   50,
+			expectedUtilization: 10,
+			expectedValue:       numContainersPerPod * 100,
+		},
+	}
+	tc.runTest(t)
+}
+
 // TestComputedToleranceAlgImplementation is a regression test which
 // back-calculates a minimal percentage for downscaling based on a small percentage
 // increase in pod utilization which is calibrated against the tolerance value.
