Merge pull request kubernetes#47523 from MaciekPytel/fix_autoscaler_flaky_e2e

Automatic merge from submit-queue

Fix flaking cluster-autoscaler e2e

Ref: kubernetes/autoscaler#89

Scale-down e2e require 5 healthy nodes to pass reliably, this PR should fix the flaking "should correctly scale down after a node is not needed and one node is broken" e2e.

Author: Kubernetes Submit Queue

test/e2e/autoscaling/cluster_size_autoscaling.go

@@ -350,8 +350,8 @@ var _ = framework.KubeDescribe("Cluster size autoscaling [Slow]", func() {
 		increasedSize := 0
 		newSizes := make(map[string]int)
 		for key, val := range originalSizes {
-			newSizes[key] = val + 2
-			increasedSize += val + 2
+			newSizes[key] = val + 2 + unready
+			increasedSize += val + 2 + unready
 		}
 		setMigSizes(newSizes)
 		framework.ExpectNoError(WaitForClusterSizeFuncWithUnready(f.ClientSet,