K8s schedule pods to removed nodes

[hongbin]

Steps to reproduce:

* Create a k8s cluster with 2 nodes
$ magnum bay-create --name k8sbay --baymodel k8sbaymodel --node-count 2

* Reduce the node count to 1
$ magnum bay-update k8sbay replace node_count=1

* Create 2 pods
$ magnum pod-create --manifest ./redis-master.yaml --bay k8sbay
$ magnum pod-create --manifest ./redis-proxy.yaml --bay k8sbay

Then, you should see one of the pods are scheduled to a removed nodes.

[hongbin (hongbin034)]

This is actually a bug from k8s https://github.com/GoogleCloudPlatform/kubernetes/issues/7222 . The fix has been committed at May 2015. Possibly, we need to upgrade the fedora image to fix that.

[Jaycen Grant (jaycen-v-grant)]

Pod commands have been removed.