[Spark] Failed to scale cluster

Environment:

Devstack with Heat and Neutron.

How to reproduce:

1. Create three node group templates. For example,

{
    "name": "slave-datanode",
    "description": "Test template for worker node",
    "flavor_id": "2",
    "plugin_name": "spark",
    "hadoop_version": "1.0.0",
    "node_processes": ["slave", "datanode"],
    "floating_ip_pool": "<ID of public network>"
}

{
    "name": "slave",
    "description": "Test template for worker node",
    "flavor_id": "2",
    "plugin_name": "spark",
    "hadoop_version": "1.0.0",
    "node_processes": ["slave"],
    "floating_ip_pool": "<ID of public network>"
}

{
    "name": "datanode",
    "description": "Test template for worker node",
    "flavor_id": "2",
    "plugin_name": "spark",
    "hadoop_version": "1.0.0",
    "node_processes": ["datanode"],
    "floating_ip_pool": "<ID of public network>"
}

2. Create a cluster template. For example,

{
    "name": "some-cluster-template",
    "description": "Test cluster template",
    "plugin_name": "spark",
    "hadoop_version": "1.0.0",
    "cluster_configs": {
      "HDFS": {
        "dfs.replication": 1
      }
    },
    "node_groups": [
        {
            "name": "master-node",
            "flavor_id": "2",
            "node_processes": ["master", "namenode"],
            "floating_ip_pool": "<ID of public network>",
            "count": 1
        },
        {
            "name": "worker-node-1",
            "node_group_template_id": "<ID of the first node group template>",
            "count": 1
        },
       {
            "name": "worker-node-2",
            "node_group_template_id": "<ID of the second node group template>",
            "count": 1
        },
       {
            "name": "worker-node-3",
            "node_group_template_id": "<ID of the third node group template>",
            "count": 1
        }
    ],
    "neutron_management_network": "<ID of private network>"
}

3. Create a cluster. For example,

{
    "name": "spark-cluster",
    "plugin_name": "spark",
    "hadoop_version": "1.0.0",
    "cluster_template_id" : "<ID of the cluster template>",
    "default_image_id": "<ID of an image>",
    "user_keypair_id": "<your key pair name>",
    "description": "Test cluster by Spark plugin"
}

4. Wait for "Active" status for cluster
5. Try to scale cluster. For example,

{
  "resize_node_groups": [
      {
        "name": "worker-node-2",
        "count": 0
      },
      {
        "name": "worker-node-3",
        "count": 0
      }
  ],
  "add_node_groups": [
    {
      "node_group_template_id": "<ID of the second node group template>",
      "count": 1,
      "name": "new-worker-node-2"
    },
    {
      "node_group_template_id": "<ID of the third node group template>",
      "count": 1,
      "name": "new-worker-node-3"
    }
  ]
}

Expected result:

Cluster can be successfully scaled.

Observed result:

Cluster hangs in "Decommissioning" status.