Cluster profile: cope with nonexistent node types (#1456)

Sometimes, one kind of a node just doesn't exist, and we don't
want to throw a KeyError on that.

Specifically, right now in the separate-services deployment, we don't
yet have the "manager" node declared.

Author: tehasdf

cloudify_cli/commands/cluster.py

@@ -261,7 +261,7 @@ def _update_cluster_nodes(nodes, nodes_type, logger):
         _update_node(node, nodes_type, logger, stored_nodes_names)
     # filter out removed nodes
     env.profile.cluster[nodes_type] = [
-        node for node in env.profile.cluster[nodes_type]
+        node for node in env.profile.cluster.get(nodes_type, {})
         if _get_node_host(node) in received_nodes_names]
     env.profile.save()