Remove redundant group host setup

It seems that all the code path that ends up calling [1] goes through
[2] before so setting up group host at [1] is redundant.

[1] https://github.com/openstack/nova/blob/c7e9e667426a6d88d396a59cb40d30763a3265f9/nova/scheduler/utils.py#L944-L950
[2] https://github.com/openstack/nova/blob/c7e9e667426a6d88d396a59cb40d30763a3265f9/nova/scheduler/utils.py#L996-L1000

Change-Id: I745aecb008f0867fcda2d3d2c7691621dd8f6168
Closes-Bug: #1788527
(cherry picked from commit f108d79)

Author: Balazs Gibizer

nova/scheduler/utils.py

@@ -898,13 +898,7 @@ def _get_group_details(context, instance_uuid, user_group_hosts=None):
             raise exception.UnsupportedPolicyException(reason=msg)
         # NOTE(melwitt): If the context is already targeted to a cell (during a
         # move operation), we don't need to scatter-gather.
-        if context.db_connection:
-            # We don't need to target the group object's context because it was
-            # retrieved with the targeted context earlier in this method.
-            group_hosts = set(group.get_hosts())
-        else:
-            group_hosts = set(_get_instance_group_hosts_all_cells(context,
-                                                                  group))
+        group_hosts = set(group.get_hosts())
         user_hosts = set(user_group_hosts) if user_group_hosts else set()
         return GroupDetails(hosts=user_hosts | group_hosts,
                             policy=group.policy, members=group.members)