diff --git a/doc/source/placement.rst b/doc/source/placement.rst
index f509314106f4..2637ae03b997 100644
--- a/doc/source/placement.rst
+++ b/doc/source/placement.rst
@@ -186,6 +186,13 @@ Ocata (15.0.0)
 * Users of CellsV1 will need to deploy a placement per cell, matching
   the scope and cardinality of the regular ``nova-scheduler`` process.
 
+Pike (16.0.0)
+~~~~~~~~~~~~~
+
+* The ``nova.scheduler.filter_scheduler.FilterScheduler`` in Pike will
+  no longer fall back to not using the Placement Service, even if older
+  computes are running in the deployment.
+
 
 REST API
 ========
diff --git a/nova/scheduler/filter_scheduler.py b/nova/scheduler/filter_scheduler.py
index 658a7169ab82..db65542a8574 100644
--- a/nova/scheduler/filter_scheduler.py
+++ b/nova/scheduler/filter_scheduler.py
@@ -27,7 +27,6 @@ from six.moves import range
 import nova.conf
 from nova import exception
 from nova.i18n import _
-from nova import objects
 from nova.objects import fields
 from nova import rpc
 from nova.scheduler import client as scheduler_client
@@ -166,18 +165,6 @@ class FilterScheduler(driver.Scheduler):
 
     def _get_all_host_states(self, context, spec_obj):
         """Template method, so a subclass can implement caching."""
-        # NOTE(sbauza): Since Newton compute nodes require a configuration
-        # change to request the Placement API, and given it goes against
-        # our rolling upgrade process, we define a graceful period for allowing
-        # clouds that are not fully upgraded to Ocata to still be able to
-        # have instances being scheduled on old nodes.
-        service_version = objects.Service.get_minimum_version(
-            context, 'nova-compute')
-        # TODO(sbauza): Remove that version check in Pike so we fully call
-        # the placement API anyway.
-        if service_version < 16:
-            LOG.debug("Skipping call to placement, as upgrade in progress.")
-            return self.host_manager.get_all_host_states(context)
         filters = {'resources': self._get_resources_per_request_spec(spec_obj)}
         reportclient = self.scheduler_client.reportclient
         rps = reportclient.get_filtered_resource_providers(filters)
diff --git a/nova/tests/unit/scheduler/test_filter_scheduler.py b/nova/tests/unit/scheduler/test_filter_scheduler.py
index d5820b5ea20a..941553cf694c 100644
--- a/nova/tests/unit/scheduler/test_filter_scheduler.py
+++ b/nova/tests/unit/scheduler/test_filter_scheduler.py
@@ -262,18 +262,6 @@ class FilterSchedulerTestCase(test_scheduler.SchedulerTestCase):
 
         self.assertEqual(50, hosts[0].weight)
 
-    # TODO(sbauza): Remove that unit test in Pike
-    @mock.patch('nova.objects.Service.get_minimum_version',
-                return_value=15)
-    def test_get_all_host_states_with_newton_computes(self, mock_get_mv):
-        """If at least one compute node is older than Ocata, then we should
-        not call the placement API.
-        """
-        with mock.patch.object(self.driver.host_manager,
-                               'get_all_host_states') as mock_get_hosts:
-            self.driver._get_all_host_states(self.context, mock.sentinel.spec)
-        mock_get_hosts.assert_called_once_with(self.context)
-
     @mock.patch('nova.objects.ServiceList.get_by_binary',
                 return_value=fakes.SERVICES)
     @mock.patch('nova.objects.InstanceList.get_by_host')
@@ -282,12 +270,10 @@ class FilterSchedulerTestCase(test_scheduler.SchedulerTestCase):
     @mock.patch('nova.scheduler.client.report.SchedulerReportClient.'
                 'get_filtered_resource_providers',
                 return_value=fakes.RESOURCE_PROVIDERS)
-    @mock.patch('nova.objects.Service.get_minimum_version',
-                return_value=objects.service.SERVICE_VERSION)
     @mock.patch('nova.db.instance_extra_get_by_instance_uuid',
                 return_value={'numa_topology': None,
                               'pci_requests': None})
-    def test_select_destinations(self, mock_get_extra, mock_get_mv,
+    def test_select_destinations(self, mock_get_extra,
                                  mock_get_rps, mock_get_all,
                                  mock_by_host, mock_get_by_binary):
         """select_destinations is basically a wrapper around _schedule().
@@ -333,7 +319,6 @@ class FilterSchedulerTestCase(test_scheduler.SchedulerTestCase):
         (host, node) = (dests[0]['host'], dests[0]['nodename'])
         self.assertEqual(host, selected_hosts[0])
         self.assertEqual(node, selected_nodes[0])
-        mock_get_mv.assert_called_once_with(mock.ANY, 'nova-compute')
 
     @mock.patch.object(filter_scheduler.FilterScheduler, '_schedule')
     def test_select_destinations_notifications(self, mock_schedule):
diff --git a/releasenotes/notes/no-placement-fallback-5db2d0645f51aca8.yaml b/releasenotes/notes/no-placement-fallback-5db2d0645f51aca8.yaml
new file mode 100644
index 000000000000..133da5fb2d46
--- /dev/null
+++ b/releasenotes/notes/no-placement-fallback-5db2d0645f51aca8.yaml
@@ -0,0 +1,6 @@
+---
+upgrade:
+  - In Ocata, the nova-scheduler would fall back to not calling the
+    placement service during instance boot if old computes were
+    running. That compatibility mode is no longer present in Pike, and
+    as such, the scheduler fully depends on the placement service.
\ No newline at end of file