Merge "Fix instance.hidden migration and querying"

nova/db/sqlalchemy/api.py

@@ -2205,9 +2205,15 @@ def instance_get_all_by_filters_sort(context, filters, limit=None, marker=None,
         else:
             filters['user_id'] = context.user_id
 
-    if 'hidden' not in filters:
+    if filters.pop('hidden', False):
-        # Filter out hidden instances by default.
+        query_prefix = query_prefix.filter(models.Instance.hidden == true())
-        filters['hidden'] = False
+    else:
+        # If the query should not include hidden instances, then
+        # filter instances with hidden=False or hidden=NULL because
+        # older records may have no value set.
+        query_prefix = query_prefix.filter(or_(
+            models.Instance.hidden == false(),
+            models.Instance.hidden == null()))
 
     # Filters for exact matches that we can do along with the SQL query...
     # For other filters that don't match this, we will do regexp matching

nova/db/sqlalchemy/migrate_repo/versions/399_add_instances_hidden.py

@@ -20,5 +20,9 @@ def upgrade(migrate_engine):
     for prefix in ('', 'shadow_'):
         instances = Table('%sinstances' % prefix, meta, autoload=True)
         if not hasattr(instances.c, 'hidden'):
-            hidden = Column('hidden', Boolean, default=False)
+            # NOTE(danms): This column originally included default=False. We
+            # discovered in bug #1862205 that this will attempt to rewrite
+            # the entire instances table with that value, which can time out
+            # for large data sets (and does not even abort).
+            hidden = Column('hidden', Boolean)
             instances.create_column(hidden)

nova/objects/instance.py

@@ -1517,7 +1517,9 @@ class InstanceList(base.ObjectListBase, base.NovaObject):
         # NOTE(mriedem): Filter out hidden instances since there should be a
         # non-hidden version of the instance in another cell database and the
         # API will only show one of them, so we don't count the hidden copy.
-        project_query = project_query.filter_by(hidden=false())
+        project_query = project_query.filter(
+            or_(models.Instance.hidden == false(),
+                models.Instance.hidden == null()))
 
         project_result = project_query.first()
         fields = ('instances', 'cores', 'ram')

nova/tests/unit/objects/test_instance.py

@@ -27,6 +27,8 @@ from oslo_versionedobjects import base as ovo_base
 from nova.compute import task_states
 from nova.compute import vm_states
 from nova.db import api as db
+from nova.db.sqlalchemy import api as sql_api
+from nova.db.sqlalchemy import models as sql_models
 from nova import exception
 from nova.network import model as network_model
 from nova import notifications
@@ -1964,7 +1966,53 @@ class _TestInstanceListObject(object):
 
 class TestInstanceListObject(test_objects._LocalTest,
                              _TestInstanceListObject):
-    pass
+    # No point in doing this db-specific test twice for remote
+    def test_hidden_filter_query(self):
+        """Check that our instance_get_by_filters() honors hidden properly
+
+        As reported in bug #1862205, we need to properly handle instances
+        with the hidden field set to NULL and not expect SQLAlchemy to
+        translate those values on SELECT.
+        """
+        values = {'user_id': self.context.user_id,
+                  'project_id': self.context.project_id,
+                  'host': 'foo'}
+
+        for hidden_value in (True, False):
+            db.instance_create(self.context,
+                               dict(values, hidden=hidden_value))
+
+        # NOTE(danms): Because the model has default=False, we can not use
+        # it to create an instance with a hidden value of NULL. So, do it
+        # manually here.
+        engine = sql_api.get_engine()
+        table = sql_models.Instance.__table__
+        with engine.connect() as conn:
+            update = table.insert().values(user_id=self.context.user_id,
+                                           project_id=self.context.project_id,
+                                           uuid=uuids.nullinst,
+                                           host='foo',
+                                           hidden=None)
+            conn.execute(update)
+
+        insts = objects.InstanceList.get_by_filters(self.context,
+                                                    {'hidden': True})
+        # We created one hidden instance above, so expect only that one
+        # to come out of this query.
+        self.assertEqual(1, len(insts))
+
+        # We created one unhidden instance above, and one specifically
+        # with a NULL value to represent an unmigrated instance, which
+        # defaults to hidden=False, so expect both of those here.
+        insts = objects.InstanceList.get_by_filters(self.context,
+                                                    {'hidden': False})
+        self.assertEqual(2, len(insts))
+
+        # Do the same check as above, but make sure hidden=False is the
+        # default behavior.
+        insts = objects.InstanceList.get_by_filters(self.context,
+                                                    {})
+        self.assertEqual(2, len(insts))
 
 
 class TestRemoteInstanceListObject(test_objects._RemoteTest,

releasenotes/notes/instances_hidden_after_upgrade_to_train-9ce4731f31bc6bd2.yaml

@@ -0,0 +1,19 @@
+---
+upgrade:
+  - |
+    Upgrading to Train on a deployment with a large database may hit
+    `bug 1862205`_, which results in instance records left in a bad
+    state, and manifests as instances not being shown in list
+    operations. Users upgrading to Train for the first time will
+    definitely want to apply a version which includes this fix.  Users
+    already on Train should upgrade to a version including this fix to
+    ensure the problem is addressed.
+
+    .. _bug 1862205: https://launchpad.net/bugs/1862205
+fixes:
+  - |
+    A fix for serious `bug 1862205`_ is provided which addresses both
+    the performance aspect of schema migration 399, as well as the
+    potential fallout for cases where this migration silently fails
+    and leaves large numbers of instances hidden from view from the
+    API.