Wait for current_state_events_membership before populate_stats_process_rooms (#7387)

populate_stats_process_rooms was added in #5971 / v1.4.0; current_state_events_membership was added in #5706 / v1.3.0.

Fixes #7380.

UPGRADE.rst

@@ -75,6 +75,37 @@ for example:
      wget https://packages.matrix.org/debian/pool/main/m/matrix-synapse-py3/matrix-synapse-py3_1.3.0+stretch1_amd64.deb
      dpkg -i matrix-synapse-py3_1.3.0+stretch1_amd64.deb
 
+Upgrading to v1.13.0
+====================
+
+Incorrect database migration in old synapse versions
+----------------------------------------------------
+
+A bug was introduced in Synapse 1.4.0 which could cause the room directory to
+be incomplete or empty if Synapse was upgraded directly from v1.2.1 or earlier,
+to versions between v1.4.0 and v1.12.x.
+
+This will *not* be a problem for Synapse installations which were:
+ * created at v1.4.0 or later,
+ * upgraded via v1.3.x, or
+ * upgraded straight from v1.2.1 or earlier to v1.13.0 or later.
+
+If completeness of the room directory is a concern, installations which are
+affected can be repaired as follows:
+
+1. Run the following sql from a `psql` or `sqlite3` console:
+
+   .. code:: sql
+
+     INSERT INTO background_updates (update_name, progress_json, depends_on) VALUES
+        ('populate_stats_process_rooms', '{}', 'current_state_events_membership');
+
+     INSERT INTO background_updates (update_name, progress_json, depends_on) VALUES
+        ('populate_stats_process_users', '{}', 'populate_stats_process_rooms');
+
+2. Restart synapse.
+
+
 Upgrading to v1.12.0
 ====================
 

changelog.d/7387.bugfix

@@ -0,0 +1 @@
+Fix a bug which would cause the room durectory to be incorrectly populated if Synapse was upgraded directly from v1.2.1 or earlier to v1.4.0 or later. Note that this fix does not apply retrospectively; see the [upgrade notes](UPGRADE.rst#upgrading-to-v1130) for more information.

synapse/storage/data_stores/main/schema/delta/56/stats_separated.sql

@@ -35,9 +35,13 @@ DELETE FROM background_updates WHERE update_name IN (
     'populate_stats_cleanup'
 );
 
+-- this relies on current_state_events.membership having been populated, so add
+-- a dependency on current_state_events_membership.
 INSERT INTO background_updates (update_name, progress_json, depends_on) VALUES
-    ('populate_stats_process_rooms', '{}', '');
+    ('populate_stats_process_rooms', '{}', 'current_state_events_membership');
 
+-- this also relies on current_state_events.membership having been populated, but
+-- we get that as a side-effect of depending on populate_stats_process_rooms.
 INSERT INTO background_updates (update_name, progress_json, depends_on) VALUES
     ('populate_stats_process_users', '{}', 'populate_stats_process_rooms');