Merge tag 'v1.47.0rc3' into develop

Synapse 1.47.0rc3 (2021-11-16)
==============================

Bugfixes
--------

- Fix a bug introduced in 1.47.0rc1 which caused worker processes to not halt startup in the presence of outstanding database migrations. ([\#11346](https://github.com/matrix-org/synapse/issues/11346))
- Fix a bug introduced in 1.47.0rc1 which prevented the 'remove deleted devices from `device_inbox` column' background process from running when updating from a recent Synapse version. ([\#11303](https://github.com/matrix-org/synapse/issues/11303), [\#11353](https://github.com/matrix-org/synapse/issues/11353))

2 files changed, 35 insertions, 19 deletions

diff --git a/synapse/storage/prepare_database.py b/synapse/storage/prepare_database.py
index 8b9c6adae2..e45adfcb55 100644
--- a/synapse/storage/prepare_database.py
+++ b/synapse/storage/prepare_database.py
@@ -131,24 +131,16 @@ def prepare_database(
                     "config==None in prepare_database, but database is not empty"
                 )
 
-            # if it's a worker app, refuse to upgrade the database, to avoid multiple
-            # workers doing it at once.
-            if config.worker.worker_app is None:
-                _upgrade_existing_database(
-                    cur,
-                    version_info,
-                    database_engine,
-                    config,
-                    databases=databases,
-                )
-            elif version_info.current_version < SCHEMA_VERSION:
-                # If the DB is on an older version than we expect then we refuse
-                # to start the worker (as the main process needs to run first to
-                # update the schema).
-                raise UpgradeDatabaseException(
-                    OUTDATED_SCHEMA_ON_WORKER_ERROR
-                    % (SCHEMA_VERSION, version_info.current_version)
-                )
+            # This should be run on all processes, master or worker. The master will
+            # apply the deltas, while workers will check if any outstanding deltas
+            # exist and raise an PrepareDatabaseException if they do.
+            _upgrade_existing_database(
+                cur,
+                version_info,
+                database_engine,
+                config,
+                databases=databases,
+            )
 
         else:
             logger.info("%r: Initialising new database", databases)
@@ -358,6 +350,18 @@ def _upgrade_existing_database(
 
     is_worker = config and config.worker.worker_app is not None
 
+    # If the schema version needs to be updated, and we are on a worker, we immediately
+    # know to bail out as workers cannot update the database schema. Only one process
+    # must update the database at the time, therefore we delegate this task to the master.
+    if is_worker and current_schema_state.current_version < SCHEMA_VERSION:
+        # If the DB is on an older version than we expect then we refuse
+        # to start the worker (as the main process needs to run first to
+        # update the schema).
+        raise UpgradeDatabaseException(
+            OUTDATED_SCHEMA_ON_WORKER_ERROR
+            % (SCHEMA_VERSION, current_schema_state.current_version)
+        )
+
     if (
         current_schema_state.compat_version is not None
         and current_schema_state.compat_version > SCHEMA_VERSION
diff --git a/synapse/storage/schema/main/delta/65/05remove_deleted_devices_from_device_inbox.sql b/synapse/storage/schema/main/delta/65/06remove_deleted_devices_from_device_inbox.sql
index 076179123d..82f6408b36 100644
--- a/synapse/storage/schema/main/delta/65/05remove_deleted_devices_from_device_inbox.sql
+++ b/synapse/storage/schema/main/delta/65/06remove_deleted_devices_from_device_inbox.sql
@@ -18,5 +18,17 @@
 -- when a device was deleted using Synapse earlier than 1.47.0.
 -- This runs as background task, but may take a bit to finish.
 
+-- Remove any existing instances of this job running. It's OK to stop and restart this job,
+-- as it's just deleting entries from a table - no progress will be lost.
+--
+-- This is necessary due a similar migration running the job accidentally
+-- being included in schema version 64 during v1.47.0rc1,rc2. If a
+-- homeserver had updated from Synapse <=v1.45.0 (schema version <=64),
+-- then they would have started running this background update already.
+-- If that update was still running, then simply inserting it again would
+-- cause an SQL failure. So we effectively do an "upsert" here instead.
+
+DELETE FROM background_updates WHERE update_name = 'remove_deleted_devices_from_device_inbox';
+
 INSERT INTO background_updates (ordering, update_name, progress_json) VALUES
-  (6505, 'remove_deleted_devices_from_device_inbox', '{}');
+  (6506, 'remove_deleted_devices_from_device_inbox', '{}');