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](#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](#11303), [\#11353](#11353))

CHANGES.md

@@ -1,3 +1,13 @@
+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))
+
+
 Synapse 1.47.0rc2 (2021-11-10)
 ==============================
 

debian/changelog

@@ -1,3 +1,9 @@
+matrix-synapse-py3 (1.47.0~rc3) stable; urgency=medium
+
+  * New synapse release 1.47.0~rc3.
+
+ -- Synapse Packaging team <packages@matrix.org>  Tue, 16 Nov 2021 14:32:47 +0000
+
 matrix-synapse-py3 (1.47.0~rc2) stable; urgency=medium
 
   [ Dan Callahan ]

synapse/__init__.py

@@ -47,7 +47,7 @@
 except ImportError:
     pass
 
-__version__ = "1.47.0rc2"
+__version__ = "1.47.0rc3"
 
 if bool(os.environ.get("SYNAPSE_TEST_PATCH_LOG_CONTEXTS", False)):
     # We import here so that we don't have to install a bunch of deps when

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

synapse/storage/schema/main/delta/65/05remove_deleted_devices_from_device_inbox.sql → 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', '{}');

tests/storage/test_rollback_worker.py

@@ -11,6 +11,9 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+from typing import List
+from unittest import mock
+
 from synapse.app.generic_worker import GenericWorkerServer
 from synapse.storage.database import LoggingDatabaseConnection
 from synapse.storage.prepare_database import PrepareDatabaseException, prepare_database
@@ -19,6 +22,22 @@
 from tests.unittest import HomeserverTestCase
 
 
+def fake_listdir(filepath: str) -> List[str]:
+    """
+    A fake implementation of os.listdir which we can use to mock out the filesystem.
+
+    Args:
+        filepath: The directory to list files for.
+
+    Returns:
+        A list of files and folders in the directory.
+    """
+    if filepath.endswith("full_schemas"):
+        return [SCHEMA_VERSION]
+
+    return ["99_add_unicorn_to_database.sql"]
+
+
 class WorkerSchemaTests(HomeserverTestCase):
     def make_homeserver(self, reactor, clock):
         hs = self.setup_test_homeserver(
@@ -51,7 +70,7 @@ def test_rolling_back(self):
 
         prepare_database(db_conn, db_pool.engine, self.hs.config)
 
-    def test_not_upgraded(self):
+    def test_not_upgraded_old_schema_version(self):
         """Test that workers don't start if the DB has an older schema version"""
         db_pool = self.hs.get_datastore().db_pool
         db_conn = LoggingDatabaseConnection(
@@ -67,3 +86,34 @@ def test_not_upgraded(self):
 
         with self.assertRaises(PrepareDatabaseException):
             prepare_database(db_conn, db_pool.engine, self.hs.config)
+
+    def test_not_upgraded_current_schema_version_with_outstanding_deltas(self):
+        """
+        Test that workers don't start if the DB is on the current schema version,
+        but there are still outstanding delta migrations to run.
+        """
+        db_pool = self.hs.get_datastore().db_pool
+        db_conn = LoggingDatabaseConnection(
+            db_pool._db_pool.connect(),
+            db_pool.engine,
+            "tests",
+        )
+
+        # Set the schema version of the database to the current version
+        cur = db_conn.cursor()
+        cur.execute("UPDATE schema_version SET version = ?", (SCHEMA_VERSION,))
+
+        db_conn.commit()
+
+        # Path `os.listdir` here to make synapse think that there is a migration
+        # file ready to be run.
+        # Note that we can't patch this function for the whole method, else Synapse
+        # will try to find the file when building the database initially.
+        with mock.patch("os.listdir", mock.Mock(side_effect=fake_listdir)):
+            with self.assertRaises(PrepareDatabaseException):
+                # Synapse should think that there is an outstanding migration file due to
+                # patching 'os.listdir' in the function decorator.
+                #
+                # We expect Synapse to raise an exception to indicate the master process
+                # needs to apply this migration file.
+                prepare_database(db_conn, db_pool.engine, self.hs.config)