Update all stream IDs after processing replication rows (#14723)

This creates a new store method, `process_replication_position` that
is called after `process_replication_rows`. By moving stream ID advances
here this guarantees any relevant cache invalidations will have been
applied before the stream is advanced.

This avoids race conditions where Python switches between threads mid
way through processing the `process_replication_rows` method where stream
IDs may be advanced before caches are invalidated due to class resolution
ordering.

See this comment/issue for further discussion:
	https://github.com/matrix-org/synapse/issues/14158#issuecomment-1344048703

1 files changed, 16 insertions, 1 deletions

diff --git a/synapse/storage/_base.py b/synapse/storage/_base.py
index 69abf6fa87..41d9111019 100644
--- a/synapse/storage/_base.py
+++ b/synapse/storage/_base.py
@@ -57,7 +57,22 @@ class SQLBaseStore(metaclass=ABCMeta):
         token: int,
         rows: Iterable[Any],
     ) -> None:
-        pass
+        """
+        Used by storage classes to invalidate caches based on incoming replication data. These
+        must not update any ID generators, use `process_replication_position`.
+        """
+
+    def process_replication_position(  # noqa: B027 (no-op by design)
+        self,
+        stream_name: str,
+        instance_name: str,
+        token: int,
+    ) -> None:
+        """
+        Used by storage classes to advance ID generators based on incoming replication data. This
+        is called after process_replication_rows such that caches are invalidated before any token
+        positions advance.
+        """
 
     def _invalidate_state_caches(
         self, room_id: str, members_changed: Collection[str]