Fix sync waiting for an invalid token from the "future" (#17386)

Fixes https://github.com/element-hq/synapse/issues/17274, hopefully.

Basically, old versions of Synapse could advance streams without
persisting anything in the DB (fixed in #17229). On restart those
updates would get lost, and so the position of the stream would revert
to an older position. If this happened across an upgrade to a later
Synapse version which included #17215, then sync could get blocked
indefinitely (until the stream advanced to the position in the token).

We fix this by bounding the stream positions we'll wait for to the
maximum position of the underlying stream ID generator.

changelog.d/17386.bugfix

@@ -0,0 +1 @@
+Fix bug where `/sync` requests could get blocked indefinitely after an upgrade from Synapse versions before v1.109.0.

synapse/notifier.py

@@ -764,6 +764,13 @@ class Notifier:
 
     async def wait_for_stream_token(self, stream_token: StreamToken) -> bool:
         """Wait for this worker to catch up with the given stream token."""
+        current_token = self.event_sources.get_current_token()
+        if stream_token.is_before_or_eq(current_token):
+            return True
+
+        # Work around a bug where older Synapse versions gave out tokens "from
+        # the future", i.e. that are ahead of the tokens persisted in the DB.
+        stream_token = await self.event_sources.bound_future_token(stream_token)
 
         start = self.clock.time_msec()
         while True:

synapse/storage/databases/main/account_data.py

@@ -43,10 +43,7 @@ from synapse.storage.database import (
 )
 from synapse.storage.databases.main.cache import CacheInvalidationWorkerStore
 from synapse.storage.databases.main.push_rule import PushRulesWorkerStore
-from synapse.storage.util.id_generators import (
+from synapse.storage.util.id_generators import MultiWriterIdGenerator
-    AbstractStreamIdGenerator,
-    MultiWriterIdGenerator,
-)
 from synapse.types import JsonDict, JsonMapping
 from synapse.util import json_encoder
 from synapse.util.caches.descriptors import cached
@@ -71,7 +68,7 @@ class AccountDataWorkerStore(PushRulesWorkerStore, CacheInvalidationWorkerStore)
             self._instance_name in hs.config.worker.writers.account_data
         )
 
-        self._account_data_id_gen: AbstractStreamIdGenerator
+        self._account_data_id_gen: MultiWriterIdGenerator
 
         self._account_data_id_gen = MultiWriterIdGenerator(
             db_conn=db_conn,
@@ -113,6 +110,9 @@ class AccountDataWorkerStore(PushRulesWorkerStore, CacheInvalidationWorkerStore)
         """
         return self._account_data_id_gen.get_current_token()
 
+    def get_account_data_id_generator(self) -> MultiWriterIdGenerator:
+        return self._account_data_id_gen
+
     @cached()
     async def get_global_account_data_for_user(
         self, user_id: str

synapse/storage/databases/main/deviceinbox.py

@@ -50,10 +50,7 @@ from synapse.storage.database import (
     LoggingTransaction,
     make_in_list_sql_clause,
 )
-from synapse.storage.util.id_generators import (
+from synapse.storage.util.id_generators import MultiWriterIdGenerator
-    AbstractStreamIdGenerator,
-    MultiWriterIdGenerator,
-)
 from synapse.types import JsonDict
 from synapse.util import json_encoder
 from synapse.util.caches.expiringcache import ExpiringCache
@@ -92,7 +89,7 @@ class DeviceInboxWorkerStore(SQLBaseStore):
             self._instance_name in hs.config.worker.writers.to_device
         )
 
-        self._to_device_msg_id_gen: AbstractStreamIdGenerator = MultiWriterIdGenerator(
+        self._to_device_msg_id_gen: MultiWriterIdGenerator = MultiWriterIdGenerator(
             db_conn=db_conn,
             db=database,
             notifier=hs.get_replication_notifier(),
@@ -169,6 +166,9 @@ class DeviceInboxWorkerStore(SQLBaseStore):
     def get_to_device_stream_token(self) -> int:
         return self._to_device_msg_id_gen.get_current_token()
 
+    def get_to_device_id_generator(self) -> MultiWriterIdGenerator:
+        return self._to_device_msg_id_gen
+
     async def get_messages_for_user_devices(
         self,
         user_ids: Collection[str],

synapse/storage/databases/main/devices.py

@@ -243,6 +243,9 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore):
     def get_device_stream_token(self) -> int:
         return self._device_list_id_gen.get_current_token()
 
+    def get_device_stream_id_generator(self) -> MultiWriterIdGenerator:
+        return self._device_list_id_gen
+
     async def count_devices_by_users(
         self, user_ids: Optional[Collection[str]] = None
     ) -> int:

synapse/storage/databases/main/events_worker.py

@@ -192,8 +192,8 @@ class EventsWorkerStore(SQLBaseStore):
     ):
         super().__init__(database, db_conn, hs)
 
-        self._stream_id_gen: AbstractStreamIdGenerator
+        self._stream_id_gen: MultiWriterIdGenerator
-        self._backfill_id_gen: AbstractStreamIdGenerator
+        self._backfill_id_gen: MultiWriterIdGenerator
 
         self._stream_id_gen = MultiWriterIdGenerator(
             db_conn=db_conn,

synapse/storage/databases/main/presence.py

@@ -42,10 +42,7 @@ from synapse.storage.database import (
 from synapse.storage.databases.main.cache import CacheInvalidationWorkerStore
 from synapse.storage.engines._base import IsolationLevel
 from synapse.storage.types import Connection
-from synapse.storage.util.id_generators import (
+from synapse.storage.util.id_generators import MultiWriterIdGenerator
-    AbstractStreamIdGenerator,
-    MultiWriterIdGenerator,
-)
 from synapse.util.caches.descriptors import cached, cachedList
 from synapse.util.caches.stream_change_cache import StreamChangeCache
 from synapse.util.iterutils import batch_iter
@@ -83,7 +80,7 @@ class PresenceStore(PresenceBackgroundUpdateStore, CacheInvalidationWorkerStore)
         super().__init__(database, db_conn, hs)
 
         self._instance_name = hs.get_instance_name()
-        self._presence_id_gen: AbstractStreamIdGenerator
+        self._presence_id_gen: MultiWriterIdGenerator
 
         self._can_persist_presence = (
             self._instance_name in hs.config.worker.writers.presence
@@ -455,6 +452,9 @@ class PresenceStore(PresenceBackgroundUpdateStore, CacheInvalidationWorkerStore)
     def get_current_presence_token(self) -> int:
         return self._presence_id_gen.get_current_token()
 
+    def get_presence_stream_id_gen(self) -> MultiWriterIdGenerator:
+        return self._presence_id_gen
+
     def _get_active_presence(self, db_conn: Connection) -> List[UserPresenceState]:
         """Fetch non-offline presence from the database so that we can register
         the appropriate time outs.

synapse/storage/databases/main/push_rule.py

@@ -178,6 +178,9 @@ class PushRulesWorkerStore(
         """
         return self._push_rules_stream_id_gen.get_current_token()
 
+    def get_push_rules_stream_id_gen(self) -> MultiWriterIdGenerator:
+        return self._push_rules_stream_id_gen
+
     def process_replication_rows(
         self, stream_name: str, instance_name: str, token: int, rows: Iterable[Any]
     ) -> None:

synapse/storage/databases/main/receipts.py

@@ -45,10 +45,7 @@ from synapse.storage.database import (
     LoggingTransaction,
 )
 from synapse.storage.engines._base import IsolationLevel
-from synapse.storage.util.id_generators import (
+from synapse.storage.util.id_generators import MultiWriterIdGenerator
-    AbstractStreamIdGenerator,
-    MultiWriterIdGenerator,
-)
 from synapse.types import (
     JsonDict,
     JsonMapping,
@@ -76,7 +73,7 @@ class ReceiptsWorkerStore(SQLBaseStore):
 
         # In the worker store this is an ID tracker which we overwrite in the non-worker
         # class below that is used on the main process.
-        self._receipts_id_gen: AbstractStreamIdGenerator
+        self._receipts_id_gen: MultiWriterIdGenerator
 
         self._can_write_to_receipts = (
             self._instance_name in hs.config.worker.writers.receipts
@@ -136,6 +133,9 @@ class ReceiptsWorkerStore(SQLBaseStore):
     def get_receipt_stream_id_for_instance(self, instance_name: str) -> int:
         return self._receipts_id_gen.get_current_token_for_writer(instance_name)
 
+    def get_receipts_stream_id_gen(self) -> MultiWriterIdGenerator:
+        return self._receipts_id_gen
+
     def get_last_unthreaded_receipt_for_user_txn(
         self,
         txn: LoggingTransaction,

synapse/storage/databases/main/room.py

@@ -59,11 +59,7 @@ from synapse.storage.database import (
 )
 from synapse.storage.databases.main.cache import CacheInvalidationWorkerStore
 from synapse.storage.types import Cursor
-from synapse.storage.util.id_generators import (
+from synapse.storage.util.id_generators import IdGenerator, MultiWriterIdGenerator
-    AbstractStreamIdGenerator,
-    IdGenerator,
-    MultiWriterIdGenerator,
-)
 from synapse.types import JsonDict, RetentionPolicy, StrCollection, ThirdPartyInstanceID
 from synapse.util import json_encoder
 from synapse.util.caches.descriptors import cached, cachedList
@@ -151,7 +147,7 @@ class RoomWorkerStore(CacheInvalidationWorkerStore):
 
         self.config: HomeServerConfig = hs.config
 
-        self._un_partial_stated_rooms_stream_id_gen: AbstractStreamIdGenerator
+        self._un_partial_stated_rooms_stream_id_gen: MultiWriterIdGenerator
 
         self._un_partial_stated_rooms_stream_id_gen = MultiWriterIdGenerator(
             db_conn=db_conn,
@@ -1409,6 +1405,9 @@ class RoomWorkerStore(CacheInvalidationWorkerStore):
             instance_name
         )
 
+    def get_un_partial_stated_rooms_id_generator(self) -> MultiWriterIdGenerator:
+        return self._un_partial_stated_rooms_stream_id_gen
+
     async def get_un_partial_stated_rooms_between(
         self, last_id: int, current_id: int, room_ids: Collection[str]
     ) -> Set[str]:

synapse/storage/databases/main/stream.py

@@ -577,6 +577,9 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore):
 
         return RoomStreamToken(stream=min_pos, instance_map=immutabledict(positions))
 
+    def get_events_stream_id_generator(self) -> MultiWriterIdGenerator:
+        return self._stream_id_gen
+
     async def get_room_events_stream_for_rooms(
         self,
         room_ids: Collection[str],

synapse/storage/util/id_generators.py

@@ -812,6 +812,11 @@ class MultiWriterIdGenerator(AbstractStreamIdGenerator):
         pos = self.get_current_token_for_writer(self._instance_name)
         txn.execute(sql, (self._stream_name, self._instance_name, pos))
 
+    async def get_max_allocated_token(self) -> int:
+        return await self._db.runInteraction(
+            "get_max_allocated_token", self._sequence_gen.get_max_allocated
+        )
+
 
 @attr.s(frozen=True, auto_attribs=True)
 class _AsyncCtxManagerWrapper(Generic[T]):

synapse/storage/util/sequence.py

@@ -88,6 +88,10 @@ class SequenceGenerator(metaclass=abc.ABCMeta):
         """
         ...
 
+    @abc.abstractmethod
+    def get_max_allocated(self, txn: Cursor) -> int:
+        """Get the maximum ID that we have allocated"""
+
 
 class PostgresSequenceGenerator(SequenceGenerator):
     """An implementation of SequenceGenerator which uses a postgres sequence"""
@@ -190,6 +194,17 @@ class PostgresSequenceGenerator(SequenceGenerator):
                 % {"seq": self._sequence_name, "stream_name": stream_name}
             )
 
+    def get_max_allocated(self, txn: Cursor) -> int:
+        # We just read from the sequence what the last value we fetched was.
+        txn.execute(f"SELECT last_value, is_called FROM {self._sequence_name}")
+        row = txn.fetchone()
+        assert row is not None
+
+        last_value, is_called = row
+        if not is_called:
+            last_value -= 1
+        return last_value
+
 
 GetFirstCallbackType = Callable[[Cursor], int]
 
@@ -248,6 +263,15 @@ class LocalSequenceGenerator(SequenceGenerator):
         # There is nothing to do for in memory sequences
         pass
 
+    def get_max_allocated(self, txn: Cursor) -> int:
+        with self._lock:
+            if self._current_max_id is None:
+                assert self._callback is not None
+                self._current_max_id = self._callback(txn)
+                self._callback = None
+
+            return self._current_max_id
+
 
 def build_sequence_generator(
     db_conn: "LoggingDatabaseConnection",

synapse/streams/events.py

@@ -30,7 +30,12 @@ from synapse.handlers.room import RoomEventSource
 from synapse.handlers.typing import TypingNotificationEventSource
 from synapse.logging.opentracing import trace
 from synapse.streams import EventSource
-from synapse.types import MultiWriterStreamToken, StreamKeyType, StreamToken
+from synapse.types import (
+    AbstractMultiWriterStreamToken,
+    MultiWriterStreamToken,
+    StreamKeyType,
+    StreamToken,
+)
 
 if TYPE_CHECKING:
     from synapse.server import HomeServer
@@ -91,6 +96,63 @@ class EventSources:
         )
         return token
 
+    async def bound_future_token(self, token: StreamToken) -> StreamToken:
+        """Bound a token that is ahead of the current token to the maximum
+        persisted values.
+
+        This ensures that if we wait for the given token we know the stream will
+        eventually advance to that point.
+
+        This works around a bug where older Synapse versions will give out
+        tokens for streams, and then after a restart will give back tokens where
+        the stream has "gone backwards".
+        """
+
+        current_token = self.get_current_token()
+
+        stream_key_to_id_gen = {
+            StreamKeyType.ROOM: self.store.get_events_stream_id_generator(),
+            StreamKeyType.PRESENCE: self.store.get_presence_stream_id_gen(),
+            StreamKeyType.RECEIPT: self.store.get_receipts_stream_id_gen(),
+            StreamKeyType.ACCOUNT_DATA: self.store.get_account_data_id_generator(),
+            StreamKeyType.PUSH_RULES: self.store.get_push_rules_stream_id_gen(),
+            StreamKeyType.TO_DEVICE: self.store.get_to_device_id_generator(),
+            StreamKeyType.DEVICE_LIST: self.store.get_device_stream_id_generator(),
+            StreamKeyType.UN_PARTIAL_STATED_ROOMS: self.store.get_un_partial_stated_rooms_id_generator(),
+        }
+
+        for _, key in StreamKeyType.__members__.items():
+            if key == StreamKeyType.TYPING:
+                # Typing stream is allowed to "reset", and so comparisons don't
+                # really make sense as is.
+                # TODO: Figure out a better way of tracking resets.
+                continue
+
+            token_value = token.get_field(key)
+            current_value = current_token.get_field(key)
+
+            if isinstance(token_value, AbstractMultiWriterStreamToken):
+                assert type(current_value) is type(token_value)
+
+                if not token_value.is_before_or_eq(current_value):  # type: ignore[arg-type]
+                    max_token = await stream_key_to_id_gen[
+                        key
+                    ].get_max_allocated_token()
+
+                    token = token.copy_and_replace(
+                        key, token.room_key.bound_stream_token(max_token)
+                    )
+            else:
+                assert isinstance(current_value, int)
+                if current_value < token_value:
+                    max_token = await stream_key_to_id_gen[
+                        key
+                    ].get_max_allocated_token()
+
+                    token = token.copy_and_replace(key, min(token_value, max_token))
+
+        return token
+
     @trace
     async def get_start_token_for_pagination(self, room_id: str) -> StreamToken:
         """Get the start token for a given room to be used to paginate

synapse/types/__init__.py

@@ -536,6 +536,16 @@ class AbstractMultiWriterStreamToken(metaclass=abc.ABCMeta):
 
         return True
 
+    def bound_stream_token(self, max_stream: int) -> "Self":
+        """Bound the stream positions to a maximum value"""
+
+        return type(self)(
+            stream=min(self.stream, max_stream),
+            instance_map=immutabledict(
+                {k: min(s, max_stream) for k, s in self.instance_map.items()}
+            ),
+        )
+
 
 @attr.s(frozen=True, slots=True, order=False)
 class RoomStreamToken(AbstractMultiWriterStreamToken):
@@ -722,6 +732,14 @@ class RoomStreamToken(AbstractMultiWriterStreamToken):
         else:
             return "s%d" % (self.stream,)
 
+    def bound_stream_token(self, max_stream: int) -> "RoomStreamToken":
+        """See super class"""
+
+        # This only makes sense for stream tokens.
+        assert self.topological is None
+
+        return super().bound_stream_token(max_stream)
+
 
 @attr.s(frozen=True, slots=True, order=False)
 class MultiWriterStreamToken(AbstractMultiWriterStreamToken):

tests/handlers/test_sync.py

@@ -22,6 +22,7 @@ from unittest.mock import AsyncMock, Mock, patch
 
 from parameterized import parameterized
 
+from twisted.internet import defer
 from twisted.test.proto_helpers import MemoryReactor
 
 from synapse.api.constants import AccountDataTypes, EventTypes, JoinRules
@@ -35,7 +36,7 @@ from synapse.handlers.sync import SyncConfig, SyncRequestKey, SyncResult, SyncVe
 from synapse.rest import admin
 from synapse.rest.client import knock, login, room
 from synapse.server import HomeServer
-from synapse.types import JsonDict, UserID, create_requester
+from synapse.types import JsonDict, StreamKeyType, UserID, create_requester
 from synapse.util import Clock
 
 import tests.unittest
@@ -959,6 +960,76 @@ class SyncTestCase(tests.unittest.HomeserverTestCase):
 
         self.fail("No push rules found")
 
+    def test_wait_for_future_sync_token(self) -> None:
+        """Test that if we receive a token that is ahead of our current token,
+        we'll wait until the stream position advances.
+
+        This can happen if replication streams start lagging, and the client's
+        previous sync request was serviced by a worker ahead of ours.
+        """
+        user = self.register_user("alice", "password")
+
+        # We simulate a lagging stream by getting a stream ID from the ID gen
+        # and then waiting to mark it as "persisted".
+        presence_id_gen = self.store.get_presence_stream_id_gen()
+        ctx_mgr = presence_id_gen.get_next()
+        stream_id = self.get_success(ctx_mgr.__aenter__())
+
+        # Create the new token based on the stream ID above.
+        current_token = self.hs.get_event_sources().get_current_token()
+        since_token = current_token.copy_and_advance(StreamKeyType.PRESENCE, stream_id)
+
+        sync_d = defer.ensureDeferred(
+            self.sync_handler.wait_for_sync_for_user(
+                create_requester(user),
+                generate_sync_config(user),
+                sync_version=SyncVersion.SYNC_V2,
+                request_key=generate_request_key(),
+                since_token=since_token,
+                timeout=0,
+            )
+        )
+
+        # This should block waiting for the presence stream to update
+        self.pump()
+        self.assertFalse(sync_d.called)
+
+        # Marking the stream ID as persisted should unblock the request.
+        self.get_success(ctx_mgr.__aexit__(None, None, None))
+
+        self.get_success(sync_d, by=1.0)
+
+    def test_wait_for_invalid_future_sync_token(self) -> None:
+        """Like the previous test, except we give a token that has a stream
+        position ahead of what is in the DB, i.e. its invalid and we shouldn't
+        wait for the stream to advance (as it may never do so).
+
+        This can happen due to older versions of Synapse giving out stream
+        positions without persisting them in the DB, and so on restart the
+        stream would get reset back to an older position.
+        """
+        user = self.register_user("alice", "password")
+
+        # Create a token and arbitrarily advance one of the streams.
+        current_token = self.hs.get_event_sources().get_current_token()
+        since_token = current_token.copy_and_advance(
+            StreamKeyType.PRESENCE, current_token.presence_key + 1
+        )
+
+        sync_d = defer.ensureDeferred(
+            self.sync_handler.wait_for_sync_for_user(
+                create_requester(user),
+                generate_sync_config(user),
+                sync_version=SyncVersion.SYNC_V2,
+                request_key=generate_request_key(),
+                since_token=since_token,
+                timeout=0,
+            )
+        )
+
+        # We should return without waiting for the presence stream to advance.
+        self.get_success(sync_d)
+
 
 def generate_sync_config(
     user_id: str,

tests/rest/client/test_sync.py

@@ -1386,10 +1386,12 @@ class SlidingSyncTestCase(unittest.HomeserverTestCase):
         # Create a future token that will cause us to wait. Since we never send a new
         # event to reach that future stream_ordering, the worker will wait until the
         # full timeout.
+        stream_id_gen = self.store.get_events_stream_id_generator()
+        stream_id = self.get_success(stream_id_gen.get_next().__aenter__())
         current_token = self.event_sources.get_current_token()
         future_position_token = current_token.copy_and_replace(
             StreamKeyType.ROOM,
-            RoomStreamToken(stream=current_token.room_key.stream + 1),
+            RoomStreamToken(stream=stream_id),
         )
 
         future_position_token_serialized = self.get_success(