From 03619324fc18632a2907ace4d3e73f3c4dd0b05e Mon Sep 17 00:00:00 2001 From: Richard van der Hoff <1389908+richvdh@users.noreply.github.com> Date: Tue, 16 Jun 2020 12:44:07 +0100 Subject: Create a ListenerConfig object (#7681) This ended up being a bit more invasive than I'd hoped for (not helped by generic_worker duplicating some of the code from homeserver), but hopefully it's an improvement. The idea is that, rather than storing unstructured `dict`s in the config for the listener configurations, we instead parse it into a structured `ListenerConfig` object. --- synapse/config/workers.py | 24 +++++++----------------- 1 file changed, 7 insertions(+), 17 deletions(-) (limited to 'synapse/config/workers.py') diff --git a/synapse/config/workers.py b/synapse/config/workers.py index ed06b91a54..dbc661630c 100644 --- a/synapse/config/workers.py +++ b/synapse/config/workers.py @@ -16,6 +16,7 @@ import attr from ._base import Config, ConfigError +from .server import ListenerConfig, parse_listener_def @attr.s @@ -52,7 +53,9 @@ class WorkerConfig(Config): if self.worker_app == "synapse.app.homeserver": self.worker_app = None - self.worker_listeners = config.get("worker_listeners", []) + self.worker_listeners = [ + parse_listener_def(x) for x in config.get("worker_listeners", []) + ] self.worker_daemonize = config.get("worker_daemonize") self.worker_pid_file = config.get("worker_pid_file") self.worker_log_config = config.get("worker_log_config") @@ -75,24 +78,11 @@ class WorkerConfig(Config): manhole = config.get("worker_manhole") if manhole: self.worker_listeners.append( - { - "port": manhole, - "bind_addresses": ["127.0.0.1"], - "type": "manhole", - "tls": False, - } + ListenerConfig( + port=manhole, bind_addresses=["127.0.0.1"], type="manhole", + ) ) - if self.worker_listeners: - for listener in self.worker_listeners: - bind_address = listener.pop("bind_address", None) - bind_addresses = listener.setdefault("bind_addresses", []) - - if bind_address: - bind_addresses.append(bind_address) - elif not bind_addresses: - bind_addresses.append("") - # A map from instance name to host/port of their HTTP replication endpoint. instance_map = config.get("instance_map") or {} self.instance_map = { -- cgit 1.5.1 From f2e38ca86711a8f80cf45d3182e426ed8967fc81 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Thu, 16 Jul 2020 15:12:54 +0100 Subject: Allow moving typing off master (#7869) --- changelog.d/7869.feature | 1 + synapse/app/generic_worker.py | 36 +---- synapse/config/workers.py | 19 +-- synapse/federation/federation_server.py | 125 +++++++++------- synapse/handlers/typing.py | 241 +++++++++++++++++++++---------- synapse/replication/tcp/handler.py | 9 ++ synapse/replication/tcp/streams/_base.py | 7 +- synapse/rest/client/v1/room.py | 9 ++ synapse/server.py | 13 +- synapse/server.pyi | 2 + 10 files changed, 284 insertions(+), 178 deletions(-) create mode 100644 changelog.d/7869.feature (limited to 'synapse/config/workers.py') diff --git a/changelog.d/7869.feature b/changelog.d/7869.feature new file mode 100644 index 0000000000..1982049a52 --- /dev/null +++ b/changelog.d/7869.feature @@ -0,0 +1 @@ +Add experimental support for moving typing off master. diff --git a/synapse/app/generic_worker.py b/synapse/app/generic_worker.py index e90695f026..c0853eef22 100644 --- a/synapse/app/generic_worker.py +++ b/synapse/app/generic_worker.py @@ -111,6 +111,7 @@ from synapse.rest.client.v1.room import ( RoomSendEventRestServlet, RoomStateEventRestServlet, RoomStateRestServlet, + RoomTypingRestServlet, ) from synapse.rest.client.v1.voip import VoipRestServlet from synapse.rest.client.v2_alpha import groups, sync, user_directory @@ -451,37 +452,6 @@ class GenericWorkerPresence(BasePresenceHandler): await self._bump_active_client(user_id=user_id) -class GenericWorkerTyping(object): - def __init__(self, hs): - self._latest_room_serial = 0 - self._reset() - - def _reset(self): - """ - Reset the typing handler's data caches. - """ - # map room IDs to serial numbers - self._room_serials = {} - # map room IDs to sets of users currently typing - self._room_typing = {} - - def process_replication_rows(self, token, rows): - if self._latest_room_serial > token: - # The master has gone backwards. To prevent inconsistent data, just - # clear everything. - self._reset() - - # Set the latest serial token to whatever the server gave us. - self._latest_room_serial = token - - for row in rows: - self._room_serials[row.room_id] = token - self._room_typing[row.room_id] = row.user_ids - - def get_current_token(self) -> int: - return self._latest_room_serial - - class GenericWorkerSlavedStore( # FIXME(#3714): We need to add UserDirectoryStore as we write directly # rather than going via the correct worker. @@ -558,6 +528,7 @@ class GenericWorkerServer(HomeServer): KeyUploadServlet(self).register(resource) AccountDataServlet(self).register(resource) RoomAccountDataServlet(self).register(resource) + RoomTypingRestServlet(self).register(resource) sync.register_servlets(self, resource) events.register_servlets(self, resource) @@ -669,9 +640,6 @@ class GenericWorkerServer(HomeServer): def build_presence_handler(self): return GenericWorkerPresence(self) - def build_typing_handler(self): - return GenericWorkerTyping(self) - class GenericWorkerReplicationHandler(ReplicationDataHandler): def __init__(self, hs): diff --git a/synapse/config/workers.py b/synapse/config/workers.py index dbc661630c..2574cd3aa1 100644 --- a/synapse/config/workers.py +++ b/synapse/config/workers.py @@ -34,9 +34,11 @@ class WriterLocations: Attributes: events: The instance that writes to the event and backfill streams. + events: The instance that writes to the typing stream. """ events = attr.ib(default="master", type=str) + typing = attr.ib(default="master", type=str) class WorkerConfig(Config): @@ -93,16 +95,15 @@ class WorkerConfig(Config): writers = config.get("stream_writers") or {} self.writers = WriterLocations(**writers) - # Check that the configured writer for events also appears in + # Check that the configured writer for events and typing also appears in # `instance_map`. - if ( - self.writers.events != "master" - and self.writers.events not in self.instance_map - ): - raise ConfigError( - "Instance %r is configured to write events but does not appear in `instance_map` config." - % (self.writers.events,) - ) + for stream in ("events", "typing"): + instance = getattr(self.writers, stream) + if instance != "master" and instance not in self.instance_map: + raise ConfigError( + "Instance %r is configured to write %s but does not appear in `instance_map` config." + % (instance, stream) + ) def read_arguments(self, args): # We support a bunch of command line arguments that override options in diff --git a/synapse/federation/federation_server.py b/synapse/federation/federation_server.py index 8c53330c49..23625ba995 100644 --- a/synapse/federation/federation_server.py +++ b/synapse/federation/federation_server.py @@ -15,7 +15,18 @@ # See the License for the specific language governing permissions and # limitations under the License. import logging -from typing import Any, Callable, Dict, List, Match, Optional, Tuple, Union +from typing import ( + TYPE_CHECKING, + Any, + Awaitable, + Callable, + Dict, + List, + Match, + Optional, + Tuple, + Union, +) from canonicaljson import json from prometheus_client import Counter, Histogram @@ -56,6 +67,9 @@ from synapse.util import glob_to_regex, unwrapFirstError from synapse.util.async_helpers import Linearizer, concurrently_execute from synapse.util.caches.response_cache import ResponseCache +if TYPE_CHECKING: + from synapse.server import HomeServer + # when processing incoming transactions, we try to handle multiple rooms in # parallel, up to this limit. TRANSACTION_CONCURRENCY_LIMIT = 10 @@ -768,11 +782,30 @@ class FederationHandlerRegistry(object): query type for incoming federation traffic. """ - def __init__(self): - self.edu_handlers = {} - self.query_handlers = {} + def __init__(self, hs: "HomeServer"): + self.config = hs.config + self.http_client = hs.get_simple_http_client() + self.clock = hs.get_clock() + self._instance_name = hs.get_instance_name() - def register_edu_handler(self, edu_type: str, handler: Callable[[str, dict], None]): + # These are safe to load in monolith mode, but will explode if we try + # and use them. However we have guards before we use them to ensure that + # we don't route to ourselves, and in monolith mode that will always be + # the case. + self._get_query_client = ReplicationGetQueryRestServlet.make_client(hs) + self._send_edu = ReplicationFederationSendEduRestServlet.make_client(hs) + + self.edu_handlers = ( + {} + ) # type: Dict[str, Callable[[str, dict], Awaitable[None]]] + self.query_handlers = {} # type: Dict[str, Callable[[dict], Awaitable[None]]] + + # Map from type to instance name that we should route EDU handling to. + self._edu_type_to_instance = {} # type: Dict[str, str] + + def register_edu_handler( + self, edu_type: str, handler: Callable[[str, dict], Awaitable[None]] + ): """Sets the handler callable that will be used to handle an incoming federation EDU of the given type. @@ -809,66 +842,56 @@ class FederationHandlerRegistry(object): self.query_handlers[query_type] = handler + def register_instance_for_edu(self, edu_type: str, instance_name: str): + """Register that the EDU handler is on a different instance than master. + """ + self._edu_type_to_instance[edu_type] = instance_name + async def on_edu(self, edu_type: str, origin: str, content: dict): + if not self.config.use_presence and edu_type == "m.presence": + return + + # Check if we have a handler on this instance handler = self.edu_handlers.get(edu_type) - if not handler: - logger.warning("No handler registered for EDU type %s", edu_type) + if handler: + with start_active_span_from_edu(content, "handle_edu"): + try: + await handler(origin, content) + except SynapseError as e: + logger.info("Failed to handle edu %r: %r", edu_type, e) + except Exception: + logger.exception("Failed to handle edu %r", edu_type) return - with start_active_span_from_edu(content, "handle_edu"): + # Check if we can route it somewhere else that isn't us + route_to = self._edu_type_to_instance.get(edu_type, "master") + if route_to != self._instance_name: try: - await handler(origin, content) + await self._send_edu( + instance_name=route_to, + edu_type=edu_type, + origin=origin, + content=content, + ) except SynapseError as e: logger.info("Failed to handle edu %r: %r", edu_type, e) except Exception: logger.exception("Failed to handle edu %r", edu_type) - - def on_query(self, query_type: str, args: dict) -> defer.Deferred: - handler = self.query_handlers.get(query_type) - if not handler: - logger.warning("No handler registered for query type %s", query_type) - raise NotFoundError("No handler for Query type '%s'" % (query_type,)) - - return handler(args) - - -class ReplicationFederationHandlerRegistry(FederationHandlerRegistry): - """A FederationHandlerRegistry for worker processes. - - When receiving EDU or queries it will check if an appropriate handler has - been registered on the worker, if there isn't one then it calls off to the - master process. - """ - - def __init__(self, hs): - self.config = hs.config - self.http_client = hs.get_simple_http_client() - self.clock = hs.get_clock() - - self._get_query_client = ReplicationGetQueryRestServlet.make_client(hs) - self._send_edu = ReplicationFederationSendEduRestServlet.make_client(hs) - - super(ReplicationFederationHandlerRegistry, self).__init__() - - async def on_edu(self, edu_type: str, origin: str, content: dict): - """Overrides FederationHandlerRegistry - """ - if not self.config.use_presence and edu_type == "m.presence": return - handler = self.edu_handlers.get(edu_type) - if handler: - return await super(ReplicationFederationHandlerRegistry, self).on_edu( - edu_type, origin, content - ) - - return await self._send_edu(edu_type=edu_type, origin=origin, content=content) + # Oh well, let's just log and move on. + logger.warning("No handler registered for EDU type %s", edu_type) async def on_query(self, query_type: str, args: dict): - """Overrides FederationHandlerRegistry - """ handler = self.query_handlers.get(query_type) if handler: return await handler(args) - return await self._get_query_client(query_type=query_type, args=args) + # Check if we can route it somewhere else that isn't us + if self._instance_name == "master": + return await self._get_query_client(query_type=query_type, args=args) + + # Uh oh, no handler! Let's raise an exception so the request returns an + # error. + logger.warning("No handler registered for query type %s", query_type) + raise NotFoundError("No handler for Query type '%s'" % (query_type,)) diff --git a/synapse/handlers/typing.py b/synapse/handlers/typing.py index 846ddbdc6c..a86ac0150e 100644 --- a/synapse/handlers/typing.py +++ b/synapse/handlers/typing.py @@ -15,15 +15,19 @@ import logging from collections import namedtuple -from typing import List, Tuple +from typing import TYPE_CHECKING, List, Set, Tuple from synapse.api.errors import AuthError, SynapseError -from synapse.logging.context import run_in_background +from synapse.metrics.background_process_metrics import run_as_background_process +from synapse.replication.tcp.streams import TypingStream from synapse.types import UserID, get_domain_from_id from synapse.util.caches.stream_change_cache import StreamChangeCache from synapse.util.metrics import Measure from synapse.util.wheel_timer import WheelTimer +if TYPE_CHECKING: + from synapse.server import HomeServer + logger = logging.getLogger(__name__) @@ -39,48 +43,48 @@ FEDERATION_TIMEOUT = 60 * 1000 FEDERATION_PING_INTERVAL = 40 * 1000 -class TypingHandler(object): - def __init__(self, hs): +class FollowerTypingHandler: + """A typing handler on a different process than the writer that is updated + via replication. + """ + + def __init__(self, hs: "HomeServer"): self.store = hs.get_datastore() self.server_name = hs.config.server_name - self.auth = hs.get_auth() - self.is_mine_id = hs.is_mine_id - self.notifier = hs.get_notifier() - self.state = hs.get_state_handler() - - self.hs = hs - self.clock = hs.get_clock() - self.wheel_timer = WheelTimer(bucket_size=5000) + self.is_mine_id = hs.is_mine_id - self.federation = hs.get_federation_sender() + self.federation = None + if hs.should_send_federation(): + self.federation = hs.get_federation_sender() - hs.get_federation_registry().register_edu_handler("m.typing", self._recv_edu) + if hs.config.worker.writers.typing != hs.get_instance_name(): + hs.get_federation_registry().register_instance_for_edu( + "m.typing", hs.config.worker.writers.typing, + ) - hs.get_distributor().observe("user_left_room", self.user_left_room) + # map room IDs to serial numbers + self._room_serials = {} + # map room IDs to sets of users currently typing + self._room_typing = {} - self._member_typing_until = {} # clock time we expect to stop self._member_last_federation_poke = {} - + self.wheel_timer = WheelTimer(bucket_size=5000) self._latest_room_serial = 0 - self._reset() - - # caches which room_ids changed at which serials - self._typing_stream_change_cache = StreamChangeCache( - "TypingStreamChangeCache", self._latest_room_serial - ) self.clock.looping_call(self._handle_timeouts, 5000) def _reset(self): - """ - Reset the typing handler's data caches. + """Reset the typing handler's data caches. """ # map room IDs to serial numbers self._room_serials = {} # map room IDs to sets of users currently typing self._room_typing = {} + self._member_last_federation_poke = {} + self.wheel_timer = WheelTimer(bucket_size=5000) + def _handle_timeouts(self): logger.debug("Checking for typing timeouts") @@ -89,30 +93,140 @@ class TypingHandler(object): members = set(self.wheel_timer.fetch(now)) for member in members: - if not self.is_typing(member): - # Nothing to do if they're no longer typing - continue - - until = self._member_typing_until.get(member, None) - if not until or until <= now: - logger.info("Timing out typing for: %s", member.user_id) - self._stopped_typing(member) - continue - - # Check if we need to resend a keep alive over federation for this - # user. - if self.hs.is_mine_id(member.user_id): - last_fed_poke = self._member_last_federation_poke.get(member, None) - if not last_fed_poke or last_fed_poke + FEDERATION_PING_INTERVAL <= now: - run_in_background(self._push_remote, member=member, typing=True) - - # Add a paranoia timer to ensure that we always have a timer for - # each person typing. - self.wheel_timer.insert(now=now, obj=member, then=now + 60 * 1000) + self._handle_timeout_for_member(now, member) + + def _handle_timeout_for_member(self, now: int, member: RoomMember): + if not self.is_typing(member): + # Nothing to do if they're no longer typing + return + + # Check if we need to resend a keep alive over federation for this + # user. + if self.federation and self.is_mine_id(member.user_id): + last_fed_poke = self._member_last_federation_poke.get(member, None) + if not last_fed_poke or last_fed_poke + FEDERATION_PING_INTERVAL <= now: + run_as_background_process( + "typing._push_remote", self._push_remote, member=member, typing=True + ) + + # Add a paranoia timer to ensure that we always have a timer for + # each person typing. + self.wheel_timer.insert(now=now, obj=member, then=now + 60 * 1000) def is_typing(self, member): return member.user_id in self._room_typing.get(member.room_id, []) + async def _push_remote(self, member, typing): + if not self.federation: + return + + try: + users = await self.store.get_users_in_room(member.room_id) + self._member_last_federation_poke[member] = self.clock.time_msec() + + now = self.clock.time_msec() + self.wheel_timer.insert( + now=now, obj=member, then=now + FEDERATION_PING_INTERVAL + ) + + for domain in {get_domain_from_id(u) for u in users}: + if domain != self.server_name: + logger.debug("sending typing update to %s", domain) + self.federation.build_and_send_edu( + destination=domain, + edu_type="m.typing", + content={ + "room_id": member.room_id, + "user_id": member.user_id, + "typing": typing, + }, + key=member, + ) + except Exception: + logger.exception("Error pushing typing notif to remotes") + + def process_replication_rows( + self, token: int, rows: List[TypingStream.TypingStreamRow] + ): + """Should be called whenever we receive updates for typing stream. + """ + + if self._latest_room_serial > token: + # The master has gone backwards. To prevent inconsistent data, just + # clear everything. + self._reset() + + # Set the latest serial token to whatever the server gave us. + self._latest_room_serial = token + + for row in rows: + self._room_serials[row.room_id] = token + + prev_typing = set(self._room_typing.get(row.room_id, [])) + now_typing = set(row.user_ids) + self._room_typing[row.room_id] = row.user_ids + + run_as_background_process( + "_handle_change_in_typing", + self._handle_change_in_typing, + row.room_id, + prev_typing, + now_typing, + ) + + async def _handle_change_in_typing( + self, room_id: str, prev_typing: Set[str], now_typing: Set[str] + ): + """Process a change in typing of a room from replication, sending EDUs + for any local users. + """ + for user_id in now_typing - prev_typing: + if self.is_mine_id(user_id): + await self._push_remote(RoomMember(room_id, user_id), True) + + for user_id in prev_typing - now_typing: + if self.is_mine_id(user_id): + await self._push_remote(RoomMember(room_id, user_id), False) + + def get_current_token(self): + return self._latest_room_serial + + +class TypingWriterHandler(FollowerTypingHandler): + def __init__(self, hs): + super().__init__(hs) + + assert hs.config.worker.writers.typing == hs.get_instance_name() + + self.auth = hs.get_auth() + self.notifier = hs.get_notifier() + + self.hs = hs + + hs.get_federation_registry().register_edu_handler("m.typing", self._recv_edu) + + hs.get_distributor().observe("user_left_room", self.user_left_room) + + self._member_typing_until = {} # clock time we expect to stop + + # caches which room_ids changed at which serials + self._typing_stream_change_cache = StreamChangeCache( + "TypingStreamChangeCache", self._latest_room_serial + ) + + def _handle_timeout_for_member(self, now: int, member: RoomMember): + super()._handle_timeout_for_member(now, member) + + if not self.is_typing(member): + # Nothing to do if they're no longer typing + return + + until = self._member_typing_until.get(member, None) + if not until or until <= now: + logger.info("Timing out typing for: %s", member.user_id) + self._stopped_typing(member) + return + async def started_typing(self, target_user, auth_user, room_id, timeout): target_user_id = target_user.to_string() auth_user_id = auth_user.to_string() @@ -179,35 +293,11 @@ class TypingHandler(object): def _push_update(self, member, typing): if self.hs.is_mine_id(member.user_id): # Only send updates for changes to our own users. - run_in_background(self._push_remote, member, typing) - - self._push_update_local(member=member, typing=typing) - - async def _push_remote(self, member, typing): - try: - users = await self.store.get_users_in_room(member.room_id) - self._member_last_federation_poke[member] = self.clock.time_msec() - - now = self.clock.time_msec() - self.wheel_timer.insert( - now=now, obj=member, then=now + FEDERATION_PING_INTERVAL + run_as_background_process( + "typing._push_remote", self._push_remote, member, typing ) - for domain in {get_domain_from_id(u) for u in users}: - if domain != self.server_name: - logger.debug("sending typing update to %s", domain) - self.federation.build_and_send_edu( - destination=domain, - edu_type="m.typing", - content={ - "room_id": member.room_id, - "user_id": member.user_id, - "typing": typing, - }, - key=member, - ) - except Exception: - logger.exception("Error pushing typing notif to remotes") + self._push_update_local(member=member, typing=typing) async def _recv_edu(self, origin, content): room_id = content["room_id"] @@ -304,8 +394,11 @@ class TypingHandler(object): return rows, current_id, limited - def get_current_token(self): - return self._latest_room_serial + def process_replication_rows( + self, token: int, rows: List[TypingStream.TypingStreamRow] + ): + # The writing process should never get updates from replication. + raise Exception("Typing writer instance got typing info over replication") class TypingNotificationEventSource(object): diff --git a/synapse/replication/tcp/handler.py b/synapse/replication/tcp/handler.py index 80f5df60f9..30d8de48fa 100644 --- a/synapse/replication/tcp/handler.py +++ b/synapse/replication/tcp/handler.py @@ -42,6 +42,7 @@ from synapse.replication.tcp.streams import ( EventsStream, FederationStream, Stream, + TypingStream, ) from synapse.util.async_helpers import Linearizer @@ -96,6 +97,14 @@ class ReplicationCommandHandler: continue + if isinstance(stream, TypingStream): + # Only add TypingStream as a source on the instance in charge of + # typing. + if hs.config.worker.writers.typing == hs.get_instance_name(): + self._streams_to_replicate.append(stream) + + continue + # Only add any other streams if we're on master. if hs.config.worker_app is not None: continue diff --git a/synapse/replication/tcp/streams/_base.py b/synapse/replication/tcp/streams/_base.py index 9076bbe9f1..7a42de3f7d 100644 --- a/synapse/replication/tcp/streams/_base.py +++ b/synapse/replication/tcp/streams/_base.py @@ -294,11 +294,12 @@ class TypingStream(Stream): def __init__(self, hs): typing_handler = hs.get_typing_handler() - if hs.config.worker_app is None: - # on the master, query the typing handler + writer_instance = hs.config.worker.writers.typing + if writer_instance == hs.get_instance_name(): + # On the writer, query the typing handler update_function = typing_handler.get_all_typing_updates else: - # Query master process + # Query the typing writer process update_function = make_http_update_function(hs, self.NAME) super().__init__( diff --git a/synapse/rest/client/v1/room.py b/synapse/rest/client/v1/room.py index ea5912d4e4..26d5a51cb2 100644 --- a/synapse/rest/client/v1/room.py +++ b/synapse/rest/client/v1/room.py @@ -817,9 +817,18 @@ class RoomTypingRestServlet(RestServlet): self.typing_handler = hs.get_typing_handler() self.auth = hs.get_auth() + # If we're not on the typing writer instance we should scream if we get + # requests. + self._is_typing_writer = ( + hs.config.worker.writers.typing == hs.get_instance_name() + ) + async def on_PUT(self, request, room_id, user_id): requester = await self.auth.get_user_by_req(request) + if not self._is_typing_writer: + raise Exception("Got /typing request on instance that is not typing writer") + room_id = urlparse.unquote(room_id) target_user = UserID.from_string(urlparse.unquote(user_id)) diff --git a/synapse/server.py b/synapse/server.py index 0e6ea96b33..8e41112530 100644 --- a/synapse/server.py +++ b/synapse/server.py @@ -44,7 +44,6 @@ from synapse.federation.federation_client import FederationClient from synapse.federation.federation_server import ( FederationHandlerRegistry, FederationServer, - ReplicationFederationHandlerRegistry, ) from synapse.federation.send_queue import FederationRemoteSendQueue from synapse.federation.sender import FederationSender @@ -84,7 +83,7 @@ from synapse.handlers.room_member_worker import RoomMemberWorkerHandler from synapse.handlers.set_password import SetPasswordHandler from synapse.handlers.stats import StatsHandler from synapse.handlers.sync import SyncHandler -from synapse.handlers.typing import TypingHandler +from synapse.handlers.typing import FollowerTypingHandler, TypingWriterHandler from synapse.handlers.user_directory import UserDirectoryHandler from synapse.http.client import InsecureInterceptableContextFactory, SimpleHttpClient from synapse.http.matrixfederationclient import MatrixFederationHttpClient @@ -378,7 +377,10 @@ class HomeServer(object): return PresenceHandler(self) def build_typing_handler(self): - return TypingHandler(self) + if self.config.worker.writers.typing == self.get_instance_name(): + return TypingWriterHandler(self) + else: + return FollowerTypingHandler(self) def build_sync_handler(self): return SyncHandler(self) @@ -534,10 +536,7 @@ class HomeServer(object): return RoomMemberMasterHandler(self) def build_federation_registry(self): - if self.config.worker_app: - return ReplicationFederationHandlerRegistry(self) - else: - return FederationHandlerRegistry() + return FederationHandlerRegistry(self) def build_server_notices_manager(self): if self.config.worker_app: diff --git a/synapse/server.pyi b/synapse/server.pyi index cd50c721b8..90a673778f 100644 --- a/synapse/server.pyi +++ b/synapse/server.pyi @@ -148,3 +148,5 @@ class HomeServer(object): self, ) -> synapse.http.matrixfederationclient.MatrixFederationHttpClient: pass + def should_send_federation(self) -> bool: + pass -- cgit 1.5.1 From 2c1b9d676322fad8cb57c92f97f81393bcfcbe56 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Wed, 29 Jul 2020 23:22:13 +0100 Subject: Update worker docs with recent enhancements (#7969) --- changelog.d/7969.doc | 1 + docs/sample_config.yaml | 54 +++ docs/synctl_workers.md | 32 ++ docs/workers.md | 459 +++++++++++---------- synapse/app/generic_worker.py | 6 +- synapse/config/federation.py | 12 +- synapse/config/homeserver.py | 2 +- synapse/config/logger.py | 2 +- synapse/config/redis.py | 23 +- synapse/config/workers.py | 49 ++- synapse/federation/send_queue.py | 2 +- synapse/federation/sender/__init__.py | 2 +- synapse/federation/sender/per_destination_queue.py | 2 +- synapse/storage/data_stores/main/stream.py | 2 +- 14 files changed, 413 insertions(+), 235 deletions(-) create mode 100644 changelog.d/7969.doc create mode 100644 docs/synctl_workers.md (limited to 'synapse/config/workers.py') diff --git a/changelog.d/7969.doc b/changelog.d/7969.doc new file mode 100644 index 0000000000..68d2ed5fad --- /dev/null +++ b/changelog.d/7969.doc @@ -0,0 +1 @@ +Update worker docs with latest enhancements. diff --git a/docs/sample_config.yaml b/docs/sample_config.yaml index 3227294e0b..b21e36bb6d 100644 --- a/docs/sample_config.yaml +++ b/docs/sample_config.yaml @@ -2398,3 +2398,57 @@ opentracing: # # logging: # false + + +## Workers ## + +# Disables sending of outbound federation transactions on the main process. +# Uncomment if using a federation sender worker. +# +#send_federation: false + +# It is possible to run multiple federation sender workers, in which case the +# work is balanced across them. +# +# This configuration must be shared between all federation sender workers, and if +# changed all federation sender workers must be stopped at the same time and then +# started, to ensure that all instances are running with the same config (otherwise +# events may be dropped). +# +#federation_sender_instances: +# - federation_sender1 + +# When using workers this should be a map from `worker_name` to the +# HTTP replication listener of the worker, if configured. +# +#instance_map: +# worker1: +# host: localhost +# port: 8034 + +# Experimental: When using workers you can define which workers should +# handle event persistence and typing notifications. Any worker +# specified here must also be in the `instance_map`. +# +#stream_writers: +# events: worker1 +# typing: worker1 + + +# Configuration for Redis when using workers. This *must* be enabled when +# using workers (unless using old style direct TCP configuration). +# +redis: + # Uncomment the below to enable Redis support. + # + #enabled: true + + # Optional host and port to use to connect to redis. Defaults to + # localhost and 6379 + # + #host: localhost + #port: 6379 + + # Optional password if configured on the Redis instance + # + #password: diff --git a/docs/synctl_workers.md b/docs/synctl_workers.md new file mode 100644 index 0000000000..8da4a31852 --- /dev/null +++ b/docs/synctl_workers.md @@ -0,0 +1,32 @@ +### Using synctl with workers + +If you want to use `synctl` to manage your synapse processes, you will need to +create an an additional configuration file for the main synapse process. That +configuration should look like this: + +```yaml +worker_app: synapse.app.homeserver +``` + +Additionally, each worker app must be configured with the name of a "pid file", +to which it will write its process ID when it starts. For example, for a +synchrotron, you might write: + +```yaml +worker_pid_file: /home/matrix/synapse/worker1.pid +``` + +Finally, to actually run your worker-based synapse, you must pass synctl the `-a` +commandline option to tell it to operate on all the worker configurations found +in the given directory, e.g.: + + synctl -a $CONFIG/workers start + +Currently one should always restart all workers when restarting or upgrading +synapse, unless you explicitly know it's safe not to. For instance, restarting +synapse without restarting all the synchrotrons may result in broken typing +notifications. + +To manipulate a specific worker, you pass the -w option to synctl: + + synctl -w $CONFIG/workers/worker1.yaml restart diff --git a/docs/workers.md b/docs/workers.md index f4cbbc0400..38bd758e57 100644 --- a/docs/workers.md +++ b/docs/workers.md @@ -16,69 +16,106 @@ workers only work with PostgreSQL-based Synapse deployments. SQLite should only be used for demo purposes and any admin considering workers should already be running PostgreSQL. -## Master/worker communication +## Main process/worker communication -The workers communicate with the master process via a Synapse-specific protocol -called 'replication' (analogous to MySQL- or Postgres-style database -replication) which feeds a stream of relevant data from the master to the -workers so they can be kept in sync with the master process and database state. +The processes communicate with each other via a Synapse-specific protocol called +'replication' (analogous to MySQL- or Postgres-style database replication) which +feeds streams of newly written data between processes so they can be kept in +sync with the database state. -Additionally, workers may make HTTP requests to the master, to send information -in the other direction. Typically this is used for operations which need to -wait for a reply - such as sending an event. +Additionally, processes may make HTTP requests to each other. Typically this is +used for operations which need to wait for a reply - such as sending an event. -## Configuration +As of Synapse v1.13.0, it is possible to configure Synapse to send replication +via a [Redis pub/sub channel](https://redis.io/topics/pubsub), and is now the +recommended way of configuring replication. This is an alternative to the old +direct TCP connections to the main process: rather than all the workers +connecting to the main process, all the workers and the main process connect to +Redis, which relays replication commands between processes. This can give a +significant cpu saving on the main process and will be a prerequisite for +upcoming performance improvements. + +(See the [Architectural diagram](#architectural-diagram) section at the end for +a visualisation of what this looks like) + + +## Setting up workers + +A Redis server is required to manage the communication between the processes. +(The older direct TCP connections are now deprecated.) The Redis server +should be installed following the normal procedure for your distribution (e.g. +`apt install redis-server` on Debian). It is safe to use an existing Redis +deployment if you have one. + +Once installed, check that Redis is running and accessible from the host running +Synapse, for example by executing `echo PING | nc -q1 localhost 6379` and seeing +a response of `+PONG`. + +The appropriate dependencies must also be installed for Synapse. If using a +virtualenv, these can be installed with: + +```sh +pip install matrix-synapse[redis] +``` + +Note that these dependencies are included when synapse is installed with `pip +install matrix-synapse[all]`. They are also included in the debian packages from +`matrix.org` and in the docker images at +https://hub.docker.com/r/matrixdotorg/synapse/. To make effective use of the workers, you will need to configure an HTTP reverse-proxy such as nginx or haproxy, which will direct incoming requests to -the correct worker, or to the main synapse instance. Note that this includes -requests made to the federation port. See [reverse_proxy.md](reverse_proxy.md) +the correct worker, or to the main synapse instance. See [reverse_proxy.md](reverse_proxy.md) for information on setting up a reverse proxy. -To enable workers, you need to add *two* replication listeners to the -main Synapse configuration file (`homeserver.yaml`). For example: +To enable workers you should create a configuration file for each worker +process. Each worker configuration file inherits the configuration of the shared +homeserver configuration file. You can then override configuration specific to +that worker, e.g. the HTTP listener that it provides (if any); logging +configuration; etc. You should minimise the number of overrides though to +maintain a usable config. + +Next you need to add both a HTTP replication listener and redis config to the +shared Synapse configuration file (`homeserver.yaml`). For example: ```yaml +# extend the existing `listeners` section. This defines the ports that the +# main process will listen on. listeners: - # The TCP replication port - - port: 9092 - bind_address: '127.0.0.1' - type: replication - # The HTTP replication port - port: 9093 bind_address: '127.0.0.1' type: http resources: - names: [replication] + +redis: + enabled: true ``` -Under **no circumstances** should these replication API listeners be exposed to -the public internet; they have no authentication and are unencrypted. +See the sample config for the full documentation of each option. -You should then create a set of configs for the various worker processes. Each -worker configuration file inherits the configuration of the main homeserver -configuration file. You can then override configuration specific to that -worker, e.g. the HTTP listener that it provides (if any); logging -configuration; etc. You should minimise the number of overrides though to -maintain a usable config. +Under **no circumstances** should the replication listener be exposed to the +public internet; it has no authentication and is unencrypted. In the config file for each worker, you must specify the type of worker -application (`worker_app`). The currently available worker applications are -listed below. You must also specify the replication endpoints that it should -talk to on the main synapse process. `worker_replication_host` should specify -the host of the main synapse, `worker_replication_port` should point to the TCP -replication listener port and `worker_replication_http_port` should point to -the HTTP replication port. +application (`worker_app`), and you should specify a unqiue name for the worker +(`worker_name`). The currently available worker applications are listed below. +You must also specify the HTTP replication endpoint that it should talk to on +the main synapse process. `worker_replication_host` should specify the host of +the main synapse and `worker_replication_http_port` should point to the HTTP +replication port. If the worker will handle HTTP requests then the +`worker_listeners` option should be set with a `http` listener, in the same way +as the `listeners` option in the shared config. For example: ```yaml -worker_app: synapse.app.synchrotron +worker_app: synapse.app.generic_worker +worker_name: worker1 -# The replication listener on the synapse to talk to. +# The replication listener on the main synapse process. worker_replication_host: 127.0.0.1 -worker_replication_port: 9092 worker_replication_http_port: 9093 worker_listeners: @@ -87,13 +124,14 @@ worker_listeners: resources: - names: - client + - federation -worker_log_config: /home/matrix/synapse/config/synchrotron_log_config.yaml +worker_log_config: /home/matrix/synapse/config/worker1_log_config.yaml ``` -...is a full configuration for a synchrotron worker instance, which will expose a -plain HTTP `/sync` endpoint on port 8083 separately from the `/sync` endpoint provided -by the main synapse. +...is a full configuration for a generic worker instance, which will expose a +plain HTTP endpoint on port 8083 separately serving various endpoints, e.g. +`/sync`, which are listed below. Obviously you should configure your reverse-proxy to route the relevant endpoints to the worker (`localhost:8083` in the above example). @@ -102,127 +140,24 @@ Finally, you need to start your worker processes. This can be done with either `synctl` or your distribution's preferred service manager such as `systemd`. We recommend the use of `systemd` where available: for information on setting up `systemd` to start synapse workers, see -[systemd-with-workers](systemd-with-workers). To use `synctl`, see below. +[systemd-with-workers](systemd-with-workers). To use `synctl`, see +[synctl_workers.md](synctl_workers.md). -### **Experimental** support for replication over redis - -As of Synapse v1.13.0, it is possible to configure Synapse to send replication -via a [Redis pub/sub channel](https://redis.io/topics/pubsub). This is an -alternative to direct TCP connections to the master: rather than all the -workers connecting to the master, all the workers and the master connect to -Redis, which relays replication commands between processes. This can give a -significant cpu saving on the master and will be a prerequisite for upcoming -performance improvements. - -Note that this support is currently experimental; you may experience lost -messages and similar problems! It is strongly recommended that admins setting -up workers for the first time use direct TCP replication as above. - -To configure Synapse to use Redis: - -1. Install Redis following the normal procedure for your distribution - for - example, on Debian, `apt install redis-server`. (It is safe to use an - existing Redis deployment if you have one: we use a pub/sub stream named - according to the `server_name` of your synapse server.) -2. Check Redis is running and accessible: you should be able to `echo PING | nc -q1 - localhost 6379` and get a response of `+PONG`. -3. Install the python prerequisites. If you installed synapse into a - virtualenv, this can be done with: - ```sh - pip install matrix-synapse[redis] - ``` - The debian packages from matrix.org already include the required - dependencies. -4. Add config to the shared configuration (`homeserver.yaml`): - ```yaml - redis: - enabled: true - ``` - Optional parameters which can go alongside `enabled` are `host`, `port`, - `password`. Normally none of these are required. -5. Restart master and all workers. - -Once redis replication is in use, `worker_replication_port` is redundant and -can be removed from the worker configuration files. Similarly, the -configuration for the `listener` for the TCP replication port can be removed -from the main configuration file. Note that the HTTP replication port is -still required. - -### Using synctl - -If you want to use `synctl` to manage your synapse processes, you will need to -create an an additional configuration file for the master synapse process. That -configuration should look like this: - -```yaml -worker_app: synapse.app.homeserver -``` - -Additionally, each worker app must be configured with the name of a "pid file", -to which it will write its process ID when it starts. For example, for a -synchrotron, you might write: - -```yaml -worker_pid_file: /home/matrix/synapse/synchrotron.pid -``` - -Finally, to actually run your worker-based synapse, you must pass synctl the `-a` -commandline option to tell it to operate on all the worker configurations found -in the given directory, e.g.: - - synctl -a $CONFIG/workers start - -Currently one should always restart all workers when restarting or upgrading -synapse, unless you explicitly know it's safe not to. For instance, restarting -synapse without restarting all the synchrotrons may result in broken typing -notifications. - -To manipulate a specific worker, you pass the -w option to synctl: - - synctl -w $CONFIG/workers/synchrotron.yaml restart ## Available worker applications -### `synapse.app.pusher` - -Handles sending push notifications to sygnal and email. Doesn't handle any -REST endpoints itself, but you should set `start_pushers: False` in the -shared configuration file to stop the main synapse sending these notifications. - -Note this worker cannot be load-balanced: only one instance should be active. - -### `synapse.app.synchrotron` +### `synapse.app.generic_worker` -The synchrotron handles `sync` requests from clients. In particular, it can -handle REST endpoints matching the following regular expressions: +This worker can handle API requests matching the following regular +expressions: + # Sync requests ^/_matrix/client/(v2_alpha|r0)/sync$ ^/_matrix/client/(api/v1|v2_alpha|r0)/events$ ^/_matrix/client/(api/v1|r0)/initialSync$ ^/_matrix/client/(api/v1|r0)/rooms/[^/]+/initialSync$ -The above endpoints should all be routed to the synchrotron worker by the -reverse-proxy configuration. - -It is possible to run multiple instances of the synchrotron to scale -horizontally. In this case the reverse-proxy should be configured to -load-balance across the instances, though it will be more efficient if all -requests from a particular user are routed to a single instance. Extracting -a userid from the access token is currently left as an exercise for the reader. - -### `synapse.app.appservice` - -Handles sending output traffic to Application Services. Doesn't handle any -REST endpoints itself, but you should set `notify_appservices: False` in the -shared configuration file to stop the main synapse sending these notifications. - -Note this worker cannot be load-balanced: only one instance should be active. - -### `synapse.app.federation_reader` - -Handles a subset of federation endpoints. In particular, it can handle REST -endpoints matching the following regular expressions: - + # Federation requests ^/_matrix/federation/v1/event/ ^/_matrix/federation/v1/state/ ^/_matrix/federation/v1/state_ids/ @@ -242,40 +177,145 @@ endpoints matching the following regular expressions: ^/_matrix/federation/v1/event_auth/ ^/_matrix/federation/v1/exchange_third_party_invite/ ^/_matrix/federation/v1/user/devices/ - ^/_matrix/federation/v1/send/ ^/_matrix/federation/v1/get_groups_publicised$ ^/_matrix/key/v2/query + # Inbound federation transaction request + ^/_matrix/federation/v1/send/ + + # Client API requests + ^/_matrix/client/(api/v1|r0|unstable)/publicRooms$ + ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/joined_members$ + ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/context/.*$ + ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/members$ + ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/state$ + ^/_matrix/client/(api/v1|r0|unstable)/account/3pid$ + ^/_matrix/client/(api/v1|r0|unstable)/keys/query$ + ^/_matrix/client/(api/v1|r0|unstable)/keys/changes$ + ^/_matrix/client/versions$ + ^/_matrix/client/(api/v1|r0|unstable)/voip/turnServer$ + ^/_matrix/client/(api/v1|r0|unstable)/joined_groups$ + ^/_matrix/client/(api/v1|r0|unstable)/publicised_groups$ + ^/_matrix/client/(api/v1|r0|unstable)/publicised_groups/ + + # Registration/login requests + ^/_matrix/client/(api/v1|r0|unstable)/login$ + ^/_matrix/client/(r0|unstable)/register$ + ^/_matrix/client/(r0|unstable)/auth/.*/fallback/web$ + + # Event sending requests + ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/send + ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/state/ + ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/(join|invite|leave|ban|unban|kick)$ + ^/_matrix/client/(api/v1|r0|unstable)/join/ + ^/_matrix/client/(api/v1|r0|unstable)/profile/ + + Additionally, the following REST endpoints can be handled for GET requests: ^/_matrix/federation/v1/groups/ -The above endpoints should all be routed to the federation_reader worker by the -reverse-proxy configuration. +Pagination requests can also be handled, but all requests for a given +room must be routed to the same instance. Additionally, care must be taken to +ensure that the purge history admin API is not used while pagination requests +for the room are in flight: + + ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/messages$ + +Note that a HTTP listener with `client` and `federation` resources must be +configured in the `worker_listeners` option in the worker config. + + +#### Load balancing + +It is possible to run multiple instances of this worker app, with incoming requests +being load-balanced between them by the reverse-proxy. However, different endpoints +have different characteristics and so admins +may wish to run multiple groups of workers handling different endpoints so that +load balancing can be done in different ways. + +For `/sync` and `/initialSync` requests it will be more efficient if all +requests from a particular user are routed to a single instance. Extracting a +user ID from the access token or `Authorization` header is currently left as an +exercise for the reader. Admins may additionally wish to separate out `/sync` +requests that have a `since` query parameter from those that don't (and +`/initialSync`), as requests that don't are known as "initial sync" that happens +when a user logs in on a new device and can be *very* resource intensive, so +isolating these requests will stop them from interfering with other users ongoing +syncs. + +Federation and client requests can be balanced via simple round robin. -The `^/_matrix/federation/v1/send/` endpoint must only be handled by a single -instance. +The inbound federation transaction request `^/_matrix/federation/v1/send/` +should be balanced by source IP so that transactions from the same remote server +go to the same process. -Note that `federation` must be added to the listener resources in the worker config: +Registration/login requests can be handled separately purely to help ensure that +unexpected load doesn't affect new logins and sign ups. + +Finally, event sending requests can be balanced by the room ID in the URI (or +the full URI, or even just round robin), the room ID is the path component after +`/rooms/`. If there is a large bridge connected that is sending or may send lots +of events, then a dedicated set of workers can be provisioned to limit the +effects of bursts of events from that bridge on events sent by normal users. + +#### Stream writers + +Additionally, there is *experimental* support for moving writing of specific +streams (such as events) off of the main process to a particular worker. (This +is only supported with Redis-based replication.) + +Currently support streams are `events` and `typing`. + +To enable this, the worker must have a HTTP replication listener configured, +have a `worker_name` and be listed in the `instance_map` config. For example to +move event persistence off to a dedicated worker, the shared configuration would +include: ```yaml -worker_app: synapse.app.federation_reader -... -worker_listeners: - - type: http - port: - resources: - - names: - - federation +instance_map: + event_persister1: + host: localhost + port: 8034 + +streams_writers: + events: event_persister1 ``` + +### `synapse.app.pusher` + +Handles sending push notifications to sygnal and email. Doesn't handle any +REST endpoints itself, but you should set `start_pushers: False` in the +shared configuration file to stop the main synapse sending push notifications. + +Note this worker cannot be load-balanced: only one instance should be active. + +### `synapse.app.appservice` + +Handles sending output traffic to Application Services. Doesn't handle any +REST endpoints itself, but you should set `notify_appservices: False` in the +shared configuration file to stop the main synapse sending appservice notifications. + +Note this worker cannot be load-balanced: only one instance should be active. + + ### `synapse.app.federation_sender` Handles sending federation traffic to other servers. Doesn't handle any REST endpoints itself, but you should set `send_federation: False` in the shared configuration file to stop the main synapse sending this traffic. -Note this worker cannot be load-balanced: only one instance should be active. +If running multiple federation senders then you must list each +instance in the `federation_sender_instances` option by their `worker_name`. +All instances must be stopped and started when adding or removing instances. +For example: + +```yaml +federation_sender_instances: + - federation_sender1 + - federation_sender2 +``` ### `synapse.app.media_repository` @@ -314,46 +354,6 @@ and you must configure a single instance to run the background tasks, e.g.: media_instance_running_background_jobs: "media-repository-1" ``` -### `synapse.app.client_reader` - -Handles client API endpoints. It can handle REST endpoints matching the -following regular expressions: - - ^/_matrix/client/(api/v1|r0|unstable)/publicRooms$ - ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/joined_members$ - ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/context/.*$ - ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/members$ - ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/state$ - ^/_matrix/client/(api/v1|r0|unstable)/login$ - ^/_matrix/client/(api/v1|r0|unstable)/account/3pid$ - ^/_matrix/client/(api/v1|r0|unstable)/keys/query$ - ^/_matrix/client/(api/v1|r0|unstable)/keys/changes$ - ^/_matrix/client/versions$ - ^/_matrix/client/(api/v1|r0|unstable)/voip/turnServer$ - ^/_matrix/client/(api/v1|r0|unstable)/joined_groups$ - ^/_matrix/client/(api/v1|r0|unstable)/publicised_groups$ - ^/_matrix/client/(api/v1|r0|unstable)/publicised_groups/ - -Additionally, the following REST endpoints can be handled for GET requests: - - ^/_matrix/client/(api/v1|r0|unstable)/pushrules/.*$ - ^/_matrix/client/(api/v1|r0|unstable)/groups/.*$ - ^/_matrix/client/(api/v1|r0|unstable)/user/[^/]*/account_data/ - ^/_matrix/client/(api/v1|r0|unstable)/user/[^/]*/rooms/[^/]*/account_data/ - -Additionally, the following REST endpoints can be handled, but all requests must -be routed to the same instance: - - ^/_matrix/client/(r0|unstable)/register$ - ^/_matrix/client/(r0|unstable)/auth/.*/fallback/web$ - -Pagination requests can also be handled, but all requests with the same path -room must be routed to the same instance. Additionally, care must be taken to -ensure that the purge history admin API is not used while pagination requests -for the room are in flight: - - ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/messages$ - ### `synapse.app.user_dir` Handles searches in the user directory. It can handle REST endpoints matching @@ -388,15 +388,48 @@ file. For example: worker_main_http_uri: http://127.0.0.1:8008 -### `synapse.app.event_creator` +### Historical apps -Handles some event creation. It can handle REST endpoints matching: +*Note:* Historically there used to be more apps, however they have been +amalgamated into a single `synapse.app.generic_worker` app. The remaining apps +are ones that do specific processing unrelated to requests, e.g. the `pusher` +that handles sending out push notifications for new events. The intention is for +all these to be folded into the `generic_worker` app and to use config to define +which processes handle the various proccessing such as push notifications. - ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/send - ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/state/ - ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/(join|invite|leave|ban|unban|kick)$ - ^/_matrix/client/(api/v1|r0|unstable)/join/ - ^/_matrix/client/(api/v1|r0|unstable)/profile/ -It will create events locally and then send them on to the main synapse -instance to be persisted and handled. +## Architectural diagram + +The following shows an example setup using Redis and a reverse proxy: + +``` + Clients & Federation + | + v + +-----------+ + | | + | Reverse | + | Proxy | + | | + +-----------+ + | | | + | | | HTTP requests + +-------------------+ | +-----------+ + | +---+ | + | | | + v v v ++--------------+ +--------------+ +--------------+ +--------------+ +| Main | | Generic | | Generic | | Event | +| Process | | Worker 1 | | Worker 2 | | Persister | ++--------------+ +--------------+ +--------------+ +--------------+ + ^ ^ | ^ | | ^ | ^ ^ + | | | | | | | | | | + | | | | | HTTP | | | | | + | +----------+<--|---|---------+ | | | | + | | +-------------|-->+----------+ | + | | | | + | | | | + v v v v +==================================================================== + Redis pub/sub channel +``` diff --git a/synapse/app/generic_worker.py b/synapse/app/generic_worker.py index ec0dbddb8c..5841454c9a 100644 --- a/synapse/app/generic_worker.py +++ b/synapse/app/generic_worker.py @@ -940,7 +940,7 @@ def start(config_options): config.server.update_user_directory = False if config.worker_app == "synapse.app.federation_sender": - if config.federation.send_federation: + if config.worker.send_federation: sys.stderr.write( "\nThe send_federation must be disabled in the main synapse process" "\nbefore they can be run in a separate worker." @@ -950,10 +950,10 @@ def start(config_options): sys.exit(1) # Force the pushers to start since they will be disabled in the main config - config.federation.send_federation = True + config.worker.send_federation = True else: # For other worker types we force this to off. - config.federation.send_federation = False + config.worker.send_federation = False synapse.events.USE_FROZEN_DICTS = config.use_frozen_dicts diff --git a/synapse/config/federation.py b/synapse/config/federation.py index 82ff9664de..2c77d8f85b 100644 --- a/synapse/config/federation.py +++ b/synapse/config/federation.py @@ -17,23 +17,13 @@ from typing import Optional from netaddr import IPSet -from ._base import Config, ConfigError, ShardedWorkerHandlingConfig +from ._base import Config, ConfigError class FederationConfig(Config): section = "federation" def read_config(self, config, **kwargs): - # Whether to send federation traffic out in this process. This only - # applies to some federation traffic, and so shouldn't be used to - # "disable" federation - self.send_federation = config.get("send_federation", True) - - federation_sender_instances = config.get("federation_sender_instances") or [] - self.federation_shard_config = ShardedWorkerHandlingConfig( - federation_sender_instances - ) - # FIXME: federation_domain_whitelist needs sytests self.federation_domain_whitelist = None # type: Optional[dict] federation_domain_whitelist = config.get("federation_domain_whitelist", None) diff --git a/synapse/config/homeserver.py b/synapse/config/homeserver.py index 8e93d31394..556e291495 100644 --- a/synapse/config/homeserver.py +++ b/synapse/config/homeserver.py @@ -78,7 +78,6 @@ class HomeServerConfig(RootConfig): JWTConfig, PasswordConfig, EmailConfig, - WorkerConfig, PasswordAuthProviderConfig, PushConfig, SpamCheckerConfig, @@ -91,6 +90,7 @@ class HomeServerConfig(RootConfig): RoomDirectoryConfig, ThirdPartyRulesConfig, TracerConfig, + WorkerConfig, RedisConfig, FederationConfig, ] diff --git a/synapse/config/logger.py b/synapse/config/logger.py index 49f6c32beb..dd775a97e8 100644 --- a/synapse/config/logger.py +++ b/synapse/config/logger.py @@ -214,7 +214,7 @@ def setup_logging( Set up the logging subsystem. Args: - config (LoggingConfig | synapse.config.workers.WorkerConfig): + config (LoggingConfig | synapse.config.worker.WorkerConfig): configuration data use_worker_options (bool): True to use the 'worker_log_config' option diff --git a/synapse/config/redis.py b/synapse/config/redis.py index d5d3ca1c9e..1373302335 100644 --- a/synapse/config/redis.py +++ b/synapse/config/redis.py @@ -21,7 +21,7 @@ class RedisConfig(Config): section = "redis" def read_config(self, config, **kwargs): - redis_config = config.get("redis", {}) + redis_config = config.get("redis") or {} self.redis_enabled = redis_config.get("enabled", False) if not self.redis_enabled: @@ -32,3 +32,24 @@ class RedisConfig(Config): self.redis_host = redis_config.get("host", "localhost") self.redis_port = redis_config.get("port", 6379) self.redis_password = redis_config.get("password") + + def generate_config_section(self, config_dir_path, server_name, **kwargs): + return """\ + # Configuration for Redis when using workers. This *must* be enabled when + # using workers (unless using old style direct TCP configuration). + # + redis: + # Uncomment the below to enable Redis support. + # + #enabled: true + + # Optional host and port to use to connect to redis. Defaults to + # localhost and 6379 + # + #host: localhost + #port: 6379 + + # Optional password if configured on the Redis instance + # + #password: + """ diff --git a/synapse/config/workers.py b/synapse/config/workers.py index 2574cd3aa1..c784a71508 100644 --- a/synapse/config/workers.py +++ b/synapse/config/workers.py @@ -15,7 +15,7 @@ import attr -from ._base import Config, ConfigError +from ._base import Config, ConfigError, ShardedWorkerHandlingConfig from .server import ListenerConfig, parse_listener_def @@ -85,6 +85,16 @@ class WorkerConfig(Config): ) ) + # Whether to send federation traffic out in this process. This only + # applies to some federation traffic, and so shouldn't be used to + # "disable" federation + self.send_federation = config.get("send_federation", True) + + federation_sender_instances = config.get("federation_sender_instances") or [] + self.federation_shard_config = ShardedWorkerHandlingConfig( + federation_sender_instances + ) + # A map from instance name to host/port of their HTTP replication endpoint. instance_map = config.get("instance_map") or {} self.instance_map = { @@ -105,6 +115,43 @@ class WorkerConfig(Config): % (instance, stream) ) + def generate_config_section(self, config_dir_path, server_name, **kwargs): + return """\ + ## Workers ## + + # Disables sending of outbound federation transactions on the main process. + # Uncomment if using a federation sender worker. + # + #send_federation: false + + # It is possible to run multiple federation sender workers, in which case the + # work is balanced across them. + # + # This configuration must be shared between all federation sender workers, and if + # changed all federation sender workers must be stopped at the same time and then + # started, to ensure that all instances are running with the same config (otherwise + # events may be dropped). + # + #federation_sender_instances: + # - federation_sender1 + + # When using workers this should be a map from `worker_name` to the + # HTTP replication listener of the worker, if configured. + # + #instance_map: + # worker1: + # host: localhost + # port: 8034 + + # Experimental: When using workers you can define which workers should + # handle event persistence and typing notifications. Any worker + # specified here must also be in the `instance_map`. + # + #stream_writers: + # events: worker1 + # typing: worker1 + """ + def read_arguments(self, args): # We support a bunch of command line arguments that override options in # the config. A lot of these options have a worker_* prefix when running diff --git a/synapse/federation/send_queue.py b/synapse/federation/send_queue.py index 4fc9ff92e5..2b0ab2dcbf 100644 --- a/synapse/federation/send_queue.py +++ b/synapse/federation/send_queue.py @@ -57,7 +57,7 @@ class FederationRemoteSendQueue(object): # We may have multiple federation sender instances, so we need to track # their positions separately. - self._sender_instances = hs.config.federation.federation_shard_config.instances + self._sender_instances = hs.config.worker.federation_shard_config.instances self._sender_positions = {} # Pending presence map user_id -> UserPresenceState diff --git a/synapse/federation/sender/__init__.py b/synapse/federation/sender/__init__.py index ba4ddd2370..6ae6522f87 100644 --- a/synapse/federation/sender/__init__.py +++ b/synapse/federation/sender/__init__.py @@ -70,7 +70,7 @@ class FederationSender(object): self._transaction_manager = TransactionManager(hs) self._instance_name = hs.get_instance_name() - self._federation_shard_config = hs.config.federation.federation_shard_config + self._federation_shard_config = hs.config.worker.federation_shard_config # map from destination to PerDestinationQueue self._per_destination_queues = {} # type: Dict[str, PerDestinationQueue] diff --git a/synapse/federation/sender/per_destination_queue.py b/synapse/federation/sender/per_destination_queue.py index 3436741783..dd150f89a6 100644 --- a/synapse/federation/sender/per_destination_queue.py +++ b/synapse/federation/sender/per_destination_queue.py @@ -75,7 +75,7 @@ class PerDestinationQueue(object): self._store = hs.get_datastore() self._transaction_manager = transaction_manager self._instance_name = hs.get_instance_name() - self._federation_shard_config = hs.config.federation.federation_shard_config + self._federation_shard_config = hs.config.worker.federation_shard_config self._should_send_on_this_instance = True if not self._federation_shard_config.should_handle( diff --git a/synapse/storage/data_stores/main/stream.py b/synapse/storage/data_stores/main/stream.py index 5e32c7aa1e..10d39b3699 100644 --- a/synapse/storage/data_stores/main/stream.py +++ b/synapse/storage/data_stores/main/stream.py @@ -255,7 +255,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): self._instance_name = hs.get_instance_name() self._send_federation = hs.should_send_federation() - self._federation_shard_config = hs.config.federation.federation_shard_config + self._federation_shard_config = hs.config.worker.federation_shard_config # If we're a process that sends federation we may need to reset the # `federation_stream_position` table to match the current sharding -- cgit 1.5.1 From 82c1ee1c22a87b9e6e3179947014b0f11c0a1ac3 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Wed, 2 Sep 2020 15:48:37 +0100 Subject: Add experimental support for sharding event persister. (#8170) This is *not* ready for production yet. Caveats: 1. We should write some tests... 2. The stream token that we use for events can get stalled at the minimum position of all writers. This means that new events may not be processed and e.g. sent down sync streams if a writer isn't writing or is slow. --- changelog.d/8170.feature | 1 + synapse/config/_base.py | 21 ++++++- synapse/config/_base.pyi | 1 + synapse/config/workers.py | 37 ++++++++---- synapse/handlers/federation.py | 44 ++++++++++----- synapse/handlers/message.py | 14 +++-- synapse/handlers/room.py | 14 +++-- synapse/handlers/room_member.py | 7 --- synapse/replication/http/federation.py | 12 +++- synapse/replication/tcp/handler.py | 2 +- synapse/replication/tcp/streams/events.py | 4 +- synapse/storage/databases/__init__.py | 2 +- synapse/storage/databases/main/event_federation.py | 2 +- synapse/storage/databases/main/events.py | 4 +- synapse/storage/databases/main/events_worker.py | 66 +++++++++++++++------- .../schema/delta/58/14events_instance_name.sql | 16 ++++++ .../delta/58/14events_instance_name.sql.postgres | 26 +++++++++ synapse/storage/util/id_generators.py | 10 ++-- 18 files changed, 206 insertions(+), 77 deletions(-) create mode 100644 changelog.d/8170.feature create mode 100644 synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql create mode 100644 synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql.postgres (limited to 'synapse/config/workers.py') diff --git a/changelog.d/8170.feature b/changelog.d/8170.feature new file mode 100644 index 0000000000..b363e929ea --- /dev/null +++ b/changelog.d/8170.feature @@ -0,0 +1 @@ +Add experimental support for sharding event persister. diff --git a/synapse/config/_base.py b/synapse/config/_base.py index 1417487427..73f0717b0d 100644 --- a/synapse/config/_base.py +++ b/synapse/config/_base.py @@ -832,11 +832,26 @@ class ShardedWorkerHandlingConfig: def should_handle(self, instance_name: str, key: str) -> bool: """Whether this instance is responsible for handling the given key. """ - - # If multiple instances are not defined we always return true. + # If multiple instances are not defined we always return true if not self.instances or len(self.instances) == 1: return True + return self.get_instance(key) == instance_name + + def get_instance(self, key: str) -> str: + """Get the instance responsible for handling the given key. + + Note: For things like federation sending the config for which instance + is sending is known only to the sender instance if there is only one. + Therefore `should_handle` should be used where possible. + """ + + if not self.instances: + return "master" + + if len(self.instances) == 1: + return self.instances[0] + # We shard by taking the hash, modulo it by the number of instances and # then checking whether this instance matches the instance at that # index. @@ -846,7 +861,7 @@ class ShardedWorkerHandlingConfig: dest_hash = sha256(key.encode("utf8")).digest() dest_int = int.from_bytes(dest_hash, byteorder="little") remainder = dest_int % (len(self.instances)) - return self.instances[remainder] == instance_name + return self.instances[remainder] __all__ = ["Config", "RootConfig", "ShardedWorkerHandlingConfig"] diff --git a/synapse/config/_base.pyi b/synapse/config/_base.pyi index eb911e8f9f..b8faafa9bd 100644 --- a/synapse/config/_base.pyi +++ b/synapse/config/_base.pyi @@ -142,3 +142,4 @@ class ShardedWorkerHandlingConfig: instances: List[str] def __init__(self, instances: List[str]) -> None: ... def should_handle(self, instance_name: str, key: str) -> bool: ... + def get_instance(self, key: str) -> str: ... diff --git a/synapse/config/workers.py b/synapse/config/workers.py index c784a71508..f23e42cdf9 100644 --- a/synapse/config/workers.py +++ b/synapse/config/workers.py @@ -13,12 +13,24 @@ # See the License for the specific language governing permissions and # limitations under the License. +from typing import List, Union + import attr from ._base import Config, ConfigError, ShardedWorkerHandlingConfig from .server import ListenerConfig, parse_listener_def +def _instance_to_list_converter(obj: Union[str, List[str]]) -> List[str]: + """Helper for allowing parsing a string or list of strings to a config + option expecting a list of strings. + """ + + if isinstance(obj, str): + return [obj] + return obj + + @attr.s class InstanceLocationConfig: """The host and port to talk to an instance via HTTP replication. @@ -33,11 +45,13 @@ class WriterLocations: """Specifies the instances that write various streams. Attributes: - events: The instance that writes to the event and backfill streams. - events: The instance that writes to the typing stream. + events: The instances that write to the event and backfill streams. + typing: The instance that writes to the typing stream. """ - events = attr.ib(default="master", type=str) + events = attr.ib( + default=["master"], type=List[str], converter=_instance_to_list_converter + ) typing = attr.ib(default="master", type=str) @@ -105,15 +119,18 @@ class WorkerConfig(Config): writers = config.get("stream_writers") or {} self.writers = WriterLocations(**writers) - # Check that the configured writer for events and typing also appears in + # Check that the configured writers for events and typing also appears in # `instance_map`. for stream in ("events", "typing"): - instance = getattr(self.writers, stream) - if instance != "master" and instance not in self.instance_map: - raise ConfigError( - "Instance %r is configured to write %s but does not appear in `instance_map` config." - % (instance, stream) - ) + instances = _instance_to_list_converter(getattr(self.writers, stream)) + for instance in instances: + if instance != "master" and instance not in self.instance_map: + raise ConfigError( + "Instance %r is configured to write %s but does not appear in `instance_map` config." + % (instance, stream) + ) + + self.events_shard_config = ShardedWorkerHandlingConfig(self.writers.events) def generate_config_section(self, config_dir_path, server_name, **kwargs): return """\ diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 16389a0dca..bd8efbb768 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -923,7 +923,8 @@ class FederationHandler(BaseHandler): ) ) - await self._handle_new_events(dest, ev_infos, backfilled=True) + if ev_infos: + await self._handle_new_events(dest, room_id, ev_infos, backfilled=True) # Step 2: Persist the rest of the events in the chunk one by one events.sort(key=lambda e: e.depth) @@ -1216,7 +1217,7 @@ class FederationHandler(BaseHandler): event_infos.append(_NewEventInfo(event, None, auth)) await self._handle_new_events( - destination, event_infos, + destination, room_id, event_infos, ) def _sanity_check_event(self, ev): @@ -1363,15 +1364,15 @@ class FederationHandler(BaseHandler): ) max_stream_id = await self._persist_auth_tree( - origin, auth_chain, state, event, room_version_obj + origin, room_id, auth_chain, state, event, room_version_obj ) # We wait here until this instance has seen the events come down # replication (if we're using replication) as the below uses caches. - # - # TODO: Currently the events stream is written to from master await self._replication.wait_for_stream_position( - self.config.worker.writers.events, "events", max_stream_id + self.config.worker.events_shard_config.get_instance(room_id), + "events", + max_stream_id, ) # Check whether this room is the result of an upgrade of a room we already know @@ -1625,7 +1626,7 @@ class FederationHandler(BaseHandler): ) context = await self.state_handler.compute_event_context(event) - await self.persist_events_and_notify([(event, context)]) + await self.persist_events_and_notify(event.room_id, [(event, context)]) return event @@ -1652,7 +1653,9 @@ class FederationHandler(BaseHandler): await self.federation_client.send_leave(host_list, event) context = await self.state_handler.compute_event_context(event) - stream_id = await self.persist_events_and_notify([(event, context)]) + stream_id = await self.persist_events_and_notify( + event.room_id, [(event, context)] + ) return event, stream_id @@ -1900,7 +1903,7 @@ class FederationHandler(BaseHandler): ) await self.persist_events_and_notify( - [(event, context)], backfilled=backfilled + event.room_id, [(event, context)], backfilled=backfilled ) except Exception: run_in_background( @@ -1913,6 +1916,7 @@ class FederationHandler(BaseHandler): async def _handle_new_events( self, origin: str, + room_id: str, event_infos: Iterable[_NewEventInfo], backfilled: bool = False, ) -> None: @@ -1944,6 +1948,7 @@ class FederationHandler(BaseHandler): ) await self.persist_events_and_notify( + room_id, [ (ev_info.event, context) for ev_info, context in zip(event_infos, contexts) @@ -1954,6 +1959,7 @@ class FederationHandler(BaseHandler): async def _persist_auth_tree( self, origin: str, + room_id: str, auth_events: List[EventBase], state: List[EventBase], event: EventBase, @@ -1968,6 +1974,7 @@ class FederationHandler(BaseHandler): Args: origin: Where the events came from + room_id, auth_events state event @@ -2042,17 +2049,20 @@ class FederationHandler(BaseHandler): events_to_context[e.event_id].rejected = RejectedReason.AUTH_ERROR await self.persist_events_and_notify( + room_id, [ (e, events_to_context[e.event_id]) for e in itertools.chain(auth_events, state) - ] + ], ) new_event_context = await self.state_handler.compute_event_context( event, old_state=state ) - return await self.persist_events_and_notify([(event, new_event_context)]) + return await self.persist_events_and_notify( + room_id, [(event, new_event_context)] + ) async def _prep_event( self, @@ -2903,6 +2913,7 @@ class FederationHandler(BaseHandler): async def persist_events_and_notify( self, + room_id: str, event_and_contexts: Sequence[Tuple[EventBase, EventContext]], backfilled: bool = False, ) -> int: @@ -2910,14 +2921,19 @@ class FederationHandler(BaseHandler): necessary. Args: - event_and_contexts: + room_id: The room ID of events being persisted. + event_and_contexts: Sequence of events with their associated + context that should be persisted. All events must belong to + the same room. backfilled: Whether these events are a result of backfilling or not """ - if self.config.worker.writers.events != self._instance_name: + instance = self.config.worker.events_shard_config.get_instance(room_id) + if instance != self._instance_name: result = await self._send_events( - instance_name=self.config.worker.writers.events, + instance_name=instance, store=self.store, + room_id=room_id, event_and_contexts=event_and_contexts, backfilled=backfilled, ) diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index 72bb638167..6398774c02 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -376,9 +376,8 @@ class EventCreationHandler(object): self.notifier = hs.get_notifier() self.config = hs.config self.require_membership_for_aliases = hs.config.require_membership_for_aliases - self._is_event_writer = ( - self.config.worker.writers.events == hs.get_instance_name() - ) + self._events_shard_config = self.config.worker.events_shard_config + self._instance_name = hs.get_instance_name() self.room_invite_state_types = self.hs.config.room_invite_state_types @@ -904,9 +903,10 @@ class EventCreationHandler(object): try: # If we're a worker we need to hit out to the master. - if not self._is_event_writer: + writer_instance = self._events_shard_config.get_instance(event.room_id) + if writer_instance != self._instance_name: result = await self.send_event( - instance_name=self.config.worker.writers.events, + instance_name=writer_instance, event_id=event.event_id, store=self.store, requester=requester, @@ -974,7 +974,9 @@ class EventCreationHandler(object): This should only be run on the instance in charge of persisting events. """ - assert self._is_event_writer + assert self._events_shard_config.should_handle( + self._instance_name, event.room_id + ) if ratelimit: # We check if this is a room admin redacting an event so that we diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index 9d5b1828df..55794c3057 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -804,7 +804,9 @@ class RoomCreationHandler(BaseHandler): # Always wait for room creation to progate before returning await self._replication.wait_for_stream_position( - self.hs.config.worker.writers.events, "events", last_stream_id + self.hs.config.worker.events_shard_config.get_instance(room_id), + "events", + last_stream_id, ) return result, last_stream_id @@ -1260,10 +1262,10 @@ class RoomShutdownHandler(object): # We now wait for the create room to come back in via replication so # that we can assume that all the joins/invites have propogated before # we try and auto join below. - # - # TODO: Currently the events stream is written to from master await self._replication.wait_for_stream_position( - self.hs.config.worker.writers.events, "events", stream_id + self.hs.config.worker.events_shard_config.get_instance(new_room_id), + "events", + stream_id, ) else: new_room_id = None @@ -1293,7 +1295,9 @@ class RoomShutdownHandler(object): # Wait for leave to come in over replication before trying to forget. await self._replication.wait_for_stream_position( - self.hs.config.worker.writers.events, "events", stream_id + self.hs.config.worker.events_shard_config.get_instance(room_id), + "events", + stream_id, ) await self.room_member_handler.forget(target_requester.user, room_id) diff --git a/synapse/handlers/room_member.py b/synapse/handlers/room_member.py index a7962b0ada..e5cb3c2d5c 100644 --- a/synapse/handlers/room_member.py +++ b/synapse/handlers/room_member.py @@ -82,13 +82,6 @@ class RoomMemberHandler(object): self._enable_lookup = hs.config.enable_3pid_lookup self.allow_per_room_profiles = self.config.allow_per_room_profiles - self._event_stream_writer_instance = hs.config.worker.writers.events - self._is_on_event_persistence_instance = ( - self._event_stream_writer_instance == hs.get_instance_name() - ) - if self._is_on_event_persistence_instance: - self.persist_event_storage = hs.get_storage().persistence - self._join_rate_limiter_local = Ratelimiter( clock=self.clock, rate_hz=hs.config.ratelimiting.rc_joins_local.per_second, diff --git a/synapse/replication/http/federation.py b/synapse/replication/http/federation.py index 6b56315148..5c8be747e1 100644 --- a/synapse/replication/http/federation.py +++ b/synapse/replication/http/federation.py @@ -65,10 +65,11 @@ class ReplicationFederationSendEventsRestServlet(ReplicationEndpoint): self.federation_handler = hs.get_handlers().federation_handler @staticmethod - async def _serialize_payload(store, event_and_contexts, backfilled): + async def _serialize_payload(store, room_id, event_and_contexts, backfilled): """ Args: store + room_id (str) event_and_contexts (list[tuple[FrozenEvent, EventContext]]) backfilled (bool): Whether or not the events are the result of backfilling @@ -88,7 +89,11 @@ class ReplicationFederationSendEventsRestServlet(ReplicationEndpoint): } ) - payload = {"events": event_payloads, "backfilled": backfilled} + payload = { + "events": event_payloads, + "backfilled": backfilled, + "room_id": room_id, + } return payload @@ -96,6 +101,7 @@ class ReplicationFederationSendEventsRestServlet(ReplicationEndpoint): with Measure(self.clock, "repl_fed_send_events_parse"): content = parse_json_object_from_request(request) + room_id = content["room_id"] backfilled = content["backfilled"] event_payloads = content["events"] @@ -120,7 +126,7 @@ class ReplicationFederationSendEventsRestServlet(ReplicationEndpoint): logger.info("Got %d events from federation", len(event_and_contexts)) max_stream_id = await self.federation_handler.persist_events_and_notify( - event_and_contexts, backfilled + room_id, event_and_contexts, backfilled ) return 200, {"max_stream_id": max_stream_id} diff --git a/synapse/replication/tcp/handler.py b/synapse/replication/tcp/handler.py index 1c303f3a46..b323841f73 100644 --- a/synapse/replication/tcp/handler.py +++ b/synapse/replication/tcp/handler.py @@ -109,7 +109,7 @@ class ReplicationCommandHandler: if isinstance(stream, (EventsStream, BackfillStream)): # Only add EventStream and BackfillStream as a source on the # instance in charge of event persistence. - if hs.config.worker.writers.events == hs.get_instance_name(): + if hs.get_instance_name() in hs.config.worker.writers.events: self._streams_to_replicate.append(stream) continue diff --git a/synapse/replication/tcp/streams/events.py b/synapse/replication/tcp/streams/events.py index 16c63ff4ec..3705618b4f 100644 --- a/synapse/replication/tcp/streams/events.py +++ b/synapse/replication/tcp/streams/events.py @@ -19,7 +19,7 @@ from typing import List, Tuple, Type import attr -from ._base import Stream, StreamUpdateResult, Token, current_token_without_instance +from ._base import Stream, StreamUpdateResult, Token """Handling of the 'events' replication stream @@ -117,7 +117,7 @@ class EventsStream(Stream): self._store = hs.get_datastore() super().__init__( hs.get_instance_name(), - current_token_without_instance(self._store.get_current_events_token), + self._store._stream_id_gen.get_current_token_for_writer, self._update_function, ) diff --git a/synapse/storage/databases/__init__.py b/synapse/storage/databases/__init__.py index 0ac854aee2..c73d54fb67 100644 --- a/synapse/storage/databases/__init__.py +++ b/synapse/storage/databases/__init__.py @@ -68,7 +68,7 @@ class Databases(object): # If we're on a process that can persist events also # instantiate a `PersistEventsStore` - if hs.config.worker.writers.events == hs.get_instance_name(): + if hs.get_instance_name() in hs.config.worker.writers.events: persist_events = PersistEventsStore(hs, database, main) if "state" in database_config.databases: diff --git a/synapse/storage/databases/main/event_federation.py b/synapse/storage/databases/main/event_federation.py index 0b69aa6a94..4c3c162acf 100644 --- a/synapse/storage/databases/main/event_federation.py +++ b/synapse/storage/databases/main/event_federation.py @@ -438,7 +438,7 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas """ if stream_ordering <= self.stream_ordering_month_ago: - raise StoreError(400, "stream_ordering too old") + raise StoreError(400, "stream_ordering too old %s" % (stream_ordering,)) sql = """ SELECT event_id FROM stream_ordering_to_exterm diff --git a/synapse/storage/databases/main/events.py b/synapse/storage/databases/main/events.py index 6313b41eef..46b11e705b 100644 --- a/synapse/storage/databases/main/events.py +++ b/synapse/storage/databases/main/events.py @@ -97,6 +97,7 @@ class PersistEventsStore: self.store = main_data_store self.database_engine = db.engine self._clock = hs.get_clock() + self._instance_name = hs.get_instance_name() self._ephemeral_messages_enabled = hs.config.enable_ephemeral_messages self.is_mine_id = hs.is_mine_id @@ -108,7 +109,7 @@ class PersistEventsStore: # This should only exist on instances that are configured to write assert ( - hs.config.worker.writers.events == hs.get_instance_name() + hs.get_instance_name() in hs.config.worker.writers.events ), "Can only instantiate EventsStore on master" async def _persist_events_and_state_updates( @@ -800,6 +801,7 @@ class PersistEventsStore: table="events", values=[ { + "instance_name": self._instance_name, "stream_ordering": event.internal_metadata.stream_ordering, "topological_ordering": event.depth, "depth": event.depth, diff --git a/synapse/storage/databases/main/events_worker.py b/synapse/storage/databases/main/events_worker.py index a7a73cc3d8..17f5997b89 100644 --- a/synapse/storage/databases/main/events_worker.py +++ b/synapse/storage/databases/main/events_worker.py @@ -42,7 +42,8 @@ from synapse.replication.tcp.streams import BackfillStream from synapse.replication.tcp.streams.events import EventsStream from synapse.storage._base import SQLBaseStore, db_to_json, make_in_list_sql_clause from synapse.storage.database import DatabasePool -from synapse.storage.util.id_generators import StreamIdGenerator +from synapse.storage.engines import PostgresEngine +from synapse.storage.util.id_generators import MultiWriterIdGenerator, StreamIdGenerator from synapse.types import Collection, get_domain_from_id from synapse.util.caches.descriptors import Cache, cached from synapse.util.iterutils import batch_iter @@ -78,27 +79,54 @@ class EventsWorkerStore(SQLBaseStore): def __init__(self, database: DatabasePool, db_conn, hs): super(EventsWorkerStore, self).__init__(database, db_conn, hs) - if hs.config.worker.writers.events == hs.get_instance_name(): - # We are the process in charge of generating stream ids for events, - # so instantiate ID generators based on the database - self._stream_id_gen = StreamIdGenerator( - db_conn, "events", "stream_ordering", + if isinstance(database.engine, PostgresEngine): + # If we're using Postgres than we can use `MultiWriterIdGenerator` + # regardless of whether this process writes to the streams or not. + self._stream_id_gen = MultiWriterIdGenerator( + db_conn=db_conn, + db=database, + instance_name=hs.get_instance_name(), + table="events", + instance_column="instance_name", + id_column="stream_ordering", + sequence_name="events_stream_seq", ) - self._backfill_id_gen = StreamIdGenerator( - db_conn, - "events", - "stream_ordering", - step=-1, - extra_tables=[("ex_outlier_stream", "event_stream_ordering")], + self._backfill_id_gen = MultiWriterIdGenerator( + db_conn=db_conn, + db=database, + instance_name=hs.get_instance_name(), + table="events", + instance_column="instance_name", + id_column="stream_ordering", + sequence_name="events_backfill_stream_seq", + positive=False, ) else: - # Another process is in charge of persisting events and generating - # stream IDs: rely on the replication streams to let us know which - # IDs we can process. - self._stream_id_gen = SlavedIdTracker(db_conn, "events", "stream_ordering") - self._backfill_id_gen = SlavedIdTracker( - db_conn, "events", "stream_ordering", step=-1 - ) + # We shouldn't be running in worker mode with SQLite, but its useful + # to support it for unit tests. + # + # If this process is the writer than we need to use + # `StreamIdGenerator`, otherwise we use `SlavedIdTracker` which gets + # updated over replication. (Multiple writers are not supported for + # SQLite). + if hs.get_instance_name() in hs.config.worker.writers.events: + self._stream_id_gen = StreamIdGenerator( + db_conn, "events", "stream_ordering", + ) + self._backfill_id_gen = StreamIdGenerator( + db_conn, + "events", + "stream_ordering", + step=-1, + extra_tables=[("ex_outlier_stream", "event_stream_ordering")], + ) + else: + self._stream_id_gen = SlavedIdTracker( + db_conn, "events", "stream_ordering" + ) + self._backfill_id_gen = SlavedIdTracker( + db_conn, "events", "stream_ordering", step=-1 + ) self._get_event_cache = Cache( "*getEvent*", diff --git a/synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql b/synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql new file mode 100644 index 0000000000..98ff76d709 --- /dev/null +++ b/synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql @@ -0,0 +1,16 @@ +/* Copyright 2020 The Matrix.org Foundation C.I.C. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * 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. + */ + +ALTER TABLE events ADD COLUMN instance_name TEXT; diff --git a/synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql.postgres b/synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql.postgres new file mode 100644 index 0000000000..97c1e6a0c5 --- /dev/null +++ b/synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql.postgres @@ -0,0 +1,26 @@ +/* Copyright 2020 The Matrix.org Foundation C.I.C. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * 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. + */ + +CREATE SEQUENCE IF NOT EXISTS events_stream_seq; + +SELECT setval('events_stream_seq', ( + SELECT COALESCE(MAX(stream_ordering), 1) FROM events +)); + +CREATE SEQUENCE IF NOT EXISTS events_backfill_stream_seq; + +SELECT setval('events_backfill_stream_seq', ( + SELECT COALESCE(-MIN(stream_ordering), 1) FROM events +)); diff --git a/synapse/storage/util/id_generators.py b/synapse/storage/util/id_generators.py index 9f3d23f0a5..8fd21c2bf8 100644 --- a/synapse/storage/util/id_generators.py +++ b/synapse/storage/util/id_generators.py @@ -231,8 +231,12 @@ class MultiWriterIdGenerator: # gaps should be relatively rare it's still worth doing the book keeping # that allows us to skip forwards when there are gapless runs of # positions. + # + # We start at 1 here as a) the first generated stream ID will be 2, and + # b) other parts of the code assume that stream IDs are strictly greater + # than 0. self._persisted_upto_position = ( - min(self._current_positions.values()) if self._current_positions else 0 + min(self._current_positions.values()) if self._current_positions else 1 ) self._known_persisted_positions = [] # type: List[int] @@ -362,9 +366,7 @@ class MultiWriterIdGenerator: equal to it have been successfully persisted. """ - # Currently we don't support this operation, as it's not obvious how to - # condense the stream positions of multiple writers into a single int. - raise NotImplementedError() + return self.get_persisted_upto_position() def get_current_token_for_writer(self, instance_name: str) -> int: """Returns the position of the given writer. -- cgit 1.5.1 From 9f8abdcc3828e942f09cc62e29dff93dbaa01ec7 Mon Sep 17 00:00:00 2001 From: Brendan Abolivier Date: Fri, 4 Sep 2020 10:19:42 +0100 Subject: Revert "Add experimental support for sharding event persister. (#8170)" (#8242) * Revert "Add experimental support for sharding event persister. (#8170)" This reverts commit 82c1ee1c22a87b9e6e3179947014b0f11c0a1ac3. * Changelog --- changelog.d/8170.feature | 1 - changelog.d/8242.feature | 1 + synapse/config/_base.py | 21 +------ synapse/config/_base.pyi | 1 - synapse/config/workers.py | 37 ++++-------- synapse/handlers/federation.py | 44 +++++---------- synapse/handlers/message.py | 14 ++--- synapse/handlers/room.py | 14 ++--- synapse/handlers/room_member.py | 7 +++ synapse/replication/http/federation.py | 12 +--- synapse/replication/tcp/handler.py | 2 +- synapse/replication/tcp/streams/events.py | 4 +- synapse/storage/databases/__init__.py | 2 +- synapse/storage/databases/main/event_federation.py | 2 +- synapse/storage/databases/main/events.py | 4 +- synapse/storage/databases/main/events_worker.py | 66 +++++++--------------- .../schema/delta/58/14events_instance_name.sql | 16 ------ .../delta/58/14events_instance_name.sql.postgres | 26 --------- synapse/storage/util/id_generators.py | 10 ++-- 19 files changed, 78 insertions(+), 206 deletions(-) delete mode 100644 changelog.d/8170.feature create mode 100644 changelog.d/8242.feature delete mode 100644 synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql delete mode 100644 synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql.postgres (limited to 'synapse/config/workers.py') diff --git a/changelog.d/8170.feature b/changelog.d/8170.feature deleted file mode 100644 index b363e929ea..0000000000 --- a/changelog.d/8170.feature +++ /dev/null @@ -1 +0,0 @@ -Add experimental support for sharding event persister. diff --git a/changelog.d/8242.feature b/changelog.d/8242.feature new file mode 100644 index 0000000000..f6891e360d --- /dev/null +++ b/changelog.d/8242.feature @@ -0,0 +1 @@ +Back out experimental support for sharding event persister. **PLEASE REMOVE THIS LINE FROM THE FINAL CHANGELOG** diff --git a/synapse/config/_base.py b/synapse/config/_base.py index 73f0717b0d..1417487427 100644 --- a/synapse/config/_base.py +++ b/synapse/config/_base.py @@ -832,26 +832,11 @@ class ShardedWorkerHandlingConfig: def should_handle(self, instance_name: str, key: str) -> bool: """Whether this instance is responsible for handling the given key. """ - # If multiple instances are not defined we always return true + + # If multiple instances are not defined we always return true. if not self.instances or len(self.instances) == 1: return True - return self.get_instance(key) == instance_name - - def get_instance(self, key: str) -> str: - """Get the instance responsible for handling the given key. - - Note: For things like federation sending the config for which instance - is sending is known only to the sender instance if there is only one. - Therefore `should_handle` should be used where possible. - """ - - if not self.instances: - return "master" - - if len(self.instances) == 1: - return self.instances[0] - # We shard by taking the hash, modulo it by the number of instances and # then checking whether this instance matches the instance at that # index. @@ -861,7 +846,7 @@ class ShardedWorkerHandlingConfig: dest_hash = sha256(key.encode("utf8")).digest() dest_int = int.from_bytes(dest_hash, byteorder="little") remainder = dest_int % (len(self.instances)) - return self.instances[remainder] + return self.instances[remainder] == instance_name __all__ = ["Config", "RootConfig", "ShardedWorkerHandlingConfig"] diff --git a/synapse/config/_base.pyi b/synapse/config/_base.pyi index b8faafa9bd..eb911e8f9f 100644 --- a/synapse/config/_base.pyi +++ b/synapse/config/_base.pyi @@ -142,4 +142,3 @@ class ShardedWorkerHandlingConfig: instances: List[str] def __init__(self, instances: List[str]) -> None: ... def should_handle(self, instance_name: str, key: str) -> bool: ... - def get_instance(self, key: str) -> str: ... diff --git a/synapse/config/workers.py b/synapse/config/workers.py index f23e42cdf9..c784a71508 100644 --- a/synapse/config/workers.py +++ b/synapse/config/workers.py @@ -13,24 +13,12 @@ # See the License for the specific language governing permissions and # limitations under the License. -from typing import List, Union - import attr from ._base import Config, ConfigError, ShardedWorkerHandlingConfig from .server import ListenerConfig, parse_listener_def -def _instance_to_list_converter(obj: Union[str, List[str]]) -> List[str]: - """Helper for allowing parsing a string or list of strings to a config - option expecting a list of strings. - """ - - if isinstance(obj, str): - return [obj] - return obj - - @attr.s class InstanceLocationConfig: """The host and port to talk to an instance via HTTP replication. @@ -45,13 +33,11 @@ class WriterLocations: """Specifies the instances that write various streams. Attributes: - events: The instances that write to the event and backfill streams. - typing: The instance that writes to the typing stream. + events: The instance that writes to the event and backfill streams. + events: The instance that writes to the typing stream. """ - events = attr.ib( - default=["master"], type=List[str], converter=_instance_to_list_converter - ) + events = attr.ib(default="master", type=str) typing = attr.ib(default="master", type=str) @@ -119,18 +105,15 @@ class WorkerConfig(Config): writers = config.get("stream_writers") or {} self.writers = WriterLocations(**writers) - # Check that the configured writers for events and typing also appears in + # Check that the configured writer for events and typing also appears in # `instance_map`. for stream in ("events", "typing"): - instances = _instance_to_list_converter(getattr(self.writers, stream)) - for instance in instances: - if instance != "master" and instance not in self.instance_map: - raise ConfigError( - "Instance %r is configured to write %s but does not appear in `instance_map` config." - % (instance, stream) - ) - - self.events_shard_config = ShardedWorkerHandlingConfig(self.writers.events) + instance = getattr(self.writers, stream) + if instance != "master" and instance not in self.instance_map: + raise ConfigError( + "Instance %r is configured to write %s but does not appear in `instance_map` config." + % (instance, stream) + ) def generate_config_section(self, config_dir_path, server_name, **kwargs): return """\ diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 310c7f7138..43f2986f89 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -923,8 +923,7 @@ class FederationHandler(BaseHandler): ) ) - if ev_infos: - await self._handle_new_events(dest, room_id, ev_infos, backfilled=True) + await self._handle_new_events(dest, ev_infos, backfilled=True) # Step 2: Persist the rest of the events in the chunk one by one events.sort(key=lambda e: e.depth) @@ -1217,7 +1216,7 @@ class FederationHandler(BaseHandler): event_infos.append(_NewEventInfo(event, None, auth)) await self._handle_new_events( - destination, room_id, event_infos, + destination, event_infos, ) def _sanity_check_event(self, ev): @@ -1364,15 +1363,15 @@ class FederationHandler(BaseHandler): ) max_stream_id = await self._persist_auth_tree( - origin, room_id, auth_chain, state, event, room_version_obj + origin, auth_chain, state, event, room_version_obj ) # We wait here until this instance has seen the events come down # replication (if we're using replication) as the below uses caches. + # + # TODO: Currently the events stream is written to from master await self._replication.wait_for_stream_position( - self.config.worker.events_shard_config.get_instance(room_id), - "events", - max_stream_id, + self.config.worker.writers.events, "events", max_stream_id ) # Check whether this room is the result of an upgrade of a room we already know @@ -1626,7 +1625,7 @@ class FederationHandler(BaseHandler): ) context = await self.state_handler.compute_event_context(event) - await self.persist_events_and_notify(event.room_id, [(event, context)]) + await self.persist_events_and_notify([(event, context)]) return event @@ -1653,9 +1652,7 @@ class FederationHandler(BaseHandler): await self.federation_client.send_leave(host_list, event) context = await self.state_handler.compute_event_context(event) - stream_id = await self.persist_events_and_notify( - event.room_id, [(event, context)] - ) + stream_id = await self.persist_events_and_notify([(event, context)]) return event, stream_id @@ -1903,7 +1900,7 @@ class FederationHandler(BaseHandler): ) await self.persist_events_and_notify( - event.room_id, [(event, context)], backfilled=backfilled + [(event, context)], backfilled=backfilled ) except Exception: run_in_background( @@ -1916,7 +1913,6 @@ class FederationHandler(BaseHandler): async def _handle_new_events( self, origin: str, - room_id: str, event_infos: Iterable[_NewEventInfo], backfilled: bool = False, ) -> None: @@ -1948,7 +1944,6 @@ class FederationHandler(BaseHandler): ) await self.persist_events_and_notify( - room_id, [ (ev_info.event, context) for ev_info, context in zip(event_infos, contexts) @@ -1959,7 +1954,6 @@ class FederationHandler(BaseHandler): async def _persist_auth_tree( self, origin: str, - room_id: str, auth_events: List[EventBase], state: List[EventBase], event: EventBase, @@ -1974,7 +1968,6 @@ class FederationHandler(BaseHandler): Args: origin: Where the events came from - room_id, auth_events state event @@ -2049,20 +2042,17 @@ class FederationHandler(BaseHandler): events_to_context[e.event_id].rejected = RejectedReason.AUTH_ERROR await self.persist_events_and_notify( - room_id, [ (e, events_to_context[e.event_id]) for e in itertools.chain(auth_events, state) - ], + ] ) new_event_context = await self.state_handler.compute_event_context( event, old_state=state ) - return await self.persist_events_and_notify( - room_id, [(event, new_event_context)] - ) + return await self.persist_events_and_notify([(event, new_event_context)]) async def _prep_event( self, @@ -2913,7 +2903,6 @@ class FederationHandler(BaseHandler): async def persist_events_and_notify( self, - room_id: str, event_and_contexts: Sequence[Tuple[EventBase, EventContext]], backfilled: bool = False, ) -> int: @@ -2921,19 +2910,14 @@ class FederationHandler(BaseHandler): necessary. Args: - room_id: The room ID of events being persisted. - event_and_contexts: Sequence of events with their associated - context that should be persisted. All events must belong to - the same room. + event_and_contexts: backfilled: Whether these events are a result of backfilling or not """ - instance = self.config.worker.events_shard_config.get_instance(room_id) - if instance != self._instance_name: + if self.config.worker.writers.events != self._instance_name: result = await self._send_events( - instance_name=instance, + instance_name=self.config.worker.writers.events, store=self.store, - room_id=room_id, event_and_contexts=event_and_contexts, backfilled=backfilled, ) diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index 6398774c02..72bb638167 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -376,8 +376,9 @@ class EventCreationHandler(object): self.notifier = hs.get_notifier() self.config = hs.config self.require_membership_for_aliases = hs.config.require_membership_for_aliases - self._events_shard_config = self.config.worker.events_shard_config - self._instance_name = hs.get_instance_name() + self._is_event_writer = ( + self.config.worker.writers.events == hs.get_instance_name() + ) self.room_invite_state_types = self.hs.config.room_invite_state_types @@ -903,10 +904,9 @@ class EventCreationHandler(object): try: # If we're a worker we need to hit out to the master. - writer_instance = self._events_shard_config.get_instance(event.room_id) - if writer_instance != self._instance_name: + if not self._is_event_writer: result = await self.send_event( - instance_name=writer_instance, + instance_name=self.config.worker.writers.events, event_id=event.event_id, store=self.store, requester=requester, @@ -974,9 +974,7 @@ class EventCreationHandler(object): This should only be run on the instance in charge of persisting events. """ - assert self._events_shard_config.should_handle( - self._instance_name, event.room_id - ) + assert self._is_event_writer if ratelimit: # We check if this is a room admin redacting an event so that we diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index 55794c3057..9d5b1828df 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -804,9 +804,7 @@ class RoomCreationHandler(BaseHandler): # Always wait for room creation to progate before returning await self._replication.wait_for_stream_position( - self.hs.config.worker.events_shard_config.get_instance(room_id), - "events", - last_stream_id, + self.hs.config.worker.writers.events, "events", last_stream_id ) return result, last_stream_id @@ -1262,10 +1260,10 @@ class RoomShutdownHandler(object): # We now wait for the create room to come back in via replication so # that we can assume that all the joins/invites have propogated before # we try and auto join below. + # + # TODO: Currently the events stream is written to from master await self._replication.wait_for_stream_position( - self.hs.config.worker.events_shard_config.get_instance(new_room_id), - "events", - stream_id, + self.hs.config.worker.writers.events, "events", stream_id ) else: new_room_id = None @@ -1295,9 +1293,7 @@ class RoomShutdownHandler(object): # Wait for leave to come in over replication before trying to forget. await self._replication.wait_for_stream_position( - self.hs.config.worker.events_shard_config.get_instance(room_id), - "events", - stream_id, + self.hs.config.worker.writers.events, "events", stream_id ) await self.room_member_handler.forget(target_requester.user, room_id) diff --git a/synapse/handlers/room_member.py b/synapse/handlers/room_member.py index e5cb3c2d5c..a7962b0ada 100644 --- a/synapse/handlers/room_member.py +++ b/synapse/handlers/room_member.py @@ -82,6 +82,13 @@ class RoomMemberHandler(object): self._enable_lookup = hs.config.enable_3pid_lookup self.allow_per_room_profiles = self.config.allow_per_room_profiles + self._event_stream_writer_instance = hs.config.worker.writers.events + self._is_on_event_persistence_instance = ( + self._event_stream_writer_instance == hs.get_instance_name() + ) + if self._is_on_event_persistence_instance: + self.persist_event_storage = hs.get_storage().persistence + self._join_rate_limiter_local = Ratelimiter( clock=self.clock, rate_hz=hs.config.ratelimiting.rc_joins_local.per_second, diff --git a/synapse/replication/http/federation.py b/synapse/replication/http/federation.py index 5c8be747e1..6b56315148 100644 --- a/synapse/replication/http/federation.py +++ b/synapse/replication/http/federation.py @@ -65,11 +65,10 @@ class ReplicationFederationSendEventsRestServlet(ReplicationEndpoint): self.federation_handler = hs.get_handlers().federation_handler @staticmethod - async def _serialize_payload(store, room_id, event_and_contexts, backfilled): + async def _serialize_payload(store, event_and_contexts, backfilled): """ Args: store - room_id (str) event_and_contexts (list[tuple[FrozenEvent, EventContext]]) backfilled (bool): Whether or not the events are the result of backfilling @@ -89,11 +88,7 @@ class ReplicationFederationSendEventsRestServlet(ReplicationEndpoint): } ) - payload = { - "events": event_payloads, - "backfilled": backfilled, - "room_id": room_id, - } + payload = {"events": event_payloads, "backfilled": backfilled} return payload @@ -101,7 +96,6 @@ class ReplicationFederationSendEventsRestServlet(ReplicationEndpoint): with Measure(self.clock, "repl_fed_send_events_parse"): content = parse_json_object_from_request(request) - room_id = content["room_id"] backfilled = content["backfilled"] event_payloads = content["events"] @@ -126,7 +120,7 @@ class ReplicationFederationSendEventsRestServlet(ReplicationEndpoint): logger.info("Got %d events from federation", len(event_and_contexts)) max_stream_id = await self.federation_handler.persist_events_and_notify( - room_id, event_and_contexts, backfilled + event_and_contexts, backfilled ) return 200, {"max_stream_id": max_stream_id} diff --git a/synapse/replication/tcp/handler.py b/synapse/replication/tcp/handler.py index b323841f73..1c303f3a46 100644 --- a/synapse/replication/tcp/handler.py +++ b/synapse/replication/tcp/handler.py @@ -109,7 +109,7 @@ class ReplicationCommandHandler: if isinstance(stream, (EventsStream, BackfillStream)): # Only add EventStream and BackfillStream as a source on the # instance in charge of event persistence. - if hs.get_instance_name() in hs.config.worker.writers.events: + if hs.config.worker.writers.events == hs.get_instance_name(): self._streams_to_replicate.append(stream) continue diff --git a/synapse/replication/tcp/streams/events.py b/synapse/replication/tcp/streams/events.py index 3705618b4f..16c63ff4ec 100644 --- a/synapse/replication/tcp/streams/events.py +++ b/synapse/replication/tcp/streams/events.py @@ -19,7 +19,7 @@ from typing import List, Tuple, Type import attr -from ._base import Stream, StreamUpdateResult, Token +from ._base import Stream, StreamUpdateResult, Token, current_token_without_instance """Handling of the 'events' replication stream @@ -117,7 +117,7 @@ class EventsStream(Stream): self._store = hs.get_datastore() super().__init__( hs.get_instance_name(), - self._store._stream_id_gen.get_current_token_for_writer, + current_token_without_instance(self._store.get_current_events_token), self._update_function, ) diff --git a/synapse/storage/databases/__init__.py b/synapse/storage/databases/__init__.py index c73d54fb67..0ac854aee2 100644 --- a/synapse/storage/databases/__init__.py +++ b/synapse/storage/databases/__init__.py @@ -68,7 +68,7 @@ class Databases(object): # If we're on a process that can persist events also # instantiate a `PersistEventsStore` - if hs.get_instance_name() in hs.config.worker.writers.events: + if hs.config.worker.writers.events == hs.get_instance_name(): persist_events = PersistEventsStore(hs, database, main) if "state" in database_config.databases: diff --git a/synapse/storage/databases/main/event_federation.py b/synapse/storage/databases/main/event_federation.py index 4c3c162acf..0b69aa6a94 100644 --- a/synapse/storage/databases/main/event_federation.py +++ b/synapse/storage/databases/main/event_federation.py @@ -438,7 +438,7 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas """ if stream_ordering <= self.stream_ordering_month_ago: - raise StoreError(400, "stream_ordering too old %s" % (stream_ordering,)) + raise StoreError(400, "stream_ordering too old") sql = """ SELECT event_id FROM stream_ordering_to_exterm diff --git a/synapse/storage/databases/main/events.py b/synapse/storage/databases/main/events.py index b94fe7ac17..b3d27a2ee7 100644 --- a/synapse/storage/databases/main/events.py +++ b/synapse/storage/databases/main/events.py @@ -97,7 +97,6 @@ class PersistEventsStore: self.store = main_data_store self.database_engine = db.engine self._clock = hs.get_clock() - self._instance_name = hs.get_instance_name() self._ephemeral_messages_enabled = hs.config.enable_ephemeral_messages self.is_mine_id = hs.is_mine_id @@ -109,7 +108,7 @@ class PersistEventsStore: # This should only exist on instances that are configured to write assert ( - hs.get_instance_name() in hs.config.worker.writers.events + hs.config.worker.writers.events == hs.get_instance_name() ), "Can only instantiate EventsStore on master" async def _persist_events_and_state_updates( @@ -801,7 +800,6 @@ class PersistEventsStore: table="events", values=[ { - "instance_name": self._instance_name, "stream_ordering": event.internal_metadata.stream_ordering, "topological_ordering": event.depth, "depth": event.depth, diff --git a/synapse/storage/databases/main/events_worker.py b/synapse/storage/databases/main/events_worker.py index 17f5997b89..a7a73cc3d8 100644 --- a/synapse/storage/databases/main/events_worker.py +++ b/synapse/storage/databases/main/events_worker.py @@ -42,8 +42,7 @@ from synapse.replication.tcp.streams import BackfillStream from synapse.replication.tcp.streams.events import EventsStream from synapse.storage._base import SQLBaseStore, db_to_json, make_in_list_sql_clause from synapse.storage.database import DatabasePool -from synapse.storage.engines import PostgresEngine -from synapse.storage.util.id_generators import MultiWriterIdGenerator, StreamIdGenerator +from synapse.storage.util.id_generators import StreamIdGenerator from synapse.types import Collection, get_domain_from_id from synapse.util.caches.descriptors import Cache, cached from synapse.util.iterutils import batch_iter @@ -79,54 +78,27 @@ class EventsWorkerStore(SQLBaseStore): def __init__(self, database: DatabasePool, db_conn, hs): super(EventsWorkerStore, self).__init__(database, db_conn, hs) - if isinstance(database.engine, PostgresEngine): - # If we're using Postgres than we can use `MultiWriterIdGenerator` - # regardless of whether this process writes to the streams or not. - self._stream_id_gen = MultiWriterIdGenerator( - db_conn=db_conn, - db=database, - instance_name=hs.get_instance_name(), - table="events", - instance_column="instance_name", - id_column="stream_ordering", - sequence_name="events_stream_seq", + if hs.config.worker.writers.events == hs.get_instance_name(): + # We are the process in charge of generating stream ids for events, + # so instantiate ID generators based on the database + self._stream_id_gen = StreamIdGenerator( + db_conn, "events", "stream_ordering", ) - self._backfill_id_gen = MultiWriterIdGenerator( - db_conn=db_conn, - db=database, - instance_name=hs.get_instance_name(), - table="events", - instance_column="instance_name", - id_column="stream_ordering", - sequence_name="events_backfill_stream_seq", - positive=False, + self._backfill_id_gen = StreamIdGenerator( + db_conn, + "events", + "stream_ordering", + step=-1, + extra_tables=[("ex_outlier_stream", "event_stream_ordering")], ) else: - # We shouldn't be running in worker mode with SQLite, but its useful - # to support it for unit tests. - # - # If this process is the writer than we need to use - # `StreamIdGenerator`, otherwise we use `SlavedIdTracker` which gets - # updated over replication. (Multiple writers are not supported for - # SQLite). - if hs.get_instance_name() in hs.config.worker.writers.events: - self._stream_id_gen = StreamIdGenerator( - db_conn, "events", "stream_ordering", - ) - self._backfill_id_gen = StreamIdGenerator( - db_conn, - "events", - "stream_ordering", - step=-1, - extra_tables=[("ex_outlier_stream", "event_stream_ordering")], - ) - else: - self._stream_id_gen = SlavedIdTracker( - db_conn, "events", "stream_ordering" - ) - self._backfill_id_gen = SlavedIdTracker( - db_conn, "events", "stream_ordering", step=-1 - ) + # Another process is in charge of persisting events and generating + # stream IDs: rely on the replication streams to let us know which + # IDs we can process. + self._stream_id_gen = SlavedIdTracker(db_conn, "events", "stream_ordering") + self._backfill_id_gen = SlavedIdTracker( + db_conn, "events", "stream_ordering", step=-1 + ) self._get_event_cache = Cache( "*getEvent*", diff --git a/synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql b/synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql deleted file mode 100644 index 98ff76d709..0000000000 --- a/synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql +++ /dev/null @@ -1,16 +0,0 @@ -/* Copyright 2020 The Matrix.org Foundation C.I.C. - * - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * 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. - */ - -ALTER TABLE events ADD COLUMN instance_name TEXT; diff --git a/synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql.postgres b/synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql.postgres deleted file mode 100644 index 97c1e6a0c5..0000000000 --- a/synapse/storage/databases/main/schema/delta/58/14events_instance_name.sql.postgres +++ /dev/null @@ -1,26 +0,0 @@ -/* Copyright 2020 The Matrix.org Foundation C.I.C. - * - * Licensed under the Apache License, Version 2.0 (the "License"); - * you may not use this file except in compliance with the License. - * You may obtain a copy of the License at - * - * http://www.apache.org/licenses/LICENSE-2.0 - * - * Unless required by applicable law or agreed to in writing, software - * distributed under the License is distributed on an "AS IS" BASIS, - * 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. - */ - -CREATE SEQUENCE IF NOT EXISTS events_stream_seq; - -SELECT setval('events_stream_seq', ( - SELECT COALESCE(MAX(stream_ordering), 1) FROM events -)); - -CREATE SEQUENCE IF NOT EXISTS events_backfill_stream_seq; - -SELECT setval('events_backfill_stream_seq', ( - SELECT COALESCE(-MIN(stream_ordering), 1) FROM events -)); diff --git a/synapse/storage/util/id_generators.py b/synapse/storage/util/id_generators.py index 8fd21c2bf8..9f3d23f0a5 100644 --- a/synapse/storage/util/id_generators.py +++ b/synapse/storage/util/id_generators.py @@ -231,12 +231,8 @@ class MultiWriterIdGenerator: # gaps should be relatively rare it's still worth doing the book keeping # that allows us to skip forwards when there are gapless runs of # positions. - # - # We start at 1 here as a) the first generated stream ID will be 2, and - # b) other parts of the code assume that stream IDs are strictly greater - # than 0. self._persisted_upto_position = ( - min(self._current_positions.values()) if self._current_positions else 1 + min(self._current_positions.values()) if self._current_positions else 0 ) self._known_persisted_positions = [] # type: List[int] @@ -366,7 +362,9 @@ class MultiWriterIdGenerator: equal to it have been successfully persisted. """ - return self.get_persisted_upto_position() + # Currently we don't support this operation, as it's not obvious how to + # condense the stream positions of multiple writers into a single int. + raise NotImplementedError() def get_current_token_for_writer(self, instance_name: str) -> int: """Returns the position of the given writer. -- cgit 1.5.1