forked-synapse/synapse/replication/tcp/resource.py
Erik Johnston fef08cbee8
Fix sending out of order POSITION over replication (#16639)
If a worker reconnects to Redis we send out the current positions of all our streams. However, if we're also trying to send out a backlog of RDATA at the same time then we can end up sending a `POSITION` with the current token *before* we've sent all the RDATA before the current token.

This doesn't cause actual bugs as the receiving servers see the POSITION, fetch the relevant rows from the DB, and then ignore the old RDATA as they come in. However, this is inefficient so it'd be better if we didn't  send out-of-order positions
2023-11-16 13:05:09 +00:00

329 lines
14 KiB
Python

# Copyright 2017 Vector Creations Ltd
#
# 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.
"""The server side of the replication stream.
"""
import logging
import random
from typing import TYPE_CHECKING, List, Optional, Tuple
from prometheus_client import Counter
from twisted.internet.interfaces import IAddress
from twisted.internet.protocol import ServerFactory
from synapse.metrics.background_process_metrics import run_as_background_process
from synapse.replication.tcp.commands import PositionCommand
from synapse.replication.tcp.protocol import ServerReplicationStreamProtocol
from synapse.replication.tcp.streams import EventsStream
from synapse.replication.tcp.streams._base import CachesStream, StreamRow, Token
from synapse.util.metrics import Measure
if TYPE_CHECKING:
from synapse.server import HomeServer
stream_updates_counter = Counter(
"synapse_replication_tcp_resource_stream_updates", "", ["stream_name"]
)
logger = logging.getLogger(__name__)
class ReplicationStreamProtocolFactory(ServerFactory):
"""Factory for new replication connections."""
def __init__(self, hs: "HomeServer"):
self.command_handler = hs.get_replication_command_handler()
self.clock = hs.get_clock()
self.server_name = hs.config.server.server_name
# If we've created a `ReplicationStreamProtocolFactory` then we're
# almost certainly registering a replication listener, so let's ensure
# that we've started a `ReplicationStreamer` instance to actually push
# data.
#
# (This is a bit of a weird place to do this, but the alternatives such
# as putting this in `HomeServer.setup()`, requires either passing the
# listener config again or always starting a `ReplicationStreamer`.)
hs.get_replication_streamer()
def buildProtocol(self, addr: IAddress) -> ServerReplicationStreamProtocol:
return ServerReplicationStreamProtocol(
self.server_name, self.clock, self.command_handler
)
class ReplicationStreamer:
"""Handles replication connections.
This needs to be poked when new replication data may be available.
When new data is available it will propagate to all Redis subscribers.
"""
def __init__(self, hs: "HomeServer"):
self.store = hs.get_datastores().main
self.clock = hs.get_clock()
self.notifier = hs.get_notifier()
self._instance_name = hs.get_instance_name()
self._replication_torture_level = hs.config.server.replication_torture_level
self.notifier.add_replication_callback(self.on_notifier_poke)
# Keeps track of whether we are currently checking for updates
self.is_looping = False
self.pending_updates = False
self.command_handler = hs.get_replication_command_handler()
# Set of streams to replicate.
self.streams = self.command_handler.get_streams_to_replicate()
# If we have streams then we must have redis enabled or on master
assert (
not self.streams
or hs.config.redis.redis_enabled
or not hs.config.worker.worker_app
)
# If we are replicating an event stream we want to periodically check if
# we should send updated POSITIONs. We do this as a looping call rather
# explicitly poking when the position advances (without new data to
# replicate) to reduce replication traffic (otherwise each writer would
# likely send a POSITION for each new event received over replication).
#
# Note that if the position hasn't advanced then we won't send anything.
if any(EventsStream.NAME == s.NAME for s in self.streams):
self.clock.looping_call(self.on_notifier_poke, 1000)
def on_notifier_poke(self) -> None:
"""Checks if there is actually any new data and sends it to the
Redis subscribers if there are.
This should get called each time new data is available, even if it
is currently being executed, so that nothing gets missed
"""
if not self.command_handler.connected() or not self.streams:
# Don't bother if nothing is listening. We still need to advance
# the stream tokens otherwise they'll fall behind forever
for stream in self.streams:
stream.discard_updates_and_advance()
return
# We check up front to see if anything has actually changed, as we get
# poked because of changes that happened on other instances.
if not self.command_handler.should_announce_positions() and all(
stream.last_token == stream.current_token(self._instance_name)
for stream in self.streams
):
return
# If there are updates then we need to set this even if we're already
# looping, as the loop needs to know that he might need to loop again.
self.pending_updates = True
if self.is_looping:
logger.debug("Notifier poke loop already running")
return
run_as_background_process("replication_notifier", self._run_notifier_loop)
async def _run_notifier_loop(self) -> None:
self.is_looping = True
try:
# Keep looping while there have been pokes about potential updates.
# This protects against the race where a stream we already checked
# gets an update while we're handling other streams.
while self.pending_updates:
self.pending_updates = False
with Measure(self.clock, "repl.stream.get_updates"):
all_streams = self.streams
if self._replication_torture_level is not None:
# there is no guarantee about ordering between the streams,
# so let's shuffle them around a bit when we are in torture mode.
all_streams = list(all_streams)
random.shuffle(all_streams)
if self.command_handler.should_announce_positions():
# We need to send out POSITIONs for all streams, usually
# because a worker has reconnected.
self.command_handler.will_announce_positions()
for stream in all_streams:
self.command_handler.send_command(
PositionCommand(
stream.NAME,
self._instance_name,
stream.last_token,
stream.last_token,
)
)
for stream in all_streams:
if stream.last_token == stream.current_token(
self._instance_name
):
continue
if self._replication_torture_level:
await self.clock.sleep(
self._replication_torture_level / 1000.0
)
last_token = stream.last_token
logger.debug(
"Getting stream: %s: %s -> %s",
stream.NAME,
stream.last_token,
stream.current_token(self._instance_name),
)
try:
updates, current_token, limited = await stream.get_updates()
self.pending_updates |= limited
except Exception:
logger.info("Failed to handle stream %s", stream.NAME)
raise
logger.debug(
"Sending %d updates",
len(updates),
)
if updates:
logger.info(
"Streaming: %s -> %s (limited: %s, updates: %s, max token: %s)",
stream.NAME,
updates[-1][0],
limited,
len(updates),
current_token,
)
stream_updates_counter.labels(stream.NAME).inc(len(updates))
else:
# The token has advanced but there is no data to
# send, so we send a `POSITION` to inform other
# workers of the updated position.
#
# There are two reasons for this: 1) this instance
# requested a stream ID but didn't use it, or 2)
# this instance advanced its own stream position due
# to receiving notifications about other instances
# advancing their stream position.
# We skip sending `POSITION` for the `caches` stream
# for the second case as a) it generates a lot of
# traffic as every worker would echo each write, and
# b) nothing cares if a given worker's caches stream
# position lags.
if stream.NAME == CachesStream.NAME:
# If there haven't been any writes since the
# `last_token` then we're in the second case.
if stream.minimal_local_current_token() <= last_token:
continue
# Note: `last_token` may not *actually* be the
# last token we sent out in a RDATA or POSITION.
# This can happen if we sent out an RDATA for
# position X when our current token was say X+1.
# Other workers will see RDATA for X and then a
# POSITION with last token of X+1, which will
# cause them to check if there were any missing
# updates between X and X+1.
logger.info(
"Sending position: %s -> %s",
stream.NAME,
current_token,
)
self.command_handler.send_command(
PositionCommand(
stream.NAME,
self._instance_name,
last_token,
current_token,
)
)
continue
# Some streams return multiple rows with the same stream IDs,
# we need to make sure they get sent out in batches. We do
# this by setting the current token to all but the last of
# a series of updates with the same token to have a None
# token. See RdataCommand for more details.
batched_updates = _batch_updates(updates)
for token, row in batched_updates:
try:
self.command_handler.stream_update(
stream.NAME, token, row
)
except Exception:
logger.exception("Failed to replicate")
# The last token we send may not match the current
# token, in which case we want to send out a `POSITION`
# to tell other workers the actual current position.
if updates[-1][0] < current_token:
logger.info(
"Sending position: %s -> %s",
stream.NAME,
current_token,
)
self.command_handler.send_command(
PositionCommand(
stream.NAME,
self._instance_name,
updates[-1][0],
current_token,
)
)
logger.debug("No more pending updates, breaking poke loop")
finally:
self.pending_updates = False
self.is_looping = False
def _batch_updates(
updates: List[Tuple[Token, StreamRow]]
) -> List[Tuple[Optional[Token], StreamRow]]:
"""Takes a list of updates of form [(token, row)] and sets the token to
None for all rows where the next row has the same token. This is used to
implement batching.
For example:
[(1, _), (1, _), (2, _), (3, _), (3, _)]
becomes:
[(None, _), (1, _), (2, _), (None, _), (3, _)]
"""
if not updates:
return []
new_updates: List[Tuple[Optional[Token], StreamRow]] = []
for i, update in enumerate(updates[:-1]):
if update[0] == updates[i + 1][0]:
new_updates.append((None, update[1]))
else:
new_updates.append(update)
new_updates.append(updates[-1])
return new_updates