summary refs log tree commit diff
path: root/synapse/replication/tcp/protocol.py
blob: 0a9a290af40afed9820a034aa4d975a055e04d83 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
# -*- coding: utf-8 -*-
# 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.
"""This module contains the implementation of both the client and server
protocols.

The basic structure of the protocol is line based, where the initial word of
each line specifies the command. The rest of the line is parsed based on the
command. For example, the `RDATA` command is defined as::

    RDATA <stream_name> <token> <row_json>

(Note that `<row_json>` may contains spaces, but cannot contain newlines.)

Blank lines are ignored.

# Example

An example iteraction is shown below. Each line is prefixed with '>' or '<' to
indicate which side is sending, these are *not* included on the wire::

    * connection established *
    > SERVER localhost:8823
    > PING 1490197665618
    < NAME synapse.app.appservice
    < PING 1490197665618
    < REPLICATE events 1
    < REPLICATE backfill 1
    < REPLICATE caches 1
    > POSITION events 1
    > POSITION backfill 1
    > POSITION caches 1
    > RDATA caches 2 ["get_user_by_id",["@01register-user:localhost:8823"],1490197670513]
    > RDATA events 14 ["$149019767112vOHxz:localhost:8823",
        "!AFDCvgApUmpdfVjIXm:localhost:8823","m.room.guest_access","",null]
    < PING 1490197675618
    > ERROR server stopping
    * connection closed by server *
"""

from twisted.internet import defer
from twisted.protocols.basic import LineOnlyReceiver
from twisted.python.failure import Failure

from commands import (
    COMMAND_MAP, VALID_CLIENT_COMMANDS, VALID_SERVER_COMMANDS,
    ErrorCommand, ServerCommand, RdataCommand, PositionCommand, PingCommand,
    NameCommand, ReplicateCommand, UserSyncCommand, SyncCommand,
)
from streams import STREAMS_MAP

from synapse.util.stringutils import random_string
from synapse.metrics.metric import CounterMetric

import logging
import synapse.metrics
import struct
import fcntl


metrics = synapse.metrics.get_metrics_for(__name__)

connection_close_counter = metrics.register_counter(
    "close_reason", labels=["reason_type"],
)


# A list of all connected protocols. This allows us to send metrics about the
# connections.
connected_connections = []


logger = logging.getLogger(__name__)


PING_TIME = 5000
PING_TIMEOUT_MULTIPLIER = 5
PING_TIMEOUT_MS = PING_TIME * PING_TIMEOUT_MULTIPLIER


class ConnectionStates(object):
    CONNECTING = "connecting"
    ESTABLISHED = "established"
    PAUSED = "paused"
    CLOSED = "closed"


class BaseReplicationStreamProtocol(LineOnlyReceiver):
    """Base replication protocol shared between client and server.

    Reads lines (ignoring blank ones) and parses them into command classes,
    asserting that they are valid for the given direction, i.e. server commands
    are only sent by the server.

    On receiving a new command it calls `on_<COMMAND_NAME>` with the parsed
    command.

    It also sends `PING` periodically, and correctly times out remote connections
    (if they send a `PING` command)
    """
    delimiter = b'\n'

    VALID_INBOUND_COMMANDS = []  # Valid commands we expect to receive
    VALID_OUTBOUND_COMMANDS = []  # Valid commans we can send

    max_line_buffer = 10000

    def __init__(self, clock):
        self.clock = clock

        self.last_received_command = self.clock.time_msec()
        self.last_sent_command = 0
        self.time_we_closed = None  # When we requested the connection be closed

        self.received_ping = False  # Have we reecived a ping from the other side

        self.state = ConnectionStates.CONNECTING

        self.name = "anon"  # The name sent by a client.
        self.conn_id = random_string(5)  # To dedupe in case of name clashes.

        # List of pending commands to send once we've established the connection
        self.pending_commands = []

        # The LoopingCall for sending pings.
        self._send_ping_loop = None

        self.inbound_commands_counter = CounterMetric(
            "inbound_commands", labels=["command"],
        )
        self.outbound_commands_counter = CounterMetric(
            "outbound_commands", labels=["command"],
        )

    def connectionMade(self):
        logger.info("[%s] Connection established", self.id())

        self.state = ConnectionStates.ESTABLISHED

        connected_connections.append(self)  # Register connection for metrics

        self.transport.registerProducer(self, True)  # For the *Producing callbacks

        self._send_pending_commands()

        # Starts sending pings
        self._send_ping_loop = self.clock.looping_call(self.send_ping, 5000)

        # Always send the initial PING so that the other side knows that they
        # can time us out.
        self.send_command(PingCommand(self.clock.time_msec()))

    def send_ping(self):
        """Periodically sends a ping and checks if we should close the connection
        due to the other side timing out.
        """
        now = self.clock.time_msec()

        if self.time_we_closed:
            if now - self.time_we_closed > PING_TIMEOUT_MS:
                logger.info(
                    "[%s] Failed to close connection gracefully, aborting", self.id()
                )
                self.transport.abortConnection()
        else:
            if now - self.last_sent_command >= PING_TIME:
                self.send_command(PingCommand(now))

            if self.received_ping and now - self.last_received_command > PING_TIMEOUT_MS:
                logger.info(
                    "[%s] Connection hasn't received command in %r ms. Closing.",
                    self.id(), now - self.last_received_command
                )
                self.send_error("ping timeout")

    def lineReceived(self, line):
        """Called when we've received a line
        """
        if line.strip() == "":
            # Ignore blank lines
            return

        line = line.decode("utf-8")
        cmd_name, rest_of_line = line.split(" ", 1)

        if cmd_name not in self.VALID_INBOUND_COMMANDS:
            logger.error("[%s] invalid command %s", self.id(), cmd_name)
            self.send_error("invalid command: %s", cmd_name)
            return

        self.last_received_command = self.clock.time_msec()

        self.inbound_commands_counter.inc(cmd_name)

        cmd_cls = COMMAND_MAP[cmd_name]
        try:
            cmd = cmd_cls.from_line(rest_of_line)
        except Exception as e:
            logger.exception(
                "[%s] failed to parse line %r: %r", self.id(), cmd_name, rest_of_line
            )
            self.send_error(
                "failed to parse line for  %r: %r (%r):" % (cmd_name, e, rest_of_line)
            )
            return

        # Now lets try and call on_<CMD_NAME> function
        try:
            getattr(self, "on_%s" % (cmd_name,))(cmd)
        except Exception:
            logger.exception("[%s] Failed to handle line: %r", self.id(), line)

    def close(self):
        logger.warn("[%s] Closing connection", self.id())
        self.time_we_closed = self.clock.time_msec()
        self.transport.loseConnection()
        self.on_connection_closed()

    def send_error(self, error_string, *args):
        """Send an error to remote and close the connection.
        """
        self.send_command(ErrorCommand(error_string % args))
        self.close()

    def send_command(self, cmd, do_buffer=True):
        """Send a command if connection has been established.

        Args:
            cmd (Command)
            do_buffer (bool): Whether to buffer the message or always attempt
                to send the command. This is mostly used to send an error
                message if we're about to close the connection due our buffers
                becoming full.
        """
        if self.state == ConnectionStates.CLOSED:
            logger.debug("[%s] Not sending, connection closed", self.id())
            return

        if do_buffer and self.state != ConnectionStates.ESTABLISHED:
            self._queue_command(cmd)
            return

        self.outbound_commands_counter.inc(cmd.NAME)

        string = "%s %s" % (cmd.NAME, cmd.to_line(),)
        if "\n" in string:
            raise Exception("Unexpected newline in command: %r", string)

        self.sendLine(string.encode("utf-8"))

        self.last_sent_command = self.clock.time_msec()

    def _queue_command(self, cmd):
        """Queue the command until the connection is ready to write to again.
        """
        logger.debug("[%s] Queing as conn %r, cmd: %r", self.id(), self.state, cmd)
        self.pending_commands.append(cmd)

        if len(self.pending_commands) > self.max_line_buffer:
            # The other side is failing to keep up and out buffers are becoming
            # full, so lets close the connection.
            # XXX: should we squawk more loudly?
            logger.error("[%s] Remote failed to keep up", self.id())
            self.send_command(ErrorCommand("Failed to keep up"), do_buffer=False)
            self.close()

    def _send_pending_commands(self):
        """Send any queued commandes
        """
        pending = self.pending_commands
        self.pending_commands = []
        for cmd in pending:
            self.send_command(cmd)

    def on_PING(self, line):
        self.received_ping = True

    def on_ERROR(self, cmd):
        logger.error("[%s] Remote reported error: %r", self.id(), cmd.data)

    def pauseProducing(self):
        """This is called when both the kernel send buffer and the twisted
        tcp connection send buffers have become full.

        We don't actually have any control over those sizes, so we buffer some
        commands ourselves before knifing the connection due to the remote
        failing to keep up.
        """
        logger.info("[%s] Pause producing", self.id())
        self.state = ConnectionStates.PAUSED

    def resumeProducing(self):
        """The remote has caught up after we started buffering!
        """
        logger.info("[%s] Resume producing", self.id())
        self.state = ConnectionStates.ESTABLISHED
        self._send_pending_commands()

    def stopProducing(self):
        """We're never going to send any more data (normally because either
        we or the remote has closed the connection)
        """
        logger.info("[%s] Stop producing", self.id())
        self.on_connection_closed()

    def connectionLost(self, reason):
        logger.info("[%s] Replication connection closed: %r", self.id(), reason)
        if isinstance(reason, Failure):
            connection_close_counter.inc(reason.type.__name__)
        else:
            connection_close_counter.inc(reason.__class__.__name__)

        try:
            # Remove us from list of connections to be monitored
            connected_connections.remove(self)
        except ValueError:
            pass

        # Stop the looping call sending pings.
        if self._send_ping_loop and self._send_ping_loop.running:
            self._send_ping_loop.stop()

        self.on_connection_closed()

    def on_connection_closed(self):
        logger.info("[%s] Connection was closed", self.id())

        self.state = ConnectionStates.CLOSED
        self.pending_commands = []

        if self.transport:
            self.transport.unregisterProducer()

    def __str__(self):
        return "ReplicationConnection<name=%s,conn_id=%s,addr=%s>" % (
            self.name, self.conn_id, self.addr,
        )

    def id(self):
        return "%s-%s" % (self.name, self.conn_id)


class ServerReplicationStreamProtocol(BaseReplicationStreamProtocol):
    VALID_INBOUND_COMMANDS = VALID_CLIENT_COMMANDS
    VALID_OUTBOUND_COMMANDS = VALID_SERVER_COMMANDS

    def __init__(self, server_name, clock, streamer, addr):
        BaseReplicationStreamProtocol.__init__(self, clock)  # Old style class

        self.server_name = server_name
        self.streamer = streamer
        self.addr = addr

        # The streams the client has subscribed to and is up to date with
        self.replication_streams = set()

        # The streams the client is currently subscribing to.
        self.connecting_streams = set()

        # Map from stream name to list of updates to send once we've finished
        # subscribing the client to the stream.
        self.pending_rdata = {}

    def connectionMade(self):
        self.send_command(ServerCommand(self.server_name))
        BaseReplicationStreamProtocol.connectionMade(self)
        self.streamer.new_connection(self)

    def on_NAME(self, cmd):
        logger.info("[%s] Renamed to %r", self.id(), cmd.data)
        self.name = cmd.data

    def on_USER_SYNC(self, cmd):
        self.streamer.on_user_sync(
            self.conn_id, cmd.user_id, cmd.is_syncing, cmd.last_sync_ms,
        )

    def on_REPLICATE(self, cmd):
        stream_name = cmd.stream_name
        token = cmd.token

        if stream_name == "ALL":
            # Subscribe to all streams we're publishing to.
            for stream in self.streamer.streams_by_name.iterkeys():
                self.subscribe_to_stream(stream, token)
        else:
            self.subscribe_to_stream(stream_name, token)

    def on_FEDERATION_ACK(self, cmd):
        self.streamer.federation_ack(cmd.token)

    def on_REMOVE_PUSHER(self, cmd):
        self.streamer.on_remove_pusher(cmd.app_id, cmd.push_key, cmd.user_id)

    def on_INVALIDATE_CACHE(self, cmd):
        self.streamer.on_invalidate_cache(cmd.cache_func, cmd.keys)

    def on_USER_IP(self, cmd):
        self.streamer.on_user_ip(
            cmd.user_id, cmd.access_token, cmd.ip, cmd.user_agent, cmd.device_id,
            cmd.last_seen,
        )

    @defer.inlineCallbacks
    def subscribe_to_stream(self, stream_name, token):
        """Subscribe the remote to a streams.

        This invloves checking if they've missed anything and sending those
        updates down if they have. During that time new updates for the stream
        are queued and sent once we've sent down any missed updates.
        """
        self.replication_streams.discard(stream_name)
        self.connecting_streams.add(stream_name)

        try:
            # Get missing updates
            updates, current_token = yield self.streamer.get_stream_updates(
                stream_name, token,
            )

            # Send all the missing updates
            for update in updates:
                token, row = update[0], update[1]
                self.send_command(RdataCommand(stream_name, token, row))

            # We send a POSITION command to ensure that they have an up to
            # date token (especially useful if we didn't send any updates
            # above)
            self.send_command(PositionCommand(stream_name, current_token))

            # Now we can send any updates that came in while we were subscribing
            pending_rdata = self.pending_rdata.pop(stream_name, [])
            for token, update in pending_rdata:
                # Only send updates newer than the current token
                if token > current_token:
                    self.send_command(RdataCommand(stream_name, token, update))

            # They're now fully subscribed
            self.replication_streams.add(stream_name)
        except Exception as e:
            logger.exception("[%s] Failed to handle REPLICATE command", self.id())
            self.send_error("failed to handle replicate: %r", e)
        finally:
            self.connecting_streams.discard(stream_name)

    def stream_update(self, stream_name, token, data):
        """Called when a new update is available to stream to clients.

        We need to check if the client is interested in the stream or not
        """
        if stream_name in self.replication_streams:
            # The client is subscribed to the stream
            self.send_command(RdataCommand(stream_name, token, data))
        elif stream_name in self.connecting_streams:
            # The client is being subscribed to the stream
            logger.debug("[%s] Queuing RDATA %r %r", self.id(), stream_name, token)
            self.pending_rdata.setdefault(stream_name, []).append((token, data))
        else:
            # The client isn't subscribed
            logger.debug("[%s] Dropping RDATA %r %r", self.id(), stream_name, token)

    def send_sync(self, data):
        self.send_command(SyncCommand(data))

    def on_connection_closed(self):
        BaseReplicationStreamProtocol.on_connection_closed(self)
        self.streamer.lost_connection(self)


class ClientReplicationStreamProtocol(BaseReplicationStreamProtocol):
    VALID_INBOUND_COMMANDS = VALID_SERVER_COMMANDS
    VALID_OUTBOUND_COMMANDS = VALID_CLIENT_COMMANDS

    def __init__(self, client_name, server_name, clock, handler):
        BaseReplicationStreamProtocol.__init__(self, clock)

        self.client_name = client_name
        self.server_name = server_name
        self.handler = handler

        # Map of stream to batched updates. See RdataCommand for info on how
        # batching works.
        self.pending_batches = {}

    def connectionMade(self):
        self.send_command(NameCommand(self.client_name))
        BaseReplicationStreamProtocol.connectionMade(self)

        # Once we've connected subscribe to the necessary streams
        for stream_name, token in self.handler.get_streams_to_replicate().iteritems():
            self.replicate(stream_name, token)

        # Tell the server if we have any users currently syncing (should only
        # happen on synchrotrons)
        currently_syncing = self.handler.get_currently_syncing_users()
        now = self.clock.time_msec()
        for user_id in currently_syncing:
            self.send_command(UserSyncCommand(user_id, True, now))

        # We've now finished connecting to so inform the client handler
        self.handler.update_connection(self)

    def on_SERVER(self, cmd):
        if cmd.data != self.server_name:
            logger.error("[%s] Connected to wrong remote: %r", self.id(), cmd.data)
            self.send_error("Wrong remote")

    def on_RDATA(self, cmd):
        stream_name = cmd.stream_name
        inbound_rdata_count.inc(stream_name)

        try:
            row = STREAMS_MAP[stream_name].ROW_TYPE(*cmd.row)
        except Exception:
            logger.exception(
                "[%s] Failed to parse RDATA: %r %r",
                self.id(), stream_name, cmd.row
            )
            raise

        if cmd.token is None:
            # I.e. this is part of a batch of updates for this stream. Batch
            # until we get an update for the stream with a non None token
            self.pending_batches.setdefault(stream_name, []).append(row)
        else:
            # Check if this is the last of a batch of updates
            rows = self.pending_batches.pop(stream_name, [])
            rows.append(row)

            self.handler.on_rdata(stream_name, cmd.token, rows)

    def on_POSITION(self, cmd):
        self.handler.on_position(cmd.stream_name, cmd.token)

    def on_SYNC(self, cmd):
        self.handler.on_sync(cmd.data)

    def replicate(self, stream_name, token):
        """Send the subscription request to the server
        """
        if stream_name not in STREAMS_MAP:
            raise Exception("Invalid stream name %r" % (stream_name,))

        logger.info(
            "[%s] Subscribing to replication stream: %r from %r",
            self.id(), stream_name, token
        )

        self.send_command(ReplicateCommand(stream_name, token))

    def on_connection_closed(self):
        BaseReplicationStreamProtocol.on_connection_closed(self)
        self.handler.update_connection(None)


# The following simply registers metrics for the replication connections

metrics.register_callback(
    "pending_commands",
    lambda: {
        (p.name, p.conn_id): len(p.pending_commands)
        for p in connected_connections
    },
    labels=["name", "conn_id"],
)


def transport_buffer_size(protocol):
    if protocol.transport:
        size = len(protocol.transport.dataBuffer) + protocol.transport._tempDataLen
        return size
    return 0


metrics.register_callback(
    "transport_send_buffer",
    lambda: {
        (p.name, p.conn_id): transport_buffer_size(p)
        for p in connected_connections
    },
    labels=["name", "conn_id"],
)


def transport_kernel_read_buffer_size(protocol, read=True):
    SIOCINQ = 0x541B
    SIOCOUTQ = 0x5411

    if protocol.transport:
        fileno = protocol.transport.getHandle().fileno()
        if read:
            op = SIOCINQ
        else:
            op = SIOCOUTQ
        size = struct.unpack("I", fcntl.ioctl(fileno, op, '\0\0\0\0'))[0]
        return size
    return 0


metrics.register_callback(
    "transport_kernel_send_buffer",
    lambda: {
        (p.name, p.conn_id): transport_kernel_read_buffer_size(p, False)
        for p in connected_connections
    },
    labels=["name", "conn_id"],
)


metrics.register_callback(
    "transport_kernel_read_buffer",
    lambda: {
        (p.name, p.conn_id): transport_kernel_read_buffer_size(p, True)
        for p in connected_connections
    },
    labels=["name", "conn_id"],
)


metrics.register_callback(
    "inbound_commands",
    lambda: {
        (k[0], p.name, p.conn_id): count
        for p in connected_connections
        for k, count in p.inbound_commands_counter.counts.iteritems()
    },
    labels=["command", "name", "conn_id"],
)

metrics.register_callback(
    "outbound_commands",
    lambda: {
        (k[0], p.name, p.conn_id): count
        for p in connected_connections
        for k, count in p.outbound_commands_counter.counts.iteritems()
    },
    labels=["command", "name", "conn_id"],
)

# number of updates received for each RDATA stream
inbound_rdata_count = metrics.register_counter(
    "inbound_rdata_count",
    labels=["stream_name"],
)