summary refs log tree commit diff
path: root/tests/replication/test_sharded_event_persister.py
blob: 0a6e4795ee922d03858c50d79f67607194d35442 (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
# 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.
import logging
from unittest.mock import patch

from synapse.api.room_versions import RoomVersion
from synapse.rest import admin
from synapse.rest.client import login, room, sync

from tests.replication._base import BaseMultiWorkerStreamTestCase
from tests.server import make_request
from tests.utils import USE_POSTGRES_FOR_TESTS

logger = logging.getLogger(__name__)


class EventPersisterShardTestCase(BaseMultiWorkerStreamTestCase):
    """Checks event persisting sharding works"""

    # Event persister sharding requires postgres (due to needing
    # `MultiWriterIdGenerator`).
    if not USE_POSTGRES_FOR_TESTS:
        skip = "Requires Postgres"

    servlets = [
        admin.register_servlets_for_client_rest_resource,
        room.register_servlets,
        login.register_servlets,
        sync.register_servlets,
    ]

    def prepare(self, reactor, clock, hs):
        # Register a user who sends a message that we'll get notified about
        self.other_user_id = self.register_user("otheruser", "pass")
        self.other_access_token = self.login("otheruser", "pass")

        self.room_creator = self.hs.get_room_creation_handler()
        self.store = hs.get_datastore()

    def default_config(self):
        conf = super().default_config()
        conf["redis"] = {"enabled": "true"}
        conf["stream_writers"] = {"events": ["worker1", "worker2"]}
        conf["instance_map"] = {
            "worker1": {"host": "testserv", "port": 1001},
            "worker2": {"host": "testserv", "port": 1002},
        }
        return conf

    def _create_room(self, room_id: str, user_id: str, tok: str):
        """Create a room with given room_id"""

        # We control the room ID generation by patching out the
        # `_generate_room_id` method
        async def generate_room(
            creator_id: str, is_public: bool, room_version: RoomVersion
        ):
            await self.store.store_room(
                room_id=room_id,
                room_creator_user_id=creator_id,
                is_public=is_public,
                room_version=room_version,
            )
            return room_id

        with patch(
            "synapse.handlers.room.RoomCreationHandler._generate_room_id"
        ) as mock:
            mock.side_effect = generate_room
            self.helper.create_room_as(user_id, tok=tok)

    def test_basic(self):
        """Simple test to ensure that multiple rooms can be created and joined,
        and that different rooms get handled by different instances.
        """

        self.make_worker_hs(
            "synapse.app.generic_worker",
            {"worker_name": "worker1"},
        )

        self.make_worker_hs(
            "synapse.app.generic_worker",
            {"worker_name": "worker2"},
        )

        persisted_on_1 = False
        persisted_on_2 = False

        store = self.hs.get_datastore()

        user_id = self.register_user("user", "pass")
        access_token = self.login("user", "pass")

        # Keep making new rooms until we see rooms being persisted on both
        # workers.
        for _ in range(10):
            # Create a room
            room = self.helper.create_room_as(user_id, tok=access_token)

            # The other user joins
            self.helper.join(
                room=room, user=self.other_user_id, tok=self.other_access_token
            )

            # The other user sends some messages
            rseponse = self.helper.send(room, body="Hi!", tok=self.other_access_token)
            event_id = rseponse["event_id"]

            # The event position includes which instance persisted the event.
            pos = self.get_success(store.get_position_for_event(event_id))

            persisted_on_1 |= pos.instance_name == "worker1"
            persisted_on_2 |= pos.instance_name == "worker2"

            if persisted_on_1 and persisted_on_2:
                break

        self.assertTrue(persisted_on_1)
        self.assertTrue(persisted_on_2)

    def test_vector_clock_token(self):
        """Tests that using a stream token with a vector clock component works
        correctly with basic /sync and /messages usage.
        """

        self.make_worker_hs(
            "synapse.app.generic_worker",
            {"worker_name": "worker1"},
        )

        worker_hs2 = self.make_worker_hs(
            "synapse.app.generic_worker",
            {"worker_name": "worker2"},
        )

        sync_hs = self.make_worker_hs(
            "synapse.app.generic_worker",
            {"worker_name": "sync"},
        )
        sync_hs_site = self._hs_to_site[sync_hs]

        # Specially selected room IDs that get persisted on different workers.
        room_id1 = "!foo:test"
        room_id2 = "!baz:test"

        self.assertEqual(
            self.hs.config.worker.events_shard_config.get_instance(room_id1), "worker1"
        )
        self.assertEqual(
            self.hs.config.worker.events_shard_config.get_instance(room_id2), "worker2"
        )

        user_id = self.register_user("user", "pass")
        access_token = self.login("user", "pass")

        store = self.hs.get_datastore()

        # Create two room on the different workers.
        self._create_room(room_id1, user_id, access_token)
        self._create_room(room_id2, user_id, access_token)

        # The other user joins
        self.helper.join(
            room=room_id1, user=self.other_user_id, tok=self.other_access_token
        )
        self.helper.join(
            room=room_id2, user=self.other_user_id, tok=self.other_access_token
        )

        # Do an initial sync so that we're up to date.
        channel = make_request(
            self.reactor, sync_hs_site, "GET", "/sync", access_token=access_token
        )
        next_batch = channel.json_body["next_batch"]

        # We now gut wrench into the events stream MultiWriterIdGenerator on
        # worker2 to mimic it getting stuck persisting an event. This ensures
        # that when we send an event on worker1 we end up in a state where
        # worker2 events stream position lags that on worker1, resulting in a
        # RoomStreamToken with a non-empty instance map component.
        #
        # Worker2's event stream position will not advance until we call
        # __aexit__ again.
        actx = worker_hs2.get_datastore()._stream_id_gen.get_next()
        self.get_success(actx.__aenter__())

        response = self.helper.send(room_id1, body="Hi!", tok=self.other_access_token)
        first_event_in_room1 = response["event_id"]

        # Assert that the current stream token has an instance map component, as
        # we are trying to test vector clock tokens.
        room_stream_token = store.get_room_max_token()
        self.assertNotEqual(len(room_stream_token.instance_map), 0)

        # Check that syncing still gets the new event, despite the gap in the
        # stream IDs.
        channel = make_request(
            self.reactor,
            sync_hs_site,
            "GET",
            f"/sync?since={next_batch}",
            access_token=access_token,
        )

        # We should only see the new event and nothing else
        self.assertIn(room_id1, channel.json_body["rooms"]["join"])
        self.assertNotIn(room_id2, channel.json_body["rooms"]["join"])

        events = channel.json_body["rooms"]["join"][room_id1]["timeline"]["events"]
        self.assertListEqual(
            [first_event_in_room1], [event["event_id"] for event in events]
        )

        # Get the next batch and makes sure its a vector clock style token.
        vector_clock_token = channel.json_body["next_batch"]
        self.assertTrue(vector_clock_token.startswith("m"))

        # Now that we've got a vector clock token we finish the fake persisting
        # an event we started above.
        self.get_success(actx.__aexit__(None, None, None))

        # Now try and send an event to the other rooom so that we can test that
        # the vector clock style token works as a `since` token.
        response = self.helper.send(room_id2, body="Hi!", tok=self.other_access_token)
        first_event_in_room2 = response["event_id"]

        channel = make_request(
            self.reactor,
            sync_hs_site,
            "GET",
            f"/sync?since={vector_clock_token}",
            access_token=access_token,
        )

        self.assertNotIn(room_id1, channel.json_body["rooms"]["join"])
        self.assertIn(room_id2, channel.json_body["rooms"]["join"])

        events = channel.json_body["rooms"]["join"][room_id2]["timeline"]["events"]
        self.assertListEqual(
            [first_event_in_room2], [event["event_id"] for event in events]
        )

        next_batch = channel.json_body["next_batch"]

        # We also want to test that the vector clock style token works with
        # pagination. We do this by sending a couple of new events into the room
        # and syncing again to get a prev_batch token for each room, then
        # paginating from there back to the vector clock token.
        self.helper.send(room_id1, body="Hi again!", tok=self.other_access_token)
        self.helper.send(room_id2, body="Hi again!", tok=self.other_access_token)

        channel = make_request(
            self.reactor,
            sync_hs_site,
            "GET",
            f"/sync?since={next_batch}",
            access_token=access_token,
        )

        prev_batch1 = channel.json_body["rooms"]["join"][room_id1]["timeline"][
            "prev_batch"
        ]
        prev_batch2 = channel.json_body["rooms"]["join"][room_id2]["timeline"][
            "prev_batch"
        ]

        # Paginating back in the first room should not produce any results, as
        # no events have happened in it. This tests that we are correctly
        # filtering results based on the vector clock portion.
        channel = make_request(
            self.reactor,
            sync_hs_site,
            "GET",
            "/rooms/{}/messages?from={}&to={}&dir=b".format(
                room_id1, prev_batch1, vector_clock_token
            ),
            access_token=access_token,
        )
        self.assertListEqual([], channel.json_body["chunk"])

        # Paginating back on the second room should produce the first event
        # again. This tests that pagination isn't completely broken.
        channel = make_request(
            self.reactor,
            sync_hs_site,
            "GET",
            "/rooms/{}/messages?from={}&to={}&dir=b".format(
                room_id2, prev_batch2, vector_clock_token
            ),
            access_token=access_token,
        )
        self.assertEqual(len(channel.json_body["chunk"]), 1)
        self.assertEqual(
            channel.json_body["chunk"][0]["event_id"], first_event_in_room2
        )

        # Paginating forwards should give the same results
        channel = make_request(
            self.reactor,
            sync_hs_site,
            "GET",
            "/rooms/{}/messages?from={}&to={}&dir=f".format(
                room_id1, vector_clock_token, prev_batch1
            ),
            access_token=access_token,
        )
        self.assertListEqual([], channel.json_body["chunk"])

        channel = make_request(
            self.reactor,
            sync_hs_site,
            "GET",
            "/rooms/{}/messages?from={}&to={}&dir=f".format(
                room_id2,
                vector_clock_token,
                prev_batch2,
            ),
            access_token=access_token,
        )
        self.assertEqual(len(channel.json_body["chunk"]), 1)
        self.assertEqual(
            channel.json_body["chunk"][0]["event_id"], first_event_in_room2
        )