summary refs log tree commit diff
path: root/synapse/util/ratelimitutils.py
blob: ea321bc6a990b3bebe62e55c8272d49bff56d265 (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
# -*- coding: utf-8 -*-
# Copyright 2015, 2016 OpenMarket 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.

from twisted.internet import defer

from synapse.api.errors import LimitExceededError

from synapse.util.async import sleep

import collections
import contextlib
import logging


logger = logging.getLogger(__name__)


class FederationRateLimiter(object):
    def __init__(self, clock, window_size, sleep_limit, sleep_msec,
                 reject_limit, concurrent_requests):
        """
        Args:
            clock (Clock)
            window_size (int): The window size in milliseconds.
            sleep_limit (int): The number of requests received in the last
                `window_size` milliseconds before we artificially start
                delaying processing of requests.
            sleep_msec (int): The number of milliseconds to delay processing
                of incoming requests by.
            reject_limit (int): The maximum number of requests that are can be
                queued for processing before we start rejecting requests with
                a 429 Too Many Requests response.
            concurrent_requests (int): The number of concurrent requests to
                process.
        """
        self.clock = clock

        self.window_size = window_size
        self.sleep_limit = sleep_limit
        self.sleep_msec = sleep_msec
        self.reject_limit = reject_limit
        self.concurrent_requests = concurrent_requests

        self.ratelimiters = {}

    def ratelimit(self, host):
        """Used to ratelimit an incoming request from given host

        Example usage:

            with rate_limiter.ratelimit(origin) as wait_deferred:
                yield wait_deferred
                # Handle request ...

        Args:
            host (str): Origin of incoming request.

        Returns:
            _PerHostRatelimiter
        """
        return self.ratelimiters.setdefault(
            host,
            _PerHostRatelimiter(
                clock=self.clock,
                window_size=self.window_size,
                sleep_limit=self.sleep_limit,
                sleep_msec=self.sleep_msec,
                reject_limit=self.reject_limit,
                concurrent_requests=self.concurrent_requests,
            )
        ).ratelimit()


class _PerHostRatelimiter(object):
    def __init__(self, clock, window_size, sleep_limit, sleep_msec,
                 reject_limit, concurrent_requests):
        self.clock = clock

        self.window_size = window_size
        self.sleep_limit = sleep_limit
        self.sleep_msec = sleep_msec
        self.reject_limit = reject_limit
        self.concurrent_requests = concurrent_requests

        self.sleeping_requests = set()
        self.ready_request_queue = collections.OrderedDict()
        self.current_processing = set()
        self.request_times = []

    def is_empty(self):
        time_now = self.clock.time_msec()
        self.request_times[:] = [
            r for r in self.request_times
            if time_now - r < self.window_size
        ]

        return not (
            self.ready_request_queue
            or self.sleeping_requests
            or self.current_processing
            or self.request_times
        )

    @contextlib.contextmanager
    def ratelimit(self):
        # `contextlib.contextmanager` takes a generator and turns it into a
        # context manager. The generator should only yield once with a value
        # to be returned by manager.
        # Exceptions will be reraised at the yield.

        request_id = object()
        ret = self._on_enter(request_id)
        try:
            yield ret
        finally:
            self._on_exit(request_id)

    def _on_enter(self, request_id):
        time_now = self.clock.time_msec()
        self.request_times[:] = [
            r for r in self.request_times
            if time_now - r < self.window_size
        ]

        queue_size = len(self.ready_request_queue) + len(self.sleeping_requests)
        if queue_size > self.reject_limit:
            raise LimitExceededError(
                retry_after_ms=int(
                    self.window_size / self.sleep_limit
                ),
            )

        self.request_times.append(time_now)

        def queue_request():
            if len(self.current_processing) > self.concurrent_requests:
                logger.debug("Ratelimit [%s]: Queue req", id(request_id))
                queue_defer = defer.Deferred()
                self.ready_request_queue[request_id] = queue_defer
                return queue_defer
            else:
                return defer.succeed(None)

        logger.debug(
            "Ratelimit [%s]: len(self.request_times)=%d",
            id(request_id), len(self.request_times),
        )

        if len(self.request_times) > self.sleep_limit:
            logger.debug(
                "Ratelimit [%s]: sleeping req",
                id(request_id),
            )
            ret_defer = sleep(self.sleep_msec / 1000.0)

            self.sleeping_requests.add(request_id)

            def on_wait_finished(_):
                logger.debug(
                    "Ratelimit [%s]: Finished sleeping",
                    id(request_id),
                )
                self.sleeping_requests.discard(request_id)
                queue_defer = queue_request()
                return queue_defer

            ret_defer.addBoth(on_wait_finished)
        else:
            ret_defer = queue_request()

        def on_start(r):
            logger.debug(
                "Ratelimit [%s]: Processing req",
                id(request_id),
            )
            self.current_processing.add(request_id)
            return r

        def on_err(r):
            self.current_processing.discard(request_id)
            return r

        def on_both(r):
            # Ensure that we've properly cleaned up.
            self.sleeping_requests.discard(request_id)
            self.ready_request_queue.pop(request_id, None)
            return r

        ret_defer.addCallbacks(on_start, on_err)
        ret_defer.addBoth(on_both)
        return ret_defer

    def _on_exit(self, request_id):
        logger.debug(
            "Ratelimit [%s]: Processed req",
            id(request_id),
        )
        self.current_processing.discard(request_id)
        try:
            request_id, deferred = self.ready_request_queue.popitem()
            self.current_processing.add(request_id)
            deferred.callback(None)
        except KeyError:
            pass