summary refs log tree commit diff
path: root/synapse/appservice/scheduler.py
blob: 54c42d1b94e4e648bf9a31a7655cd0e90b2ca495 (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
# -*- coding: utf-8 -*-
# Copyright 2015 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.
"""
This module controls the reliability for application service transactions.

The nominal flow through this module looks like:
              __________
1---ASa[e]-->|  Service |--> Queue ASa[f]
2----ASb[e]->|  Queuer  |
3--ASa[f]--->|__________|-----------+ ASa[e], ASb[e]
                                    V
      -````````-            +------------+
      |````````|<--StoreTxn-|Transaction |
      |Database|            | Controller |---> SEND TO AS
      `--------`            +------------+
What happens on SEND TO AS depends on the state of the Application Service:
 - If the AS is marked as DOWN, do nothing.
 - If the AS is marked as UP, send the transaction.
     * SUCCESS : Increment where the AS is up to txn-wise and nuke the txn
                 contents from the db.
     * FAILURE : Marked AS as DOWN and start Recoverer.

Recoverer attempts to recover ASes who have died. The flow for this looks like:
                ,--------------------- backoff++ --------------.
               V                                               |
  START ---> Wait exp ------> Get oldest txn ID from ----> FAILURE
             backoff           DB and try to send it
                                 ^                |___________
Mark AS as                       |                            V
UP & quit           +---------- YES                       SUCCESS
    |               |                                         |
    NO <--- Have more txns? <------ Mark txn success & nuke <-+
                                      from db; incr AS pos.
                                         Reset backoff.

This is all tied together by the AppServiceScheduler which DIs the required
components.
"""

from synapse.appservice import ApplicationServiceState
from twisted.internet import defer
import logging

logger = logging.getLogger(__name__)


class AppServiceScheduler(object):
    """ Public facing API for this module. Does the required DI to tie the
    components together. This also serves as the "event_pool", which in this
    case is a simple array.
    """

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

        def create_recoverer(service, callback):
            return _Recoverer(clock, store, as_api, service, callback)

        self.txn_ctrl = _TransactionController(
            clock, store, as_api, create_recoverer
        )
        self.queuer = _ServiceQueuer(self.txn_ctrl)

    @defer.inlineCallbacks
    def start(self):
        logger.info("Starting appservice scheduler")
        # check for any DOWN ASes and start recoverers for them.
        recoverers = yield _Recoverer.start(
            self.clock, self.store, self.as_api, self.txn_ctrl.on_recovered
        )
        self.txn_ctrl.add_recoverers(recoverers)
        self.txn_ctrl.start_polling()

    def submit_event_for_as(self, service, event):
        self.queuer.enqueue(service, event)


class _ServiceQueuer(object):
    """Queues events for the same application service together, sending
    transactions as soon as possible. Once a transaction is sent successfully,
    this schedules any other events in the queue to run.
    """

    def __init__(self, txn_ctrl):
        self.groups = {}  # dict of {service: [events]}
        self.txn_ctrl = txn_ctrl

    def enqueue(self, service, event):
        # if nothing in queue for this service, send event immediately and add
        # callbacks.
        self.txn_ctrl.send(service, [event])

        # else add to queue for this service

        if service not in self.groups:
            self.groups[service] = []
        self.groups[service].append(event)

    def drain_groups(self):
        groups = self.groups
        self.groups = {}
        return groups


class _TransactionController(object):

    def __init__(self, clock, store, as_api, recoverer_fn):
        self.clock = clock
        self.store = store
        self.as_api = as_api
        self.recoverer_fn = recoverer_fn
        # keep track of how many recoverers there are
        self.recoverers = []

    @defer.inlineCallbacks
    def send(self, service, events):
        try:
            txn = yield self.store.create_appservice_txn(
                service=service,
                events=events
            )
            service_is_up = yield self._is_service_up(service)
            if service_is_up:
                sent = yield txn.send(self.as_api)
                if sent:
                    txn.complete(self.store)
                else:
                    self._start_recoverer(service)
        except Exception as e:
            logger.exception(e)

    @defer.inlineCallbacks
    def on_recovered(self, recoverer):
        self.recoverers.remove(recoverer)
        logger.info("Successfully recovered application service AS ID %s",
                    recoverer.service.id)
        logger.info("Remaining active recoverers: %s", len(self.recoverers))
        yield self.store.set_appservice_state(
            recoverer.service,
            ApplicationServiceState.UP
        )

    def add_recoverers(self, recoverers):
        for r in recoverers:
            self.recoverers.append(r)
        if len(recoverers) > 0:
            logger.info("New active recoverers: %s", len(self.recoverers))

    @defer.inlineCallbacks
    def _start_recoverer(self, service):
        yield self.store.set_appservice_state(
            service,
            ApplicationServiceState.DOWN
        )
        logger.info(
            "Application service falling behind. Starting recoverer. AS ID %s",
            service.id
        )
        recoverer = self.recoverer_fn(service, self.on_recovered)
        self.add_recoverers([recoverer])
        recoverer.recover()

    @defer.inlineCallbacks
    def _is_service_up(self, service):
        state = yield self.store.get_appservice_state(service)
        defer.returnValue(state == ApplicationServiceState.UP or state is None)


class _Recoverer(object):

    @staticmethod
    @defer.inlineCallbacks
    def start(clock, store, as_api, callback):
        services = yield store.get_appservices_by_state(
            ApplicationServiceState.DOWN
        )
        recoverers = [
            _Recoverer(clock, store, as_api, s, callback) for s in services
        ]
        for r in recoverers:
            logger.info("Starting recoverer for AS ID %s which was marked as "
                        "DOWN", r.service.id)
            r.recover()
        defer.returnValue(recoverers)

    def __init__(self, clock, store, as_api, service, callback):
        self.clock = clock
        self.store = store
        self.as_api = as_api
        self.service = service
        self.callback = callback
        self.backoff_counter = 1

    def recover(self):
        self.clock.call_later((2 ** self.backoff_counter), self.retry)

    def _backoff(self):
        # cap the backoff to be around 18h => (2^16) = 65536 secs
        if self.backoff_counter < 16:
            self.backoff_counter += 1
        self.recover()

    @defer.inlineCallbacks
    def retry(self):
        try:
            txn = yield self.store.get_oldest_unsent_txn(self.service)
            if txn:
                logger.info("Retrying transaction %s for AS ID %s",
                            txn.id, txn.service.id)
                sent = yield txn.send(self.as_api)
                if sent:
                    yield txn.complete(self.store)
                    # reset the backoff counter and retry immediately
                    self.backoff_counter = 1
                    yield self.retry()
                else:
                    self._backoff()
            else:
                self._set_service_recovered()
        except Exception as e:
            logger.exception(e)
            self._backoff()

    def _set_service_recovered(self):
        self.callback(self)