diff --git a/docs/replication.rst b/docs/replication.rst index 7e37e71987..310abb3488 100644 --- a/docs/replication.rst +++ b/docs/replication.rst @@ -26,28 +26,10 @@ expose the append-only log to the readers should be fairly minimal. Architecture ------------ -The Replication API -~~~~~~~~~~~~~~~~~~~ +The Replication Protocol +~~~~~~~~~~~~~~~~~~~~~~~~ -Synapse will optionally expose a long poll HTTP API for extracting updates. The -API will have a similar shape to /sync in that clients provide tokens -indicating where in the log they have reached and a timeout. The synapse server -then either responds with updates immediately if it already has updates or it -waits until the timeout for more updates. If the timeout expires and nothing -happened then the server returns an empty response. - -However unlike the /sync API this replication API is returning synapse specific -data rather than trying to implement a matrix specification. The replication -results are returned as arrays of rows where the rows are mostly lifted -directly from the database. This avoids unnecessary JSON parsing on the server -and hopefully avoids an impedance mismatch between the data returned and the -required updates to the datastore. - -This does not replicate all the database tables as many of the database tables -are indexes that can be recovered from the contents of other tables. - -The format and parameters for the api are documented in -``synapse/replication/resource.py``. +See ``tcp_replication.rst`` The Slaved DataStore diff --git a/docs/workers.rst b/docs/workers.rst index 65b6e690f7..2d3df91593 100644 --- a/docs/workers.rst +++ b/docs/workers.rst @@ -12,7 +12,7 @@ across multiple processes is a recipe for disaster, plus you should be using postgres anyway if you care about scalability). The workers communicate with the master synapse process via a synapse-specific -HTTP protocol called 'replication' - analogous to MySQL or Postgres style +TCP protocol called 'replication' - analogous to MySQL or Postgres style database replication; feeding a stream of relevant data to the workers so they can be kept in sync with the main synapse process and database state. @@ -21,16 +21,11 @@ To enable workers, you need to add a replication listener to the master synapse, listeners: - port: 9092 bind_address: '127.0.0.1' - type: http - tls: false - x_forwarded: false - resources: - - names: [replication] - compress: false + type: replication Under **no circumstances** should this replication API listener be exposed to the public internet; it currently implements no authentication whatsoever and is -unencrypted HTTP. +unencrypted. You then create a set of configs for the various worker processes. These should be worker configuration files should be stored in a dedicated subdirectory, to allow @@ -50,14 +45,16 @@ e.g. the HTTP listener that it provides (if any); logging configuration; etc. You should minimise the number of overrides though to maintain a usable config. You must specify the type of worker application (worker_app) and the replication -endpoint that it's talking to on the main synapse process (worker_replication_url). +endpoint that it's talking to on the main synapse process (worker_replication_host +and worker_replication_port). For instance:: worker_app: synapse.app.synchrotron # The replication listener on the synapse to talk to. - worker_replication_url: http://127.0.0.1:9092/_synapse/replication + worker_replication_host: 127.0.0.1 + worker_replication_port: 9092 worker_listeners: - type: http @@ -95,4 +92,3 @@ To manipulate a specific worker, you pass the -w option to synctl:: All of the above is highly experimental and subject to change as Synapse evolves, but documenting it here to help folks needing highly scalable Synapses similar to the one running matrix.org! - diff --git a/synapse/app/homeserver.py b/synapse/app/homeserver.py index 990eb477e5..6f5924d2c7 100755 --- a/synapse/app/homeserver.py +++ b/synapse/app/homeserver.py @@ -55,7 +55,6 @@ from synapse.crypto import context_factory from synapse.util.logcontext import LoggingContext, PreserveLoggingContext from synapse.metrics import register_memory_metrics, get_metrics_for from synapse.metrics.resource import MetricsResource, METRICS_PREFIX -from synapse.replication.resource import ReplicationResource, REPLICATION_PREFIX from synapse.replication.tcp.resource import ReplicationStreamProtocolFactory from synapse.federation.transport.server import TransportLayerServer @@ -167,9 +166,6 @@ class SynapseHomeServer(HomeServer): if name == "metrics" and self.get_config().enable_metrics: resources[METRICS_PREFIX] = MetricsResource(self) - if name == "replication": - resources[REPLICATION_PREFIX] = ReplicationResource(self) - if WEB_CLIENT_PREFIX in resources: root_resource = RootRedirect(WEB_CLIENT_PREFIX) else: diff --git a/synapse/replication/expire_cache.py b/synapse/replication/expire_cache.py deleted file mode 100644 index c05a50d7a6..0000000000 --- a/synapse/replication/expire_cache.py +++ /dev/null @@ -1,60 +0,0 @@ -# Copyright 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 synapse.http.server import respond_with_json_bytes, request_handler -from synapse.http.servlet import parse_json_object_from_request - -from twisted.web.resource import Resource -from twisted.web.server import NOT_DONE_YET - - -class ExpireCacheResource(Resource): - """ - HTTP endpoint for expiring storage caches. - - POST /_synapse/replication/expire_cache HTTP/1.1 - Content-Type: application/json - - { - "invalidate": [ - { - "name": "func_name", - "keys": ["key1", "key2"] - } - ] - } - """ - - def __init__(self, hs): - Resource.__init__(self) # Resource is old-style, so no super() - - self.store = hs.get_datastore() - self.version_string = hs.version_string - self.clock = hs.get_clock() - - def render_POST(self, request): - self._async_render_POST(request) - return NOT_DONE_YET - - @request_handler() - def _async_render_POST(self, request): - content = parse_json_object_from_request(request) - - for row in content["invalidate"]: - name = row["name"] - keys = tuple(row["keys"]) - - getattr(self.store, name).invalidate(keys) - - respond_with_json_bytes(request, 200, "{}") diff --git a/synapse/replication/presence_resource.py b/synapse/replication/presence_resource.py deleted file mode 100644 index fc18130ab4..0000000000 --- a/synapse/replication/presence_resource.py +++ /dev/null @@ -1,59 +0,0 @@ -# Copyright 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 synapse.http.server import respond_with_json_bytes, request_handler -from synapse.http.servlet import parse_json_object_from_request - -from twisted.web.resource import Resource -from twisted.web.server import NOT_DONE_YET -from twisted.internet import defer - - -class PresenceResource(Resource): - """ - HTTP endpoint for marking users as syncing. - - POST /_synapse/replication/presence HTTP/1.1 - Content-Type: application/json - - { - "process_id": "", - "syncing_users": [""] - } - """ - - def __init__(self, hs): - Resource.__init__(self) # Resource is old-style, so no super() - - self.version_string = hs.version_string - self.clock = hs.get_clock() - self.presence_handler = hs.get_presence_handler() - - def render_POST(self, request): - self._async_render_POST(request) - return NOT_DONE_YET - - @request_handler() - @defer.inlineCallbacks - def _async_render_POST(self, request): - content = parse_json_object_from_request(request) - - process_id = content["process_id"] - syncing_user_ids = content["syncing_users"] - - yield self.presence_handler.update_external_syncs( - process_id, set(syncing_user_ids) - ) - - respond_with_json_bytes(request, 200, "{}") diff --git a/synapse/replication/pusher_resource.py b/synapse/replication/pusher_resource.py deleted file mode 100644 index 9b01ab3c13..0000000000 --- a/synapse/replication/pusher_resource.py +++ /dev/null @@ -1,54 +0,0 @@ -# Copyright 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 synapse.http.server import respond_with_json_bytes, request_handler -from synapse.http.servlet import parse_json_object_from_request - -from twisted.web.resource import Resource -from twisted.web.server import NOT_DONE_YET -from twisted.internet import defer - - -class PusherResource(Resource): - """ - HTTP endpoint for deleting rejected pushers - """ - - def __init__(self, hs): - Resource.__init__(self) # Resource is old-style, so no super() - - self.version_string = hs.version_string - self.store = hs.get_datastore() - self.notifier = hs.get_notifier() - self.clock = hs.get_clock() - - def render_POST(self, request): - self._async_render_POST(request) - return NOT_DONE_YET - - @request_handler() - @defer.inlineCallbacks - def _async_render_POST(self, request): - content = parse_json_object_from_request(request) - - for remove in content["remove"]: - yield self.store.delete_pusher_by_app_id_pushkey_user_id( - remove["app_id"], - remove["push_key"], - remove["user_id"], - ) - - self.notifier.on_new_replication_data() - - respond_with_json_bytes(request, 200, "{}") diff --git a/synapse/replication/resource.py b/synapse/replication/resource.py deleted file mode 100644 index abd3fe7665..0000000000 --- a/synapse/replication/resource.py +++ /dev/null @@ -1,576 +0,0 @@ -# -*- 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. - -from synapse.http.servlet import parse_integer, parse_string -from synapse.http.server import request_handler, finish_request -from synapse.replication.pusher_resource import PusherResource -from synapse.replication.presence_resource import PresenceResource -from synapse.replication.expire_cache import ExpireCacheResource -from synapse.api.errors import SynapseError - -from twisted.web.resource import Resource -from twisted.web.server import NOT_DONE_YET -from twisted.internet import defer - -import ujson as json - -import collections -import logging - -logger = logging.getLogger(__name__) - -REPLICATION_PREFIX = "/_synapse/replication" - -STREAM_NAMES = ( - ("events",), - ("presence",), - ("typing",), - ("receipts",), - ("user_account_data", "room_account_data", "tag_account_data",), - ("backfill",), - ("push_rules",), - ("pushers",), - ("caches",), - ("to_device",), - ("public_rooms",), - ("federation",), - ("device_lists",), -) - - -class ReplicationResource(Resource): - """ - HTTP endpoint for extracting data from synapse. - - The streams of data returned by the endpoint are controlled by the - parameters given to the API. To return a given stream pass a query - parameter with a position in the stream to return data from or the - special value "-1" to return data from the start of the stream. - - If there is no data for any of the supplied streams after the given - position then the request will block until there is data for one - of the streams. This allows clients to long-poll this API. - - The possible streams are: - - * "streams": A special stream returing the positions of other streams. - * "events": The new events seen on the server. - * "presence": Presence updates. - * "typing": Typing updates. - * "receipts": Receipt updates. - * "user_account_data": Top-level per user account data. - * "room_account_data: Per room per user account data. - * "tag_account_data": Per room per user tags. - * "backfill": Old events that have been backfilled from other servers. - * "push_rules": Per user changes to push rules. - * "pushers": Per user changes to their pushers. - * "caches": Cache invalidations. - - The API takes two additional query parameters: - - * "timeout": How long to wait before returning an empty response. - * "limit": The maximum number of rows to return for the selected streams. - - The response is a JSON object with keys for each stream with updates. Under - each key is a JSON object with: - - * "position": The current position of the stream. - * "field_names": The names of the fields in each row. - * "rows": The updates as an array of arrays. - - There are a number of ways this API could be used: - - 1) To replicate the contents of the backing database to another database. - 2) To be notified when the contents of a shared backing database changes. - 3) To "tail" the activity happening on a server for debugging. - - In the first case the client would track all of the streams and store it's - own copy of the data. - - In the second case the client might theoretically just be able to follow - the "streams" stream to track where the other streams are. However in - practise it will probably need to get the contents of the streams in - order to expire the any in-memory caches. Whether it gets the contents - of the streams from this replication API or directly from the backing - store is a matter of taste. - - In the third case the client would use the "streams" stream to find what - streams are available and their current positions. Then it can start - long-polling this replication API for new data on those streams. - """ - - def __init__(self, hs): - Resource.__init__(self) # Resource is old-style, so no super() - - self.version_string = hs.version_string - self.store = hs.get_datastore() - self.sources = hs.get_event_sources() - self.presence_handler = hs.get_presence_handler() - self.typing_handler = hs.get_typing_handler() - self.federation_sender = hs.get_federation_sender() - self.notifier = hs.notifier - self.clock = hs.get_clock() - self.config = hs.get_config() - - self.putChild("remove_pushers", PusherResource(hs)) - self.putChild("syncing_users", PresenceResource(hs)) - self.putChild("expire_cache", ExpireCacheResource(hs)) - - def render_GET(self, request): - self._async_render_GET(request) - return NOT_DONE_YET - - @defer.inlineCallbacks - def current_replication_token(self): - stream_token = yield self.sources.get_current_token() - backfill_token = yield self.store.get_current_backfill_token() - push_rules_token, room_stream_token = self.store.get_push_rules_stream_token() - pushers_token = self.store.get_pushers_stream_token() - caches_token = self.store.get_cache_stream_token() - public_rooms_token = self.store.get_current_public_room_stream_id() - federation_token = self.federation_sender.get_current_token() - device_list_token = self.store.get_device_stream_token() - - defer.returnValue(_ReplicationToken( - room_stream_token, - int(stream_token.presence_key), - int(stream_token.typing_key), - int(stream_token.receipt_key), - int(stream_token.account_data_key), - backfill_token, - push_rules_token, - pushers_token, - 0, # State stream is no longer a thing - caches_token, - int(stream_token.to_device_key), - int(public_rooms_token), - int(federation_token), - int(device_list_token), - )) - - @request_handler() - @defer.inlineCallbacks - def _async_render_GET(self, request): - limit = parse_integer(request, "limit", 100) - timeout = parse_integer(request, "timeout", 10 * 1000) - - request.setHeader(b"Content-Type", b"application/json") - - request_streams = { - name: parse_integer(request, name) - for names in STREAM_NAMES for name in names - } - request_streams["streams"] = parse_string(request, "streams") - - federation_ack = parse_integer(request, "federation_ack", None) - - def replicate(): - return self.replicate( - request_streams, limit, - federation_ack=federation_ack - ) - - writer = yield self.notifier.wait_for_replication(replicate, timeout) - result = writer.finish() - - for stream_name, stream_content in result.items(): - logger.info( - "Replicating %d rows of %s from %s -> %s", - len(stream_content["rows"]), - stream_name, - request_streams.get(stream_name), - stream_content["position"], - ) - - request.write(json.dumps(result, ensure_ascii=False)) - finish_request(request) - - @defer.inlineCallbacks - def replicate(self, request_streams, limit, federation_ack=None): - writer = _Writer() - current_token = yield self.current_replication_token() - logger.debug("Replicating up to %r", current_token) - - if limit == 0: - raise SynapseError(400, "Limit cannot be 0") - - yield self.account_data(writer, current_token, limit, request_streams) - yield self.events(writer, current_token, limit, request_streams) - # TODO: implement limit - yield self.presence(writer, current_token, request_streams) - yield self.typing(writer, current_token, request_streams) - yield self.receipts(writer, current_token, limit, request_streams) - yield self.push_rules(writer, current_token, limit, request_streams) - yield self.pushers(writer, current_token, limit, request_streams) - yield self.caches(writer, current_token, limit, request_streams) - yield self.to_device(writer, current_token, limit, request_streams) - yield self.public_rooms(writer, current_token, limit, request_streams) - yield self.device_lists(writer, current_token, limit, request_streams) - self.federation(writer, current_token, limit, request_streams, federation_ack) - self.streams(writer, current_token, request_streams) - - logger.debug("Replicated %d rows", writer.total) - defer.returnValue(writer) - - def streams(self, writer, current_token, request_streams): - request_token = request_streams.get("streams") - - streams = [] - - if request_token is not None: - if request_token == "-1": - for names, position in zip(STREAM_NAMES, current_token): - streams.extend((name, position) for name in names) - else: - items = zip( - STREAM_NAMES, - current_token, - _ReplicationToken(request_token) - ) - for names, current_id, last_id in items: - if last_id < current_id: - streams.extend((name, current_id) for name in names) - - if streams: - writer.write_header_and_rows( - "streams", streams, ("name", "position"), - position=str(current_token) - ) - - @defer.inlineCallbacks - def events(self, writer, current_token, limit, request_streams): - request_events = request_streams.get("events") - request_backfill = request_streams.get("backfill") - - if request_events is not None or request_backfill is not None: - if request_events is None: - request_events = current_token.events - if request_backfill is None: - request_backfill = current_token.backfill - - no_new_tokens = ( - request_events == current_token.events - and request_backfill == current_token.backfill - ) - if no_new_tokens: - return - - res = yield self.store.get_all_new_events( - request_backfill, request_events, - current_token.backfill, current_token.events, - limit - ) - - upto_events_token = _position_from_rows( - res.new_forward_events, current_token.events - ) - - upto_backfill_token = _position_from_rows( - res.new_backfill_events, current_token.backfill - ) - - if request_events != upto_events_token: - writer.write_header_and_rows("events", res.new_forward_events, ( - "position", "event_id", "room_id", "type", "state_key", - ), position=upto_events_token) - - if request_backfill != upto_backfill_token: - writer.write_header_and_rows("backfill", res.new_backfill_events, ( - "position", "event_id", "room_id", "type", "state_key", "redacts", - ), position=upto_backfill_token) - - writer.write_header_and_rows( - "forward_ex_outliers", res.forward_ex_outliers, - ("position", "event_id", "state_group"), - ) - writer.write_header_and_rows( - "backward_ex_outliers", res.backward_ex_outliers, - ("position", "event_id", "state_group"), - ) - - @defer.inlineCallbacks - def presence(self, writer, current_token, request_streams): - current_position = current_token.presence - - request_presence = request_streams.get("presence") - - if request_presence is not None and request_presence != current_position: - presence_rows = yield self.presence_handler.get_all_presence_updates( - request_presence, current_position - ) - upto_token = _position_from_rows(presence_rows, current_position) - writer.write_header_and_rows("presence", presence_rows, ( - "position", "user_id", "state", "last_active_ts", - "last_federation_update_ts", "last_user_sync_ts", - "status_msg", "currently_active", - ), position=upto_token) - - @defer.inlineCallbacks - def typing(self, writer, current_token, request_streams): - current_position = current_token.typing - - request_typing = request_streams.get("typing") - - if request_typing is not None and request_typing != current_position: - # If they have a higher token than current max, we can assume that - # they had been talking to a previous instance of the master. Since - # we reset the token on restart, the best (but hacky) thing we can - # do is to simply resend down all the typing notifications. - if request_typing > current_position: - request_typing = 0 - - typing_rows = yield self.typing_handler.get_all_typing_updates( - request_typing, current_position - ) - upto_token = _position_from_rows(typing_rows, current_position) - writer.write_header_and_rows("typing", typing_rows, ( - "position", "room_id", "typing" - ), position=upto_token) - - @defer.inlineCallbacks - def receipts(self, writer, current_token, limit, request_streams): - current_position = current_token.receipts - - request_receipts = request_streams.get("receipts") - - if request_receipts is not None and request_receipts != current_position: - receipts_rows = yield self.store.get_all_updated_receipts( - request_receipts, current_position, limit - ) - upto_token = _position_from_rows(receipts_rows, current_position) - writer.write_header_and_rows("receipts", receipts_rows, ( - "position", "room_id", "receipt_type", "user_id", "event_id", "data" - ), position=upto_token) - - @defer.inlineCallbacks - def account_data(self, writer, current_token, limit, request_streams): - current_position = current_token.account_data - - user_account_data = request_streams.get("user_account_data") - room_account_data = request_streams.get("room_account_data") - tag_account_data = request_streams.get("tag_account_data") - - if user_account_data is not None or room_account_data is not None: - if user_account_data is None: - user_account_data = current_position - if room_account_data is None: - room_account_data = current_position - - no_new_tokens = ( - user_account_data == current_position - and room_account_data == current_position - ) - if no_new_tokens: - return - - user_rows, room_rows = yield self.store.get_all_updated_account_data( - user_account_data, room_account_data, current_position, limit - ) - - upto_users_token = _position_from_rows(user_rows, current_position) - upto_rooms_token = _position_from_rows(room_rows, current_position) - - writer.write_header_and_rows("user_account_data", user_rows, ( - "position", "user_id", "type", "content" - ), position=upto_users_token) - writer.write_header_and_rows("room_account_data", room_rows, ( - "position", "user_id", "room_id", "type", "content" - ), position=upto_rooms_token) - - if tag_account_data is not None: - tag_rows = yield self.store.get_all_updated_tags( - tag_account_data, current_position, limit - ) - upto_tag_token = _position_from_rows(tag_rows, current_position) - writer.write_header_and_rows("tag_account_data", tag_rows, ( - "position", "user_id", "room_id", "tags" - ), position=upto_tag_token) - - @defer.inlineCallbacks - def push_rules(self, writer, current_token, limit, request_streams): - current_position = current_token.push_rules - - push_rules = request_streams.get("push_rules") - - if push_rules is not None and push_rules != current_position: - rows = yield self.store.get_all_push_rule_updates( - push_rules, current_position, limit - ) - upto_token = _position_from_rows(rows, current_position) - writer.write_header_and_rows("push_rules", rows, ( - "position", "event_stream_ordering", "user_id", "rule_id", "op", - "priority_class", "priority", "conditions", "actions" - ), position=upto_token) - - @defer.inlineCallbacks - def pushers(self, writer, current_token, limit, request_streams): - current_position = current_token.pushers - - pushers = request_streams.get("pushers") - - if pushers is not None and pushers != current_position: - updated, deleted = yield self.store.get_all_updated_pushers( - pushers, current_position, limit - ) - upto_token = _position_from_rows(updated, current_position) - writer.write_header_and_rows("pushers", updated, ( - "position", "user_id", "access_token", "profile_tag", "kind", - "app_id", "app_display_name", "device_display_name", "pushkey", - "ts", "lang", "data" - ), position=upto_token) - writer.write_header_and_rows("deleted_pushers", deleted, ( - "position", "user_id", "app_id", "pushkey" - ), position=upto_token) - - @defer.inlineCallbacks - def caches(self, writer, current_token, limit, request_streams): - current_position = current_token.caches - - caches = request_streams.get("caches") - - if caches is not None and caches != current_position: - updated_caches = yield self.store.get_all_updated_caches( - caches, current_position, limit - ) - upto_token = _position_from_rows(updated_caches, current_position) - writer.write_header_and_rows("caches", updated_caches, ( - "position", "cache_func", "keys", "invalidation_ts" - ), position=upto_token) - - @defer.inlineCallbacks - def to_device(self, writer, current_token, limit, request_streams): - current_position = current_token.to_device - - to_device = request_streams.get("to_device") - - if to_device is not None and to_device != current_position: - to_device_rows = yield self.store.get_all_new_device_messages( - to_device, current_position, limit - ) - upto_token = _position_from_rows(to_device_rows, current_position) - writer.write_header_and_rows("to_device", to_device_rows, ( - "position", "user_id", "device_id", "message_json" - ), position=upto_token) - - @defer.inlineCallbacks - def public_rooms(self, writer, current_token, limit, request_streams): - current_position = current_token.public_rooms - - public_rooms = request_streams.get("public_rooms") - - if public_rooms is not None and public_rooms != current_position: - public_rooms_rows = yield self.store.get_all_new_public_rooms( - public_rooms, current_position, limit - ) - upto_token = _position_from_rows(public_rooms_rows, current_position) - writer.write_header_and_rows("public_rooms", public_rooms_rows, ( - "position", "room_id", "visibility", "appservice_id", "network_id", - ), position=upto_token) - - def federation(self, writer, current_token, limit, request_streams, federation_ack): - if self.config.send_federation: - return - - current_position = current_token.federation - - federation = request_streams.get("federation") - - if federation is not None and federation != current_position: - federation_rows = self.federation_sender.get_replication_rows( - federation, current_position, limit, federation_ack=federation_ack, - ) - upto_token = _position_from_rows(federation_rows, current_position) - writer.write_header_and_rows("federation", federation_rows, ( - "position", "type", "content", - ), position=upto_token) - - @defer.inlineCallbacks - def device_lists(self, writer, current_token, limit, request_streams): - current_position = current_token.device_lists - - device_lists = request_streams.get("device_lists") - - if device_lists is not None and device_lists != current_position: - changes = yield self.store.get_all_device_list_changes_for_remotes( - device_lists, current_position, - ) - writer.write_header_and_rows("device_lists", changes, ( - "position", "user_id", "destination", - ), position=current_position) - - -class _Writer(object): - """Writes the streams as a JSON object as the response to the request""" - def __init__(self): - self.streams = {} - self.total = 0 - - def write_header_and_rows(self, name, rows, fields, position=None): - if position is None: - if rows: - position = rows[-1][0] - else: - return - - self.streams[name] = { - "position": position if type(position) is int else str(position), - "field_names": fields, - "rows": rows, - } - - self.total += len(rows) - - def __nonzero__(self): - return bool(self.total) - - def finish(self): - return self.streams - - -class _ReplicationToken(collections.namedtuple("_ReplicationToken", ( - "events", "presence", "typing", "receipts", "account_data", "backfill", - "push_rules", "pushers", "state", "caches", "to_device", "public_rooms", - "federation", "device_lists", -))): - __slots__ = [] - - def __new__(cls, *args): - if len(args) == 1: - streams = [int(value) for value in args[0].split("_")] - if len(streams) < len(cls._fields): - streams.extend([0] * (len(cls._fields) - len(streams))) - return cls(*streams) - else: - return super(_ReplicationToken, cls).__new__(cls, *args) - - def __str__(self): - return "_".join(str(value) for value in self) - - -def _position_from_rows(rows, current_position): - """Calculates a position to return for a stream. Ideally we want to return the - position of the last row, as that will be the most correct. However, if there - are no rows we fall back to using the current position to stop us from - repeatedly hitting the storage layer unncessarily thinking there are updates. - (Not all advances of the token correspond to an actual update) - - We can't just always return the current position, as we often limit the - number of rows we replicate, and so the stream may lag. The assumption is - that if the storage layer returns no new rows then we are not lagging and - we are at the `current_position`. - """ - if rows: - return rows[-1][0] - return current_position diff --git a/tests/replication/test_resource.py b/tests/replication/test_resource.py deleted file mode 100644 index 429b37e360..0000000000 --- a/tests/replication/test_resource.py +++ /dev/null @@ -1,204 +0,0 @@ -# -*- coding: utf-8 -*- -# Copyright 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. - -import contextlib -import json - -from mock import Mock, NonCallableMock -from twisted.internet import defer - -import synapse.types -from synapse.replication.resource import ReplicationResource -from synapse.types import UserID -from tests import unittest -from tests.utils import setup_test_homeserver - - -class ReplicationResourceCase(unittest.TestCase): - @defer.inlineCallbacks - def setUp(self): - self.hs = yield setup_test_homeserver( - "red", - http_client=None, - replication_layer=Mock(), - ratelimiter=NonCallableMock(spec_set=[ - "send_message", - ]), - ) - self.user_id = "@seeing:red" - self.user = UserID.from_string(self.user_id) - - self.hs.get_ratelimiter().send_message.return_value = (True, 0) - - self.resource = ReplicationResource(self.hs) - - @defer.inlineCallbacks - def test_streams(self): - # Passing "-1" returns the current stream positions - code, body = yield self.get(streams="-1") - self.assertEquals(code, 200) - self.assertEquals(body["streams"]["field_names"], ["name", "position"]) - position = body["streams"]["position"] - # Passing the current position returns an empty response after the - # timeout - get = self.get(streams=str(position), timeout="0") - self.hs.clock.advance_time_msec(1) - code, body = yield get - self.assertEquals(code, 200) - self.assertEquals(body, {}) - - @defer.inlineCallbacks - def test_events(self): - get = self.get(events="-1", timeout="0") - yield self.hs.get_handlers().room_creation_handler.create_room( - synapse.types.create_requester(self.user), {} - ) - code, body = yield get - self.assertEquals(code, 200) - self.assertEquals(body["events"]["field_names"], [ - "position", "event_id", "room_id", "type", "state_key", - ]) - - @defer.inlineCallbacks - def test_presence(self): - get = self.get(presence="-1") - yield self.hs.get_presence_handler().set_state( - self.user, {"presence": "online"} - ) - code, body = yield get - self.assertEquals(code, 200) - self.assertEquals(body["presence"]["field_names"], [ - "position", "user_id", "state", "last_active_ts", - "last_federation_update_ts", "last_user_sync_ts", - "status_msg", "currently_active", - ]) - - @defer.inlineCallbacks - def test_typing(self): - room_id = yield self.create_room() - get = self.get(typing="-1") - yield self.hs.get_typing_handler().started_typing( - self.user, self.user, room_id, timeout=2 - ) - code, body = yield get - self.assertEquals(code, 200) - self.assertEquals(body["typing"]["field_names"], [ - "position", "room_id", "typing" - ]) - - @defer.inlineCallbacks - def test_receipts(self): - room_id = yield self.create_room() - event_id = yield self.send_text_message(room_id, "Hello, World") - get = self.get(receipts="-1") - yield self.hs.get_receipts_handler().received_client_receipt( - room_id, "m.read", self.user_id, event_id - ) - code, body = yield get - self.assertEquals(code, 200) - self.assertEquals(body["receipts"]["field_names"], [ - "position", "room_id", "receipt_type", "user_id", "event_id", "data" - ]) - - def _test_timeout(stream): - """Check that a request for the given stream timesout""" - @defer.inlineCallbacks - def test_timeout(self): - get = self.get(**{stream: "-1", "timeout": "0"}) - self.hs.clock.advance_time_msec(1) - code, body = yield get - self.assertEquals(code, 200) - self.assertEquals(body.get("rows", []), []) - test_timeout.__name__ = "test_timeout_%s" % (stream) - return test_timeout - - test_timeout_events = _test_timeout("events") - test_timeout_presence = _test_timeout("presence") - test_timeout_typing = _test_timeout("typing") - test_timeout_receipts = _test_timeout("receipts") - test_timeout_user_account_data = _test_timeout("user_account_data") - test_timeout_room_account_data = _test_timeout("room_account_data") - test_timeout_tag_account_data = _test_timeout("tag_account_data") - test_timeout_backfill = _test_timeout("backfill") - test_timeout_push_rules = _test_timeout("push_rules") - test_timeout_pushers = _test_timeout("pushers") - test_timeout_state = _test_timeout("state") - - @defer.inlineCallbacks - def send_text_message(self, room_id, message): - handler = self.hs.get_handlers().message_handler - event = yield handler.create_and_send_nonmember_event( - synapse.types.create_requester(self.user), - { - "type": "m.room.message", - "content": {"body": "message", "msgtype": "m.text"}, - "room_id": room_id, - "sender": self.user.to_string(), - } - ) - defer.returnValue(event.event_id) - - @defer.inlineCallbacks - def create_room(self): - result = yield self.hs.get_handlers().room_creation_handler.create_room( - synapse.types.create_requester(self.user), {} - ) - defer.returnValue(result["room_id"]) - - @defer.inlineCallbacks - def get(self, **params): - request = NonCallableMock(spec_set=[ - "write", "finish", "setResponseCode", "setHeader", "args", - "method", "processing" - ]) - - request.method = "GET" - request.args = {k: [v] for k, v in params.items()} - - @contextlib.contextmanager - def processing(): - yield - request.processing = processing - - yield self.resource._async_render_GET(request) - self.assertTrue(request.finish.called) - - if request.setResponseCode.called: - response_code = request.setResponseCode.call_args[0][0] - else: - response_code = 200 - - response_json = "".join( - call[0][0] for call in request.write.call_args_list - ) - response_body = json.loads(response_json) - - if response_code == 200: - self.check_response(response_body) - - defer.returnValue((response_code, response_body)) - - def check_response(self, response_body): - for name, stream in response_body.items(): - self.assertIn("field_names", stream) - field_names = stream["field_names"] - self.assertIn("rows", stream) - for row in stream["rows"]: - self.assertEquals( - len(row), len(field_names), - "%s: len(row = %r) == len(field_names = %r)" % ( - name, row, field_names - ) - )