synapse-product/synapse/storage/databases/main/pusher.py

351 lines
12 KiB
Python
Raw Normal View History

# -*- coding: utf-8 -*-
2016-01-06 23:26:29 -05:00
# Copyright 2014-2016 OpenMarket Ltd
2018-02-23 05:41:34 -05:00
# Copyright 2018 New Vector 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.
2018-07-09 02:09:20 -04:00
import logging
from typing import Iterable, Iterator, List, Tuple
from canonicaljson import encode_canonical_json
from synapse.storage._base import SQLBaseStore, db_to_json
from synapse.util.caches.descriptors import cached, cachedList
logger = logging.getLogger(__name__)
2018-02-21 05:54:18 -05:00
class PusherWorkerStore(SQLBaseStore):
def _decode_pushers_rows(self, rows: Iterable[dict]) -> Iterator[dict]:
"""JSON-decode the data in the rows returned from the `pushers` table
Drops any rows whose data cannot be decoded
"""
for r in rows:
data_json = r["data"]
try:
r["data"] = db_to_json(data_json)
except Exception as e:
logger.warning(
"Invalid JSON in data for pusher %d: %s, %s",
2019-06-20 05:32:02 -04:00
r["id"],
data_json,
e.args[0],
)
continue
yield r
async def user_has_pusher(self, user_id):
ret = await self.db_pool.simple_select_one_onecol(
"pushers", {"user_name": user_id}, "id", allow_none=True
)
return ret is not None
def get_pushers_by_app_id_and_pushkey(self, app_id, pushkey):
return self.get_pushers_by({"app_id": app_id, "pushkey": pushkey})
2016-04-12 08:35:08 -04:00
def get_pushers_by_user_id(self, user_id):
return self.get_pushers_by({"user_name": user_id})
async def get_pushers_by(self, keyvalues):
ret = await self.db_pool.simple_select_list(
"pushers",
keyvalues,
[
"id",
"user_name",
"access_token",
"profile_tag",
"kind",
"app_id",
"app_display_name",
"device_display_name",
"pushkey",
"ts",
"lang",
"data",
"last_stream_ordering",
"last_success",
"failing_since",
],
desc="get_pushers_by",
)
return self._decode_pushers_rows(ret)
async def get_all_pushers(self):
2015-04-29 13:03:42 -04:00
def get_pushers(txn):
txn.execute("SELECT * FROM pushers")
rows = self.db_pool.cursor_to_dict(txn)
return self._decode_pushers_rows(rows)
2015-04-29 13:07:36 -04:00
return await self.db_pool.runInteraction("get_all_pushers", get_pushers)
async def get_all_updated_pushers_rows(
self, instance_name: str, last_id: int, current_id: int, limit: int
) -> Tuple[List[Tuple[int, tuple]], int, bool]:
"""Get updates for pushers replication stream.
2016-03-15 13:01:43 -04:00
Args:
instance_name: The writer we want to fetch updates from. Unused
here since there is only ever one writer.
last_id: The token to fetch updates from. Exclusive.
current_id: The token to fetch updates up to. Inclusive.
limit: The requested limit for the number of rows to return. The
function may return more or fewer rows.
Returns:
A tuple consisting of: the updates, a token to use to fetch
subsequent updates, and whether we returned fewer rows than exists
between the requested tokens due to the limit.
2016-03-15 13:01:43 -04:00
The token returned can be used in a subsequent call to this
function to get further updatees.
The updates are a list of 2-tuples of stream ID and the row data
"""
if last_id == current_id:
return [], current_id, False
def get_all_updated_pushers_rows_txn(txn):
sql = """
SELECT id, user_name, app_id, pushkey
FROM pushers
WHERE ? < id AND id <= ?
ORDER BY id ASC LIMIT ?
"""
txn.execute(sql, (last_id, current_id, limit))
updates = [
(stream_id, (user_name, app_id, pushkey, False))
for stream_id, user_name, app_id, pushkey in txn
]
sql = """
SELECT stream_id, user_id, app_id, pushkey
FROM deleted_pushers
WHERE ? < stream_id AND stream_id <= ?
ORDER BY stream_id ASC LIMIT ?
"""
txn.execute(sql, (last_id, current_id, limit))
updates.extend(
(stream_id, (user_name, app_id, pushkey, True))
for stream_id, user_name, app_id, pushkey in txn
)
updates.sort() # Sort so that they're ordered by stream id
limited = False
upper_bound = current_id
if len(updates) >= limit:
limited = True
upper_bound = updates[-1][0]
return updates, upper_bound, limited
return await self.db_pool.runInteraction(
"get_all_updated_pushers_rows", get_all_updated_pushers_rows_txn
)
@cached(num_args=1, max_entries=15000)
async def get_if_user_has_pusher(self, user_id):
2016-11-04 06:53:42 -04:00
# This only exists for the cachedList decorator
2016-11-04 06:48:20 -04:00
raise NotImplementedError()
@cachedList(
cached_method_name="get_if_user_has_pusher", list_name="user_ids", num_args=1,
)
async def get_if_users_have_pushers(self, user_ids):
rows = await self.db_pool.simple_select_many_batch(
2019-06-20 05:32:02 -04:00
table="pushers",
column="user_name",
iterable=user_ids,
2019-06-20 05:32:02 -04:00
retcols=["user_name"],
desc="get_if_users_have_pushers",
)
result = {user_id: False for user_id in user_ids}
2019-06-20 05:32:02 -04:00
result.update({r["user_name"]: True for r in rows})
return result
async def update_pusher_last_stream_ordering(
2020-02-25 11:56:55 -05:00
self, app_id, pushkey, user_id, last_stream_ordering
) -> None:
await self.db_pool.simple_update_one(
2020-02-25 11:56:55 -05:00
"pushers",
{"app_id": app_id, "pushkey": pushkey, "user_name": user_id},
{"last_stream_ordering": last_stream_ordering},
desc="update_pusher_last_stream_ordering",
)
async def update_pusher_last_stream_ordering_and_success(
self,
app_id: str,
pushkey: str,
user_id: str,
last_stream_ordering: int,
last_success: int,
) -> bool:
2020-02-25 11:56:55 -05:00
"""Update the last stream ordering position we've processed up to for
the given pusher.
Args:
app_id
pushkey
user_id
last_stream_ordering
last_success
2020-02-25 11:56:55 -05:00
Returns:
True if the pusher still exists; False if it has been deleted.
2020-02-25 11:56:55 -05:00
"""
updated = await self.db_pool.simple_update(
2020-02-25 11:56:55 -05:00
table="pushers",
keyvalues={"app_id": app_id, "pushkey": pushkey, "user_name": user_id},
updatevalues={
"last_stream_ordering": last_stream_ordering,
"last_success": last_success,
},
desc="update_pusher_last_stream_ordering_and_success",
)
return bool(updated)
async def update_pusher_failing_since(
self, app_id, pushkey, user_id, failing_since
) -> None:
await self.db_pool.simple_update(
2020-02-25 11:56:55 -05:00
table="pushers",
keyvalues={"app_id": app_id, "pushkey": pushkey, "user_name": user_id},
updatevalues={"failing_since": failing_since},
desc="update_pusher_failing_since",
)
async def get_throttle_params_by_room(self, pusher_id):
res = await self.db_pool.simple_select_list(
2020-02-25 11:56:55 -05:00
"pusher_throttle",
{"pusher": pusher_id},
["room_id", "last_sent_ts", "throttle_ms"],
desc="get_throttle_params_by_room",
)
params_by_room = {}
for row in res:
params_by_room[row["room_id"]] = {
"last_sent_ts": row["last_sent_ts"],
"throttle_ms": row["throttle_ms"],
}
return params_by_room
async def set_throttle_params(self, pusher_id, room_id, params) -> None:
2020-02-25 11:56:55 -05:00
# no need to lock because `pusher_throttle` has a primary key on
# (pusher, room_id) so simple_upsert will retry
await self.db_pool.simple_upsert(
2020-02-25 11:56:55 -05:00
"pusher_throttle",
{"pusher": pusher_id, "room_id": room_id},
params,
desc="set_throttle_params",
lock=False,
)
class PusherStore(PusherWorkerStore):
def get_pushers_stream_token(self):
return self._pushers_id_gen.get_current_token()
async def add_pusher(
self,
user_id,
access_token,
kind,
app_id,
app_display_name,
device_display_name,
pushkey,
pushkey_ts,
lang,
data,
last_stream_ordering,
profile_tag="",
) -> None:
async with self._pushers_id_gen.get_next() as stream_id:
# no need to lock because `pushers` has a unique key on
# (app_id, pushkey, user_name) so simple_upsert will retry
await self.db_pool.simple_upsert(
table="pushers",
keyvalues={"app_id": app_id, "pushkey": pushkey, "user_name": user_id},
values={
"access_token": access_token,
"kind": kind,
"app_display_name": app_display_name,
"device_display_name": device_display_name,
"ts": pushkey_ts,
"lang": lang,
"data": bytearray(encode_canonical_json(data)),
"last_stream_ordering": last_stream_ordering,
"profile_tag": profile_tag,
"id": stream_id,
},
desc="add_pusher",
lock=False,
)
2016-05-13 06:25:02 -04:00
user_has_pusher = self.get_if_user_has_pusher.cache.get(
(user_id,), None, update_metrics=False
)
if user_has_pusher is not True:
# invalidate, since we the user might not have had a pusher before
await self.db_pool.runInteraction(
"add_pusher",
self._invalidate_cache_and_stream,
self.get_if_user_has_pusher,
(user_id,),
)
async def delete_pusher_by_app_id_pushkey_user_id(
self, app_id, pushkey, user_id
) -> None:
2016-03-15 13:01:43 -04:00
def delete_pusher_txn(txn, stream_id):
self._invalidate_cache_and_stream(
txn, self.get_if_user_has_pusher, (user_id,)
)
2016-05-13 06:25:02 -04:00
self.db_pool.simple_delete_one_txn(
2016-03-15 13:01:43 -04:00
txn,
"pushers",
{"app_id": app_id, "pushkey": pushkey, "user_name": user_id},
2016-03-15 13:01:43 -04:00
)
# it's possible for us to end up with duplicate rows for
# (app_id, pushkey, user_id) at different stream_ids, but that
# doesn't really matter.
self.db_pool.simple_insert_txn(
2016-03-15 13:01:43 -04:00
txn,
table="deleted_pushers",
values={
"stream_id": stream_id,
"app_id": app_id,
"pushkey": pushkey,
"user_id": user_id,
},
2016-03-15 13:01:43 -04:00
)
2016-05-13 06:25:02 -04:00
async with self._pushers_id_gen.get_next() as stream_id:
await self.db_pool.runInteraction(
"delete_pusher", delete_pusher_txn, stream_id
)