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

498 lines
17 KiB
Python
Raw Normal View History

2014-08-12 10:10:52 -04:00
# -*- coding: utf-8 -*-
2016-01-06 23:26:29 -05:00
# Copyright 2014-2016 OpenMarket Ltd
2014-08-12 10:10:52 -04:00
#
# 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 collections import namedtuple
from typing import Iterable, List, Optional, Tuple
2014-08-12 10:10:52 -04:00
2018-08-30 10:19:58 -04:00
from canonicaljson import encode_canonical_json
from synapse.metrics.background_process_metrics import wrap_as_background_process
from synapse.storage._base import SQLBaseStore, db_to_json
from synapse.storage.database import DatabasePool, LoggingTransaction
from synapse.storage.engines import PostgresEngine, Sqlite3Engine
from synapse.types import JsonDict
from synapse.util.caches.expiringcache import ExpiringCache
2018-07-09 02:09:20 -04:00
db_binary_type = memoryview
2018-05-29 11:42:43 -04:00
2014-08-12 10:10:52 -04:00
logger = logging.getLogger(__name__)
_TransactionRow = namedtuple(
"_TransactionRow",
("id", "transaction_id", "destination", "ts", "response_code", "response_json"),
)
_UpdateTransactionRow = namedtuple(
"_TransactionRow", ("response_code", "response_json")
)
SENTINEL = object()
class TransactionWorkerStore(SQLBaseStore):
def __init__(self, database: DatabasePool, db_conn, hs):
super().__init__(database, db_conn, hs)
if hs.config.run_background_tasks:
self._clock.looping_call(self._cleanup_transactions, 30 * 60 * 1000)
@wrap_as_background_process("cleanup_transactions")
async def _cleanup_transactions(self) -> None:
now = self._clock.time_msec()
month_ago = now - 30 * 24 * 60 * 60 * 1000
def _cleanup_transactions_txn(txn):
txn.execute("DELETE FROM received_transactions WHERE ts < ?", (month_ago,))
await self.db_pool.runInteraction(
"_cleanup_transactions", _cleanup_transactions_txn
)
class TransactionStore(TransactionWorkerStore):
"""A collection of queries for handling PDUs."""
2014-12-10 05:16:09 -05:00
def __init__(self, database: DatabasePool, db_conn, hs):
super().__init__(database, db_conn, hs)
self._destination_retry_cache = ExpiringCache(
cache_name="get_destination_retry_timings",
clock=self._clock,
expiry_ms=5 * 60 * 1000,
)
async def get_received_txn_response(
self, transaction_id: str, origin: str
) -> Optional[Tuple[int, JsonDict]]:
2014-08-12 10:10:52 -04:00
"""For an incoming transaction from a given origin, check if we have
already responded to it. If so, return the response code and response
body (as a dict).
Args:
transaction_id
origin
2014-08-12 10:10:52 -04:00
Returns:
None if we have not previously responded to this transaction or a
2-tuple of (int, dict)
2014-08-12 10:10:52 -04:00
"""
return await self.db_pool.runInteraction(
"get_received_txn_response",
self._get_received_txn_response,
transaction_id,
origin,
2014-08-12 10:10:52 -04:00
)
def _get_received_txn_response(self, txn, transaction_id, origin):
result = self.db_pool.simple_select_one_txn(
2015-03-20 11:05:44 -04:00
txn,
table="received_transactions",
keyvalues={"transaction_id": transaction_id, "origin": origin},
retcols=(
"transaction_id",
"origin",
"ts",
"response_code",
"response_json",
"has_been_referenced",
),
2015-03-20 11:05:44 -04:00
allow_none=True,
)
2014-08-12 10:10:52 -04:00
if result and result["response_code"]:
2018-08-30 10:19:58 -04:00
return result["response_code"], db_to_json(result["response_json"])
2014-08-12 10:10:52 -04:00
else:
return None
async def set_received_txn_response(
self, transaction_id: str, origin: str, code: int, response_dict: JsonDict
) -> None:
"""Persist the response we returned for an incoming transaction, and
2014-08-12 10:10:52 -04:00
should return for subsequent transactions with the same transaction_id
and origin.
Args:
transaction_id: The incoming transaction ID.
origin: The origin server.
code: The response code.
response_dict: The response, to be encoded into JSON.
2014-08-12 10:10:52 -04:00
"""
await self.db_pool.simple_insert(
table="received_transactions",
values={
2015-03-23 09:43:21 -04:00
"transaction_id": transaction_id,
"origin": origin,
"response_code": code,
2018-05-29 11:42:43 -04:00
"response_json": db_binary_type(encode_canonical_json(response_dict)),
2016-08-22 11:29:46 -04:00
"ts": self._clock.time_msec(),
},
or_ignore=True,
desc="set_received_txn_response",
2015-03-23 09:43:21 -04:00
)
2014-08-12 10:10:52 -04:00
async def get_destination_retry_timings(self, destination):
"""Gets the current retry timings (if any) for a given destination.
2014-12-10 05:16:09 -05:00
Args:
destination (str)
2014-12-10 05:16:09 -05:00
Returns:
None if not retrying
2015-03-23 09:43:21 -04:00
Otherwise a dict for the retry scheme
"""
result = self._destination_retry_cache.get(destination, SENTINEL)
if result is not SENTINEL:
return result
result = await self.db_pool.runInteraction(
"get_destination_retry_timings",
self._get_destination_retry_timings,
destination,
)
2014-12-10 05:16:09 -05:00
# We don't hugely care about race conditions between getting and
# invalidating the cache, since we time out fairly quickly anyway.
self._destination_retry_cache[destination] = result
return result
2015-03-23 09:43:21 -04:00
def _get_destination_retry_timings(self, txn, destination):
result = self.db_pool.simple_select_one_txn(
2015-03-23 09:43:21 -04:00
txn,
table="destinations",
keyvalues={"destination": destination},
retcols=("destination", "failure_ts", "retry_last_ts", "retry_interval"),
2015-03-23 09:43:21 -04:00
allow_none=True,
)
# check we have a row and retry_last_ts is not null or zero
# (retry_last_ts can't be negative)
if result and result["retry_last_ts"]:
2015-03-23 09:43:21 -04:00
return result
else:
return None
2014-12-10 05:16:09 -05:00
async def set_destination_retry_timings(
self,
destination: str,
failure_ts: Optional[int],
retry_last_ts: int,
retry_interval: int,
) -> None:
"""Sets the current retry timings for a given destination.
2021-02-12 11:01:48 -05:00
Both timings should be zero if retrying is no longer occurring.
2014-12-10 05:16:09 -05:00
Args:
destination
failure_ts: when the server started failing (ms since epoch)
retry_last_ts: time of last retry attempt in unix epoch ms
retry_interval: how long until next retry in ms
"""
2014-12-10 05:16:09 -05:00
self._destination_retry_cache.pop(destination, None)
Use autocommit mode for single statement DB functions. (#8542) Autocommit means that we don't wrap the functions in transactions, and instead get executed directly. Introduced in #8456. This will help: 1. reduce the number of `could not serialize access due to concurrent delete` errors that we see (though there are a few functions that often cause serialization errors that we don't fix here); 2. improve the DB performance, as it no longer needs to deal with the overhead of `REPEATABLE READ` isolation levels; and 3. improve wall clock speed of these functions, as we no longer need to send `BEGIN` and `COMMIT` to the DB. Some notes about the differences between autocommit mode and our default `REPEATABLE READ` transactions: 1. Currently `autocommit` only applies when using PostgreSQL, and is ignored when using SQLite (due to silliness with [Twisted DB classes](https://twistedmatrix.com/trac/ticket/9998)). 2. Autocommit functions may get retried on error, which means they can get applied *twice* (or more) to the DB (since they are not in a transaction the previous call would not get rolled back). This means that the functions need to be idempotent (or otherwise not care about being called multiple times). Read queries, simple deletes, and updates/upserts that replace rows (rather than generating new values from existing rows) are all idempotent. 3. Autocommit functions no longer get executed in [`REPEATABLE READ`](https://www.postgresql.org/docs/current/transaction-iso.html) isolation level, and so data can change queries, which is fine for single statement queries.
2020-10-14 10:50:59 -04:00
if self.database_engine.can_native_upsert:
return await self.db_pool.runInteraction(
"set_destination_retry_timings",
self._set_destination_retry_timings_native,
destination,
failure_ts,
retry_last_ts,
retry_interval,
db_autocommit=True, # Safe as its a single upsert
)
else:
return await self.db_pool.runInteraction(
"set_destination_retry_timings",
self._set_destination_retry_timings_emulated,
destination,
failure_ts,
retry_last_ts,
retry_interval,
)
2014-12-10 05:16:09 -05:00
Use autocommit mode for single statement DB functions. (#8542) Autocommit means that we don't wrap the functions in transactions, and instead get executed directly. Introduced in #8456. This will help: 1. reduce the number of `could not serialize access due to concurrent delete` errors that we see (though there are a few functions that often cause serialization errors that we don't fix here); 2. improve the DB performance, as it no longer needs to deal with the overhead of `REPEATABLE READ` isolation levels; and 3. improve wall clock speed of these functions, as we no longer need to send `BEGIN` and `COMMIT` to the DB. Some notes about the differences between autocommit mode and our default `REPEATABLE READ` transactions: 1. Currently `autocommit` only applies when using PostgreSQL, and is ignored when using SQLite (due to silliness with [Twisted DB classes](https://twistedmatrix.com/trac/ticket/9998)). 2. Autocommit functions may get retried on error, which means they can get applied *twice* (or more) to the DB (since they are not in a transaction the previous call would not get rolled back). This means that the functions need to be idempotent (or otherwise not care about being called multiple times). Read queries, simple deletes, and updates/upserts that replace rows (rather than generating new values from existing rows) are all idempotent. 3. Autocommit functions no longer get executed in [`REPEATABLE READ`](https://www.postgresql.org/docs/current/transaction-iso.html) isolation level, and so data can change queries, which is fine for single statement queries.
2020-10-14 10:50:59 -04:00
def _set_destination_retry_timings_native(
self, txn, destination, failure_ts, retry_last_ts, retry_interval
):
Use autocommit mode for single statement DB functions. (#8542) Autocommit means that we don't wrap the functions in transactions, and instead get executed directly. Introduced in #8456. This will help: 1. reduce the number of `could not serialize access due to concurrent delete` errors that we see (though there are a few functions that often cause serialization errors that we don't fix here); 2. improve the DB performance, as it no longer needs to deal with the overhead of `REPEATABLE READ` isolation levels; and 3. improve wall clock speed of these functions, as we no longer need to send `BEGIN` and `COMMIT` to the DB. Some notes about the differences between autocommit mode and our default `REPEATABLE READ` transactions: 1. Currently `autocommit` only applies when using PostgreSQL, and is ignored when using SQLite (due to silliness with [Twisted DB classes](https://twistedmatrix.com/trac/ticket/9998)). 2. Autocommit functions may get retried on error, which means they can get applied *twice* (or more) to the DB (since they are not in a transaction the previous call would not get rolled back). This means that the functions need to be idempotent (or otherwise not care about being called multiple times). Read queries, simple deletes, and updates/upserts that replace rows (rather than generating new values from existing rows) are all idempotent. 3. Autocommit functions no longer get executed in [`REPEATABLE READ`](https://www.postgresql.org/docs/current/transaction-iso.html) isolation level, and so data can change queries, which is fine for single statement queries.
2020-10-14 10:50:59 -04:00
assert self.database_engine.can_native_upsert
# Upsert retry time interval if retry_interval is zero (i.e. we're
# resetting it) or greater than the existing retry interval.
#
# WARNING: This is executed in autocommit, so we shouldn't add any more
# SQL calls in here (without being very careful).
sql = """
INSERT INTO destinations (
destination, failure_ts, retry_last_ts, retry_interval
)
VALUES (?, ?, ?, ?)
ON CONFLICT (destination) DO UPDATE SET
failure_ts = EXCLUDED.failure_ts,
retry_last_ts = EXCLUDED.retry_last_ts,
retry_interval = EXCLUDED.retry_interval
WHERE
EXCLUDED.retry_interval = 0
OR destinations.retry_interval IS NULL
OR destinations.retry_interval < EXCLUDED.retry_interval
"""
Use autocommit mode for single statement DB functions. (#8542) Autocommit means that we don't wrap the functions in transactions, and instead get executed directly. Introduced in #8456. This will help: 1. reduce the number of `could not serialize access due to concurrent delete` errors that we see (though there are a few functions that often cause serialization errors that we don't fix here); 2. improve the DB performance, as it no longer needs to deal with the overhead of `REPEATABLE READ` isolation levels; and 3. improve wall clock speed of these functions, as we no longer need to send `BEGIN` and `COMMIT` to the DB. Some notes about the differences between autocommit mode and our default `REPEATABLE READ` transactions: 1. Currently `autocommit` only applies when using PostgreSQL, and is ignored when using SQLite (due to silliness with [Twisted DB classes](https://twistedmatrix.com/trac/ticket/9998)). 2. Autocommit functions may get retried on error, which means they can get applied *twice* (or more) to the DB (since they are not in a transaction the previous call would not get rolled back). This means that the functions need to be idempotent (or otherwise not care about being called multiple times). Read queries, simple deletes, and updates/upserts that replace rows (rather than generating new values from existing rows) are all idempotent. 3. Autocommit functions no longer get executed in [`REPEATABLE READ`](https://www.postgresql.org/docs/current/transaction-iso.html) isolation level, and so data can change queries, which is fine for single statement queries.
2020-10-14 10:50:59 -04:00
txn.execute(sql, (destination, failure_ts, retry_last_ts, retry_interval))
Use autocommit mode for single statement DB functions. (#8542) Autocommit means that we don't wrap the functions in transactions, and instead get executed directly. Introduced in #8456. This will help: 1. reduce the number of `could not serialize access due to concurrent delete` errors that we see (though there are a few functions that often cause serialization errors that we don't fix here); 2. improve the DB performance, as it no longer needs to deal with the overhead of `REPEATABLE READ` isolation levels; and 3. improve wall clock speed of these functions, as we no longer need to send `BEGIN` and `COMMIT` to the DB. Some notes about the differences between autocommit mode and our default `REPEATABLE READ` transactions: 1. Currently `autocommit` only applies when using PostgreSQL, and is ignored when using SQLite (due to silliness with [Twisted DB classes](https://twistedmatrix.com/trac/ticket/9998)). 2. Autocommit functions may get retried on error, which means they can get applied *twice* (or more) to the DB (since they are not in a transaction the previous call would not get rolled back). This means that the functions need to be idempotent (or otherwise not care about being called multiple times). Read queries, simple deletes, and updates/upserts that replace rows (rather than generating new values from existing rows) are all idempotent. 3. Autocommit functions no longer get executed in [`REPEATABLE READ`](https://www.postgresql.org/docs/current/transaction-iso.html) isolation level, and so data can change queries, which is fine for single statement queries.
2020-10-14 10:50:59 -04:00
def _set_destination_retry_timings_emulated(
self, txn, destination, failure_ts, retry_last_ts, retry_interval
):
self.database_engine.lock_table(txn, "destinations")
# We need to be careful here as the data may have changed from under us
# due to a worker setting the timings.
prev_row = self.db_pool.simple_select_one_txn(
txn,
table="destinations",
keyvalues={"destination": destination},
retcols=("failure_ts", "retry_last_ts", "retry_interval"),
allow_none=True,
2015-03-23 09:43:21 -04:00
)
if not prev_row:
self.db_pool.simple_insert_txn(
txn,
table="destinations",
values={
"destination": destination,
"failure_ts": failure_ts,
"retry_last_ts": retry_last_ts,
"retry_interval": retry_interval,
},
)
elif (
retry_interval == 0
or prev_row["retry_interval"] is None
or prev_row["retry_interval"] < retry_interval
):
self.db_pool.simple_update_one_txn(
txn,
"destinations",
keyvalues={"destination": destination},
updatevalues={
"failure_ts": failure_ts,
"retry_last_ts": retry_last_ts,
"retry_interval": retry_interval,
},
)
async def store_destination_rooms_entries(
self,
destinations: Iterable[str],
room_id: str,
stream_ordering: int,
) -> None:
"""
Updates or creates `destination_rooms` entries in batch for a single event.
Args:
destinations: list of destinations
room_id: the room_id of the event
stream_ordering: the stream_ordering of the event
"""
return await self.db_pool.runInteraction(
"store_destination_rooms_entries",
self._store_destination_rooms_entries_txn,
destinations,
room_id,
stream_ordering,
)
def _store_destination_rooms_entries_txn(
self,
txn: LoggingTransaction,
destinations: Iterable[str],
room_id: str,
stream_ordering: int,
) -> None:
# ensure we have a `destinations` row for this destination, as there is
# a foreign key constraint.
if isinstance(self.database_engine, PostgresEngine):
q = """
INSERT INTO destinations (destination)
VALUES (?)
ON CONFLICT DO NOTHING;
"""
elif isinstance(self.database_engine, Sqlite3Engine):
q = """
INSERT OR IGNORE INTO destinations (destination)
VALUES (?);
"""
else:
raise RuntimeError("Unknown database engine")
txn.execute_batch(q, ((destination,) for destination in destinations))
rows = [(destination, room_id) for destination in destinations]
self.db_pool.simple_upsert_many_txn(
txn,
"destination_rooms",
["destination", "room_id"],
rows,
["stream_ordering"],
[(stream_ordering,)] * len(rows),
)
async def get_destination_last_successful_stream_ordering(
self, destination: str
) -> Optional[int]:
"""
Gets the stream ordering of the PDU most-recently successfully sent
to the specified destination, or None if this information has not been
tracked yet.
Args:
destination: the destination to query
"""
return await self.db_pool.simple_select_one_onecol(
"destinations",
{"destination": destination},
"last_successful_stream_ordering",
allow_none=True,
desc="get_last_successful_stream_ordering",
)
async def set_destination_last_successful_stream_ordering(
self, destination: str, last_successful_stream_ordering: int
) -> None:
"""
Marks that we have successfully sent the PDUs up to and including the
one specified.
Args:
destination: the destination we have successfully sent to
last_successful_stream_ordering: the stream_ordering of the most
recent successfully-sent PDU
"""
return await self.db_pool.simple_upsert(
"destinations",
keyvalues={"destination": destination},
values={"last_successful_stream_ordering": last_successful_stream_ordering},
desc="set_last_successful_stream_ordering",
)
async def get_catch_up_room_event_ids(
self,
destination: str,
last_successful_stream_ordering: int,
) -> List[str]:
"""
Returns at most 50 event IDs and their corresponding stream_orderings
that correspond to the oldest events that have not yet been sent to
the destination.
Args:
destination: the destination in question
last_successful_stream_ordering: the stream_ordering of the
most-recently successfully-transmitted event to the destination
Returns:
list of event_ids
"""
return await self.db_pool.runInteraction(
"get_catch_up_room_event_ids",
self._get_catch_up_room_event_ids_txn,
destination,
last_successful_stream_ordering,
)
@staticmethod
def _get_catch_up_room_event_ids_txn(
txn: LoggingTransaction,
destination: str,
last_successful_stream_ordering: int,
) -> List[str]:
q = """
SELECT event_id FROM destination_rooms
JOIN events USING (stream_ordering)
WHERE destination = ?
AND stream_ordering > ?
ORDER BY stream_ordering
LIMIT 50
"""
txn.execute(
q,
(destination, last_successful_stream_ordering),
)
event_ids = [row[0] for row in txn]
return event_ids
async def get_catch_up_outstanding_destinations(
self, after_destination: Optional[str]
) -> List[str]:
"""
Gets at most 25 destinations which have outstanding PDUs to be caught up,
and are not being backed off from
Args:
after_destination:
If provided, all destinations must be lexicographically greater
than this one.
Returns:
list of up to 25 destinations with outstanding catch-up.
These are the lexicographically first destinations which are
lexicographically greater than after_destination (if provided).
"""
time = self.hs.get_clock().time_msec()
return await self.db_pool.runInteraction(
"get_catch_up_outstanding_destinations",
self._get_catch_up_outstanding_destinations_txn,
time,
after_destination,
)
@staticmethod
def _get_catch_up_outstanding_destinations_txn(
txn: LoggingTransaction, now_time_ms: int, after_destination: Optional[str]
) -> List[str]:
q = """
SELECT DISTINCT destination FROM destinations
INNER JOIN destination_rooms USING (destination)
WHERE
stream_ordering > last_successful_stream_ordering
AND destination > ?
AND (
retry_last_ts IS NULL OR
retry_last_ts + retry_interval < ?
)
ORDER BY destination
LIMIT 25
"""
txn.execute(
q,
(
# everything is lexicographically greater than "" so this gives
# us the first batch of up to 25.
after_destination or "",
now_time_ms,
),
)
destinations = [row[0] for row in txn]
return destinations