Skip to content
This repository has been archived by the owner on Apr 26, 2024. It is now read-only.

Move things out of SQLBaseStore #6454

Merged
merged 7 commits into from
Dec 4, 2019
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions changelog.d/6454.misc
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
Move data store specific code out of `SQLBaseStore`.
5 changes: 3 additions & 2 deletions synapse/replication/slave/storage/_base.py
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,8 @@

import six

from synapse.storage._base import _CURRENT_STATE_CACHE_NAME, SQLBaseStore
from synapse.storage._base import SQLBaseStore
from synapse.storage.data_stores.main.cache import CURRENT_STATE_CACHE_NAME
from synapse.storage.engines import PostgresEngine

from ._slaved_id_tracker import SlavedIdTracker
Expand Down Expand Up @@ -62,7 +63,7 @@ def process_replication_rows(self, stream_name, token, rows):
if stream_name == "caches":
self._cache_id_gen.advance(token)
for row in rows:
if row.cache_func == _CURRENT_STATE_CACHE_NAME:
if row.cache_func == CURRENT_STATE_CACHE_NAME:
room_id = row.keys[0]
members_changed = set(row.keys[1:])
self._invalidate_state_caches(room_id, members_changed)
Expand Down
182 changes: 0 additions & 182 deletions synapse/storage/_base.py
Original file line number Diff line number Diff line change
Expand Up @@ -14,11 +14,9 @@
# 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 itertools
import logging
import random
import sys
import threading
import time
from typing import Iterable, Tuple

Expand All @@ -35,8 +33,6 @@
from synapse.metrics.background_process_metrics import run_as_background_process
from synapse.storage.engines import PostgresEngine, Sqlite3Engine
from synapse.types import get_domain_from_id
from synapse.util import batch_iter
from synapse.util.caches.descriptors import Cache
from synapse.util.stringutils import exception_to_unicode

# import a function which will return a monotonic time, in seconds
Expand Down Expand Up @@ -79,10 +75,6 @@
"event_search": "event_search_event_id_idx",
}

# This is a special cache name we use to batch multiple invalidations of caches
# based on the current state when notifying workers over replication.
_CURRENT_STATE_CACHE_NAME = "cs_cache_fake"


class LoggingTransaction(object):
"""An object that almost-transparently proxies for the 'txn' object
Expand Down Expand Up @@ -237,23 +229,11 @@ def __init__(self, db_conn, hs):
# to watch it
self._txn_perf_counters = PerformanceCounters()

self._get_event_cache = Cache(
"*getEvent*", keylen=3, max_entries=hs.config.event_cache_size
)

self._event_fetch_lock = threading.Condition()
self._event_fetch_list = []
self._event_fetch_ongoing = 0

self._pending_ds = []

self.database_engine = hs.database_engine

# A set of tables that are not safe to use native upserts in.
self._unsafe_to_upsert_tables = set(UNIQUE_INDEX_BACKGROUND_UPDATES.keys())

self._account_validity = self.hs.config.account_validity

# We add the user_directory_search table to the blacklist on SQLite
# because the existing search table does not have an index, making it
# unsafe to use native upserts.
Expand All @@ -272,14 +252,6 @@ def __init__(self, db_conn, hs):

self.rand = random.SystemRandom()

if self._account_validity.enabled:
self._clock.call_later(
0.0,
run_as_background_process,
"account_validity_set_expiration_dates",
self._set_expiration_date_when_missing,
)

@defer.inlineCallbacks
def _check_safe_to_upsert(self):
"""
Expand Down Expand Up @@ -312,62 +284,6 @@ def _check_safe_to_upsert(self):
self._check_safe_to_upsert,
)

@defer.inlineCallbacks
def _set_expiration_date_when_missing(self):
"""
Retrieves the list of registered users that don't have an expiration date, and
adds an expiration date for each of them.
"""

def select_users_with_no_expiration_date_txn(txn):
"""Retrieves the list of registered users with no expiration date from the
database, filtering out deactivated users.
"""
sql = (
"SELECT users.name FROM users"
" LEFT JOIN account_validity ON (users.name = account_validity.user_id)"
" WHERE account_validity.user_id is NULL AND users.deactivated = 0;"
)
txn.execute(sql, [])

res = self.cursor_to_dict(txn)
if res:
for user in res:
self.set_expiration_date_for_user_txn(
txn, user["name"], use_delta=True
)

yield self.runInteraction(
"get_users_with_no_expiration_date",
select_users_with_no_expiration_date_txn,
)

def set_expiration_date_for_user_txn(self, txn, user_id, use_delta=False):
"""Sets an expiration date to the account with the given user ID.

Args:
user_id (str): User ID to set an expiration date for.
use_delta (bool): If set to False, the expiration date for the user will be
now + validity period. If set to True, this expiration date will be a
random value in the [now + period - d ; now + period] range, d being a
delta equal to 10% of the validity period.
"""
now_ms = self._clock.time_msec()
expiration_ts = now_ms + self._account_validity.period

if use_delta:
expiration_ts = self.rand.randrange(
expiration_ts - self._account_validity.startup_job_max_delta,
expiration_ts,
)

self._simple_upsert_txn(
txn,
"account_validity",
keyvalues={"user_id": user_id},
values={"expiration_ts_ms": expiration_ts, "email_sent": False},
)

def start_profiling(self):
self._previous_loop_ts = monotonic_time()

Expand Down Expand Up @@ -1400,47 +1316,6 @@ def _get_cache_dict(

return cache, min_val

def _invalidate_cache_and_stream(self, txn, cache_func, keys):
"""Invalidates the cache and adds it to the cache stream so slaves
will know to invalidate their caches.

This should only be used to invalidate caches where slaves won't
otherwise know from other replication streams that the cache should
be invalidated.
"""
txn.call_after(cache_func.invalidate, keys)
self._send_invalidation_to_replication(txn, cache_func.__name__, keys)

def _invalidate_state_caches_and_stream(self, txn, room_id, members_changed):
"""Special case invalidation of caches based on current state.

We special case this so that we can batch the cache invalidations into a
single replication poke.

Args:
txn
room_id (str): Room where state changed
members_changed (iterable[str]): The user_ids of members that have changed
"""
txn.call_after(self._invalidate_state_caches, room_id, members_changed)

if members_changed:
# We need to be careful that the size of the `members_changed` list
# isn't so large that it causes problems sending over replication, so we
# send them in chunks.
# Max line length is 16K, and max user ID length is 255, so 50 should
# be safe.
for chunk in batch_iter(members_changed, 50):
keys = itertools.chain([room_id], chunk)
self._send_invalidation_to_replication(
txn, _CURRENT_STATE_CACHE_NAME, keys
)
else:
# if no members changed, we still need to invalidate the other caches.
self._send_invalidation_to_replication(
txn, _CURRENT_STATE_CACHE_NAME, [room_id]
)

def _invalidate_state_caches(self, room_id, members_changed):
"""Invalidates caches that are based on the current state, but does
not stream invalidations down replication.
Expand Down Expand Up @@ -1474,63 +1349,6 @@ def _attempt_to_invalidate_cache(self, cache_name, key):
# which is fine.
pass

def _send_invalidation_to_replication(self, txn, cache_name, keys):
"""Notifies replication that given cache has been invalidated.

Note that this does *not* invalidate the cache locally.

Args:
txn
cache_name (str)
keys (iterable[str])
"""

if isinstance(self.database_engine, PostgresEngine):
# get_next() returns a context manager which is designed to wrap
# the transaction. However, we want to only get an ID when we want
# to use it, here, so we need to call __enter__ manually, and have
# __exit__ called after the transaction finishes.
ctx = self._cache_id_gen.get_next()
stream_id = ctx.__enter__()
txn.call_on_exception(ctx.__exit__, None, None, None)
txn.call_after(ctx.__exit__, None, None, None)
txn.call_after(self.hs.get_notifier().on_new_replication_data)

self._simple_insert_txn(
txn,
table="cache_invalidation_stream",
values={
"stream_id": stream_id,
"cache_func": cache_name,
"keys": list(keys),
"invalidation_ts": self.clock.time_msec(),
},
)

def get_all_updated_caches(self, last_id, current_id, limit):
if last_id == current_id:
return defer.succeed([])

def get_all_updated_caches_txn(txn):
# We purposefully don't bound by the current token, as we want to
# send across cache invalidations as quickly as possible. Cache
# invalidations are idempotent, so duplicates are fine.
sql = (
"SELECT stream_id, cache_func, keys, invalidation_ts"
" FROM cache_invalidation_stream"
" WHERE stream_id > ? ORDER BY stream_id ASC LIMIT ?"
)
txn.execute(sql, (last_id, limit))
return txn.fetchall()

return self.runInteraction("get_all_updated_caches", get_all_updated_caches_txn)

def get_cache_stream_token(self):
if self._cache_id_gen:
return self._cache_id_gen.get_current_token()
else:
return 0

def _simple_select_list_paginate(
self,
table,
Expand Down
2 changes: 2 additions & 0 deletions synapse/storage/data_stores/main/__init__.py
Original file line number Diff line number Diff line change
Expand Up @@ -32,6 +32,7 @@

from .account_data import AccountDataStore
from .appservice import ApplicationServiceStore, ApplicationServiceTransactionStore
from .cache import CacheInvalidationStore
from .client_ips import ClientIpStore
from .deviceinbox import DeviceInboxStore
from .devices import DeviceStore
Expand Down Expand Up @@ -110,6 +111,7 @@ class DataStore(
MonthlyActiveUsersStore,
StatsStore,
RelationsStore,
CacheInvalidationStore,
):
def __init__(self, db_conn, hs):
self.hs = hs
Expand Down
Loading