From e8f30a76caa4394ebb3e77c56df951e3626b3fdd Mon Sep 17 00:00:00 2001 From: David Robertson Date: Fri, 30 Sep 2022 11:54:53 +0100 Subject: [PATCH] Fix overflows in /messages backfill calculation (#13936) * Reproduce bug * Compute `least_function` first * Substitute `least_function` with an f-string * Bugfix: avoid overflow Co-authored-by: Eric Eastwood --- changelog.d/13936.feature | 1 + .../databases/main/event_federation.py | 82 ++++++++++++------- tests/storage/test_event_federation.py | 61 +++++++++++--- 3 files changed, 103 insertions(+), 41 deletions(-) create mode 100644 changelog.d/13936.feature diff --git a/changelog.d/13936.feature b/changelog.d/13936.feature new file mode 100644 index 0000000000..d86bf7ed80 --- /dev/null +++ b/changelog.d/13936.feature @@ -0,0 +1 @@ +Exponentially backoff from backfilling the same event over and over. diff --git a/synapse/storage/databases/main/event_federation.py b/synapse/storage/databases/main/event_federation.py index 17f2fd4458..6b9a629edd 100644 --- a/synapse/storage/databases/main/event_federation.py +++ b/synapse/storage/databases/main/event_federation.py @@ -73,13 +73,30 @@ pdus_pruned_from_federation_queue = Counter( logger = logging.getLogger(__name__) -BACKFILL_EVENT_BACKOFF_UPPER_BOUND_SECONDS: int = int( - datetime.timedelta(days=7).total_seconds() -) -BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_SECONDS: int = int( - datetime.timedelta(hours=1).total_seconds() +# Parameters controlling exponential backoff between backfill failures. +# After the first failure to backfill, we wait 2 hours before trying again. If the +# second attempt fails, we wait 4 hours before trying again. If the third attempt fails, +# we wait 8 hours before trying again, ... and so on. +# +# Each successive backoff period is twice as long as the last. However we cap this +# period at a maximum of 2^8 = 256 hours: a little over 10 days. (This is the smallest +# power of 2 which yields a maximum backoff period of at least 7 days---which was the +# original maximum backoff period.) Even when we hit this cap, we will continue to +# make backfill attempts once every 10 days. +BACKFILL_EVENT_EXPONENTIAL_BACKOFF_MAXIMUM_DOUBLING_STEPS = 8 +BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_MILLISECONDS = int( + datetime.timedelta(hours=1).total_seconds() * 1000 ) +# We need a cap on the power of 2 or else the backoff period +# 2^N * (milliseconds per hour) +# will overflow when calcuated within the database. We ensure overflow does not occur +# by checking that the largest backoff period fits in a 32-bit signed integer. +_LONGEST_BACKOFF_PERIOD_MILLISECONDS = ( + 2**BACKFILL_EVENT_EXPONENTIAL_BACKOFF_MAXIMUM_DOUBLING_STEPS +) * BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_MILLISECONDS +assert 0 < _LONGEST_BACKOFF_PERIOD_MILLISECONDS <= ((2**31) - 1) + # All the info we need while iterating the DAG while backfilling @attr.s(frozen=True, slots=True, auto_attribs=True) @@ -767,7 +784,15 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas # persisted in our database yet (meaning we don't know their depth # specifically). So we need to look for the approximate depth from # the events connected to the current backwards extremeties. - sql = """ + + if isinstance(self.database_engine, PostgresEngine): + least_function = "LEAST" + elif isinstance(self.database_engine, Sqlite3Engine): + least_function = "MIN" + else: + raise RuntimeError("Unknown database engine") + + sql = f""" SELECT backward_extrem.event_id, event.depth FROM events AS event /** * Get the edge connections from the event_edges table @@ -825,7 +850,10 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas */ AND ( failed_backfill_attempt_info.event_id IS NULL - OR ? /* current_time */ >= failed_backfill_attempt_info.last_attempt_ts + /*least*/%s((1 << failed_backfill_attempt_info.num_attempts) * ? /* step */, ? /* upper bound */) + OR ? /* current_time */ >= failed_backfill_attempt_info.last_attempt_ts + ( + (1 << {least_function}(failed_backfill_attempt_info.num_attempts, ? /* max doubling steps */)) + * ? /* step */ + ) ) /** * Sort from highest (closest to the `current_depth`) to the lowest depth @@ -837,22 +865,15 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas LIMIT ? """ - if isinstance(self.database_engine, PostgresEngine): - least_function = "least" - elif isinstance(self.database_engine, Sqlite3Engine): - least_function = "min" - else: - raise RuntimeError("Unknown database engine") - txn.execute( - sql % (least_function,), + sql, ( room_id, False, current_depth, self._clock.time_msec(), - 1000 * BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_SECONDS, - 1000 * BACKFILL_EVENT_BACKOFF_UPPER_BOUND_SECONDS, + BACKFILL_EVENT_EXPONENTIAL_BACKOFF_MAXIMUM_DOUBLING_STEPS, + BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_MILLISECONDS, limit, ), ) @@ -902,7 +923,14 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas def get_insertion_event_backward_extremities_in_room_txn( txn: LoggingTransaction, room_id: str ) -> List[Tuple[str, int]]: - sql = """ + if isinstance(self.database_engine, PostgresEngine): + least_function = "LEAST" + elif isinstance(self.database_engine, Sqlite3Engine): + least_function = "MIN" + else: + raise RuntimeError("Unknown database engine") + + sql = f""" SELECT insertion_event_extremity.event_id, event.depth /* We only want insertion events that are also marked as backwards extremities */ @@ -942,7 +970,10 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas */ AND ( failed_backfill_attempt_info.event_id IS NULL - OR ? /* current_time */ >= failed_backfill_attempt_info.last_attempt_ts + /*least*/%s((1 << failed_backfill_attempt_info.num_attempts) * ? /* step */, ? /* upper bound */) + OR ? /* current_time */ >= failed_backfill_attempt_info.last_attempt_ts + ( + (1 << {least_function}(failed_backfill_attempt_info.num_attempts, ? /* max doubling steps */)) + * ? /* step */ + ) ) /** * Sort from highest (closest to the `current_depth`) to the lowest depth @@ -954,21 +985,14 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas LIMIT ? """ - if isinstance(self.database_engine, PostgresEngine): - least_function = "least" - elif isinstance(self.database_engine, Sqlite3Engine): - least_function = "min" - else: - raise RuntimeError("Unknown database engine") - txn.execute( - sql % (least_function,), + sql, ( room_id, current_depth, self._clock.time_msec(), - 1000 * BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_SECONDS, - 1000 * BACKFILL_EVENT_BACKOFF_UPPER_BOUND_SECONDS, + BACKFILL_EVENT_EXPONENTIAL_BACKOFF_MAXIMUM_DOUBLING_STEPS, + BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_MILLISECONDS, limit, ), ) diff --git a/tests/storage/test_event_federation.py b/tests/storage/test_event_federation.py index 398f338b66..59b8910907 100644 --- a/tests/storage/test_event_federation.py +++ b/tests/storage/test_event_federation.py @@ -766,9 +766,7 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase): self.store.get_backfill_points_in_room(room_id, depth_map["B"], limit=100) ) backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points] - self.assertListEqual( - backfill_event_ids, ["b6", "b5", "b4", "2", "b3", "b2", "b1"] - ) + self.assertEqual(backfill_event_ids, ["b6", "b5", "b4", "2", "b3", "b2", "b1"]) # Try at "A" backfill_points = self.get_success( @@ -814,7 +812,7 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase): ) backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points] # Only the backfill points that we didn't record earlier exist here. - self.assertListEqual(backfill_event_ids, ["b6", "2", "b1"]) + self.assertEqual(backfill_event_ids, ["b6", "2", "b1"]) def test_get_backfill_points_in_room_attempted_event_retry_after_backoff_duration( self, @@ -860,7 +858,7 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase): self.store.get_backfill_points_in_room(room_id, depth_map["A"], limit=100) ) backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points] - self.assertListEqual(backfill_event_ids, ["b3", "b2"]) + self.assertEqual(backfill_event_ids, ["b3", "b2"]) # Now advance time by 20 hours (above 2^4 because we made 4 attemps) and # see if we can now backfill it @@ -871,7 +869,48 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase): self.store.get_backfill_points_in_room(room_id, depth_map["A"], limit=100) ) backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points] - self.assertListEqual(backfill_event_ids, ["b3", "b2", "b1"]) + self.assertEqual(backfill_event_ids, ["b3", "b2", "b1"]) + + def test_get_backfill_points_in_room_works_after_many_failed_pull_attempts_that_could_naively_overflow( + self, + ) -> None: + """ + A test that reproduces #13929 (Postgres only). + + Test to make sure we can still get backfill points after many failed pull + attempts that cause us to backoff to the limit. Even if the backoff formula + would tell us to wait for more seconds than can be expressed in a 32 bit + signed int. + """ + setup_info = self._setup_room_for_backfill_tests() + room_id = setup_info.room_id + depth_map = setup_info.depth_map + + # Pretend that we have tried and failed 10 times to backfill event b1. + for _ in range(10): + self.get_success( + self.store.record_event_failed_pull_attempt(room_id, "b1", "fake cause") + ) + + # If the backoff periods grow without limit: + # After the first failed attempt, we would have backed off for 1 << 1 = 2 hours. + # After the second failed attempt we would have backed off for 1 << 2 = 4 hours, + # so after the 10th failed attempt we should backoff for 1 << 10 == 1024 hours. + # Wait 1100 hours just so we have a nice round number. + self.reactor.advance(datetime.timedelta(hours=1100).total_seconds()) + + # 1024 hours in milliseconds is 1024 * 3600000, which exceeds the largest 32 bit + # signed integer. The bug we're reproducing is that this overflow causes an + # error in postgres preventing us from fetching a set of backwards extremities + # to retry fetching. + backfill_points = self.get_success( + self.store.get_backfill_points_in_room(room_id, depth_map["A"], limit=100) + ) + + # We should aim to fetch all backoff points: b1's latest backoff period has + # expired, and we haven't tried the rest. + backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points] + self.assertEqual(backfill_event_ids, ["b3", "b2", "b1"]) def _setup_room_for_insertion_backfill_tests(self) -> _BackfillSetupInfo: """ @@ -965,9 +1004,7 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase): ) ) backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points] - self.assertListEqual( - backfill_event_ids, ["insertion_eventB", "insertion_eventA"] - ) + self.assertEqual(backfill_event_ids, ["insertion_eventB", "insertion_eventA"]) # Try at "insertion_eventA" backfill_points = self.get_success( @@ -1011,7 +1048,7 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase): ) backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points] # Only the backfill points that we didn't record earlier exist here. - self.assertListEqual(backfill_event_ids, ["insertion_eventB"]) + self.assertEqual(backfill_event_ids, ["insertion_eventB"]) def test_get_insertion_event_backward_extremities_in_room_attempted_event_retry_after_backoff_duration( self, @@ -1069,7 +1106,7 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase): ) ) backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points] - self.assertListEqual(backfill_event_ids, []) + self.assertEqual(backfill_event_ids, []) # Now advance time by 20 hours (above 2^4 because we made 4 attemps) and # see if we can now backfill it @@ -1083,7 +1120,7 @@ class EventFederationWorkerStoreTestCase(tests.unittest.HomeserverTestCase): ) ) backfill_event_ids = [backfill_point[0] for backfill_point in backfill_points] - self.assertListEqual(backfill_event_ids, ["insertion_eventA"]) + self.assertEqual(backfill_event_ids, ["insertion_eventA"]) @attr.s