From f0a860908ba0309c89c9dba452d99b4f9c6928f7 Mon Sep 17 00:00:00 2001 From: Mathieu Velten Date: Thu, 3 Aug 2023 20:36:55 +0200 Subject: Allow config of the backoff algorithm for the federation client. (#15754) Adds three new configuration variables: * destination_min_retry_interval is identical to before (10mn). * destination_retry_multiplier is now 2 instead of 5, the maximum value will be reached slower. * destination_max_retry_interval is one day instead of (essentially) infinity. Capping this will cause destinations to continue to be retried sometimes instead of being lost forever. The previous value was 2 ^ 62 milliseconds. --- synapse/util/retryutils.py | 29 ++++++++++++++++------------- 1 file changed, 16 insertions(+), 13 deletions(-) (limited to 'synapse/util') diff --git a/synapse/util/retryutils.py b/synapse/util/retryutils.py index dcc037b982..27e9fc976c 100644 --- a/synapse/util/retryutils.py +++ b/synapse/util/retryutils.py @@ -27,15 +27,6 @@ if TYPE_CHECKING: logger = logging.getLogger(__name__) -# the initial backoff, after the first transaction fails -MIN_RETRY_INTERVAL = 10 * 60 * 1000 - -# how much we multiply the backoff by after each subsequent fail -RETRY_MULTIPLIER = 5 - -# a cap on the backoff. (Essentially none) -MAX_RETRY_INTERVAL = 2**62 - class NotRetryingDestination(Exception): def __init__(self, retry_last_ts: int, retry_interval: int, destination: str): @@ -169,6 +160,16 @@ class RetryDestinationLimiter: self.notifier = notifier self.replication_client = replication_client + self.destination_min_retry_interval_ms = ( + self.store.hs.config.federation.destination_min_retry_interval_ms + ) + self.destination_retry_multiplier = ( + self.store.hs.config.federation.destination_retry_multiplier + ) + self.destination_max_retry_interval_ms = ( + self.store.hs.config.federation.destination_max_retry_interval_ms + ) + def __enter__(self) -> None: pass @@ -220,13 +221,15 @@ class RetryDestinationLimiter: # We couldn't connect. if self.retry_interval: self.retry_interval = int( - self.retry_interval * RETRY_MULTIPLIER * random.uniform(0.8, 1.4) + self.retry_interval + * self.destination_retry_multiplier + * random.uniform(0.8, 1.4) ) - if self.retry_interval >= MAX_RETRY_INTERVAL: - self.retry_interval = MAX_RETRY_INTERVAL + if self.retry_interval >= self.destination_max_retry_interval_ms: + self.retry_interval = self.destination_max_retry_interval_ms else: - self.retry_interval = MIN_RETRY_INTERVAL + self.retry_interval = self.destination_min_retry_interval_ms logger.info( "Connection to %s was unsuccessful (%s(%s)); backoff now %i", -- cgit 1.4.1