Unreleased Release Notes

6.4.0-25

New Features

• The Redis job board backend now allows the retry_on_timeout and health_check_interval redis-py client options to be passed through the job board configuration. These are opt-in and let deployments make the Redis client resilient to transient backend disruptions such as a Redis Sentinel master failover.

Upgrade Notes

• The taskflow.utils.async_utils module has been removed. This module has been unused in taskflow.

Deprecation Notes

• The retry_on_timeout option of the redis jobboard driver has been deprecated, because the same option in redis-py has been deprecated since redis-py 6.0.0 . Now retry on timeout is enabled by default.

Bug Fixes

• The executor conductor no longer terminates when a transient job board failure (for example a Redis Sentinel master failover) is raised while iterating or claiming jobs. It now logs the failure, backs off, and continues, reconnecting to the backend on the next cycle.

• Fixed log flooding caused by redis-py 8.0’s maintenance notification probes on Redis servers older than 8.x. Taskflow now explicitly disables these probes by passing MaintNotificationsConfig(enabled=False) to the Redis client, as Taskflow has its own reconnection and failover logic and does not consume maintenance notifications.

• The redis jobboard driver now raises taskflow.exceptions.IncompatibleVersion instead of taskflow.exceptions.JobFailure when redis backend is not compatible.