This minor release provides improved logic for resilience to network problems, a number of improvements to edge cases of fault detector implementations, and includes an updated version of StackExchange.Redis binaries with better cluster node detection logic and improved handling of faulted connections.
-
Added – Allow using a LUA-based fault detector with the new
UseExperimentalFaultDetector
storage option. - Added – Periodic explicit checks for new jobs while in the subscription mode.
- Added – Periodic role checks for the current primary endpoint to detect silent fail-overs.
- Added – Log background jobs evicted from their queue by fault detector.
- Changed – Don’t create a fake state transition entry when fault detector re-queues a job.
- Changed – Use LUA script instead of conditional transactions when removing a server by heartbeat.
- Changed – Update StackExchange.Redis binaries to 1.2.7-build-00086 with better endpoints detection in a cluster mode.
- Fixed – Possible scheduled jobs getting stuck when a queue is set explicitly.
- Fixed – Remove fetched and expired jobs immediately, instead of re-queueing them.
- Fixed – Ensure we always ping the primary endpoint when waiting for events.
- Fixed – Increase the chances of immediate requeue on shutdown.
- Fixed – Decrease the chances of leaving a lock active on shutdown.
- Fixed – Clean up all flags related to different fault detectors in the LUA-based implementation.
- Fixed – Fault detector doesn’t process the same identifier multiple times.
-
Fixed – Don’t pass
AddressFamily.Unspecified
text to log messages to avoid confusion.