The Lettuce community recently fixed a bug where a client could take up to 15 minutes to recover when a Redis server shuts down abnormally without sending an RST (Reset) packet. For details, see Issue #2082. The Alibaba Cloud Tair team fixed this bug in PR #2499.
Bug fix and optimization
The Lettuce 6.3.0.RELEASE version has addressed the issue where the client does not automatically reconnect to the server after experiencing multiple request timeouts. For more information about the background of this issue, methods to reproduce the issue, and solutions, see Why does Lettuce result in longer downtimes?
Version suggestions
If Lettuce is included as a declared dependency, upgrade Lettuce to 6.3.0.RELEASE or later and enable the TCP_USER_TIMEOUT option. For the complete sample code, see Lettuce.
If Spring Data Redis is included as a declared dependency, upgrade the underlying Lettuce client to 6.3.0.RELEASE or later and enable the TCP_USER_TIMEOUT option. For the complete sample code, see Spring Data Redis.
If you are not a user of Alibaba Cloud Tair or Tair, we recommend that you upgrade the Lettuce client in your on-premises Redis environment to enhance the overall stability of your on-premises system.