Exit walsender before confirming remote flush in logical replication

[grantzhou]

--- As of 2026-01-26T14:08:39.000Z ---

Fujii Masao provides detailed feedback on the walsender shutdown patch. He suggests using LOG_SERVER_ONLY instead of WARNING for ereport to avoid blocking on full output buffers, but notes that proc_exit() could still emit messages at other levels. He proposes resetting whereToSendOutput like in WalSndShutdown(). Masao questions a comment about handling serial execution of parallel-planned scans, noting it may be outdated. He observes inconsistent statistics tracking between parallel and non-parallel seq scans. Alternative approaches are discussed, including having postmaster send SIGTERM to walsenders for immediate shutdown. Documentation improvements are suggested for clarity, parameter ordering, and config examples. Several code improvements are recommended: adding WalSndShutdownMode to typedefs.list, updating comments about shutdown behavior, and removing unnecessary abort() call. The feedback covers correctness concerns, consistency issues, and code maintenance aspects of the proposed changes.
藤井雅夫对 walsender 关闭补丁提供了详细反馈。他建议在 ereport 中使用 LOG_SERVER_ONLY 而不是 WARNING 来避免输出缓冲区满时的阻塞，但指出 proc_exit() 仍可能发出其他级别的消息。他提议像在 WalSndShutdown() 中那样重置 whereToSendOutput。雅夫质疑关于处理并行规划扫描的串行执行的注释，认为可能已过时。他观察到并行和非并行顺序扫描之间的统计跟踪不一致。讨论了替代方法，包括让 postmaster 向 walsenders 发送 SIGTERM 来实现立即关闭。建议改进文档的清晰度、参数排序和配置示例。推荐了几项代码改进：将 WalSndShutdownMode 添加到 typedefs.list、更新关于关闭行为的注释，以及移除不必要的 abort() 调用。反馈涵盖了所提议更改的正确性问题、一致性问题和代码维护方面。

participants - 参与者:

• a.silitskiy@postgrespro.ru
• aekorotkov@gmail.com
• amit.kapila16@gmail.com
• andres@anarazel.de
• dilipbalaut@gmail.com
• horikyota.ntt@gmail.com
• htamfids@gmail.com
• kuroda.hayato@fujitsu.com
• masao.fujii@gmail.com
• michael@paquier.xyz
• osumi.takamichi@fujitsu.com
• peter.eisentraut@enterprisedb.com
• sawada.mshk@gmail.com
• smithpb2250@gmail.com
• v.davydov@postgrespro.ru