📝 1 新 Issues
✅ 1 关闭
✨ 1 新 PRs
🎉 1 合并
🔥 核心要点
- 🎉 PR #2635 文档更新：添加 Lakekeeper catalog 引用参考
- ✨ Issue #3235 新增 System Tables 的 Umbrella issue
- 🐛 PR #3208 修复启用 datalake 的空表批量查询失败问题
- 🐛 PR #3227 修复 Spark 批量读取 COUNT(*)/COUNT(1) 空投影导致的元数据错误
- 🚀 PR #3236 新增 Flink 支持带启动时间戳过滤的 lake 联合读取
⚠️ 重点关注
- ⚠️ Issue #3188 测试不稳定：CoordinatorEventManagerTest.testMetricsUpdatedImmediatelyOnStartup 出现失败
💬 Issue/PR 动态
Issue 讨论
Issue #2371 Tiering may stuck for first row merge engine table
- @luoyuxia: 提醒 zuston 查看此 issue
- @zuston: 感谢帮助，请继续处理此 issue
Issue #3188 [test] Unstable test CoordinatorEventManagerTest.testMetricsUpdatedImmediatelyOnStartup
- @github-actions[bot]: 报告测试失败：AssertionFailedError 异常
Issue #3043 [client] release pending fetch semaphore on discarding
- @fresh-borzoni: 请求 rebase 并关联 tracking issue，询问是否可以合并
Issue #3233 [server] Support altering table.log.ttl table option
- @Kaixuan-Duan: 已根据反馈修改，请 luoyuxia 审查
Issue #3236 [flink] union read from lake with startup timestamp filter
- @binary-signal: 讨论主键表使用起始偏移时间戳的可行性方案
- @zuston: 认同方案，计划后续支持该功能
Issue #3230 Streaming union read from the lake and Fluss log table using a specified offset timestamp
- @matrixsparse: 表示愿意参与此 issue 的开发工作
- @zuston: 感谢关注，正在 #3236 PR 上进行开发
Issue #3210 [Test] Fixing flaky test CoordinatorEventManagerTest.testMetricsUpdatedImmediatelyOnStartup
- @vamossagar12: 提醒 swuferhong 审查此 PR
Issue #3227 [spark] Fix empty projection causing Invalid metadata length for COUNT(*)/COUNT(1)
- @luoyuxia: 指出此 PR 可能与 #2725 重复
- @Kaixuan-Duan: 道歉表示未注意到已有 PR 指向相同 issue
Issue #2881 [lake/iceberg] Support snapshot expiration for lake tables
- @vamossagar12: 提醒 leekeiabstraction 审查此 PR
Issue #2876 Fixing issue with row update not working for PK table:
- @vamossagar12: 提醒 loserwang1024 和 wuchong 审查此 PR
PR Review
PR #3208 [flink] Fix batch query on empty datalake-enabled table to return 0 rows instead of failing
- @luoyuxia: 质疑返回空列表的逻辑，建议参考 Spark 的回退方案
- @matrixsparse: 已修改为回退到 Fluss-only split 生成，并添加分区验证
- @fresh-borzoni: 建议添加日志以便追踪，参考 stream-mode 的做法
PR #3227 [spark] Fix empty projection causing Invalid metadata length for COUNT(*)/COUNT(1)
- @Yohahaha: 建议验证 Flink connector 中的行为并添加修复
- @Kaixuan-Duan: 确认 Flink connector 未受影响，服务端修复是通用的
PR #3200 [server] Support configurable time partition format for auto-partitioned tables
- @Copilot: 指出 time-format 验证不足，可能产生无效分区值或碰撞
- @Copilot: 建议文档补充分区值字符约束，避免用户使用无效格式
PR #2725 [spark] Fix empty projection issue in Spark batch read to avoid Arrow metadata errors with COUNT(*)/COUNT(1)
- @Copilot: 指出方法注释需要更新以匹配实际支持的 COUNT 下推场景
- @Copilot: 建议添加 COUNT(1) 测试用例确保覆盖
- @Copilot: 警告测试数据修改可能影响其他测试的断言
PR #3233 [server] Support altering table.log.ttl table option
- @Copilot: 指出 @VisibleForTesting 注解与生产代码使用不一致
- @Copilot: 建议使用非测试 API 替代异常驱动的控制流
- @Copilot: 建议降低日志级别避免噪音，并改进消息说明
- @Copilot: 建议从配置项获取默认 TTL 避免硬编码
📝 新建 Issue/PR
Issues
- #3235 [Umbrella] Umbrella issue for System Tables @gyang94
Pull Requests
- #3236 [flink] union read from lake with startup timestamp filter @zuston
✅ 关闭 Issue/PR
已关闭 Issues
- #2611 [Lake/Iceberg] create doc to integrate fluss + Iceberg via flink with Lakekeeper (IRC) @MehulBatra
已合并 PRs
- #2635 [docs] Add Lakekeeper catalog reference @fresh-borzoni