📝 5 新 Issues
✅ 2 关闭
✨ 7 新 PRs
🎉 2 合并
🔥 核心要点
- 🎉 PR #2839 新增 InfluxDB 指标上报器支持，完善监控生态
- 🎉 PR #3317 Spark lake-batch 回退优化，改用 kv snapshot 提升读取效率
- 🐛 PR #3322 修复 Flink/Spark 非字符串分区键的分区裁剪 ClassCastException 问题
- ✨ PR #3316 引入 HudiBucketingFunction 支持 Hudi bucket 策略
- ✨ Issue #3314 新特性提案：Fluss Tiering 支持 Partition Mark Done
💬 Issue/PR 动态
Issue 讨论
Issue #3304 [flink] currentFetchEventTimeLag is misleading when subscribing multiple partitions/buckets with uneven lag
- @zuston: 担心方案1对于无数据流入的过期分区会导致永久高延迟警告，可能误导用户
Issue #3222 [server] Respect the option of `table.log.ttl` for local segments
- @zuston: 友善提醒 luoyuxia 和 swuferhong 进行代码审查
Issue #3260 [spark] Add scan.maxRecordsPerPartition config to split log table input partitions
- @Yohahaha: 询问其他审查人员是否有更多意见
- @fresh-borzoni: 提醒需要更新测试中的配置键以修复测试用例
Issue #3316 [lake/hudi] Introduce HudiBucketingFunction for bucket strategy
- @fhan688: 请求 XuQianJin-Stars 帮忙审查代码
Issue #3317 [spark] Use kv snapshot in lake-batch fallback
- @fresh-borzoni: 请求 Yohahaha、YannByron、luoyuxia 审查，并已处理审查意见
Issue #3322 [flink][spark] Fix partition pruning for non-string partition keys
- @fresh-borzoni: 请求审查并已处理审查意见
PR Review
PR #2839 [metric] Support influxdb reporter
- @zhuyufeng0809: 确认 InfluxDB Tag Value 规范无需过滤，已完成修改
PR #3260 [spark] Add scan.maxRecordsPerPartition config to split log table input partitions
- @YannByron: 建议使用 scan.maxRecordsPerPartition 命名更符合 Spark 风格
- @fresh-borzoni: 质疑值为 0 时是否会触发 RPC 调用，建议要求值大于 0
- @Yohahaha: 发现遗漏修改处并表示感谢提醒
PR #3296 [flink] Fix batch fallback generating mixed split types for primary-key tables
- @matrixsparse: 已修复使用 stoppingOffsetsInitializer.getBucketOffsets() 传递停止偏移量
PR #3316 [lake/hudi] Introduce HudiBucketingFunction for bucket strategy
- @Copilot: 指出 HudiKeyEncoder 返回 4 字节哈希会导致主键碰撞问题，建议特殊处理
- @fhan688: 感谢审查意见
PR #3317 [spark] Use kv snapshot in lake-batch fallback
- @Yohahaha: 质疑 fallback 命名是否准确，snapshotId>=0 意为避免读取全量历史日志
- @fresh-borzoni: 已处理审查意见，请求 luoyuxia 快速审查
PR #3322 [flink][spark] Fix partition pruning for non-string partition keys
- @Yohahaha: 建议移除 toInternalRow 直接使用 PartitionUtils.toPartitionRow
- @luoyuxia: 解释 DATE 类型在分区裁剪时会导致 LocalDate 与 Integer 比较抛出 ClassCastException
- @Copilot: 建议缓存 partitionRowType、增加 TIMESTAMP_LTZ 测试覆盖、优雅处理解析失败
- @fresh-borzoni: 感谢发现有趣的问题
📝 新建 Issue/PR
Issues
- #3321 [spark] Partition filter pushdown for lake-enabled tables @fresh-borzoni
- #3320 [spark] Port TreeMap-collapse log buffer to FlussUpsertPartitionReader @fresh-borzoni
- #3318 [spark] Use Fluss kv snapshot in lake-batch fallback @fresh-borzoni
- #3315 Respect the log.* config from the table level options @zuston
- #3314 [Feature] Support Partition Mark Done for Fluss Tiering (Paimon Lake) @beryllw
Pull Requests
- #3326 [INFRA] Set up default rulesets for default and release branches @asf-gitbox-commits
- #3325 [kv] Treat full-schema partial writes as full row for DefaultRowMerger @zuston
- #3324 [lake] Order lake snapshot metadata by commit_timestamp @Kaixuan-Duan
- #3323 Bump com.github.tomakehurst:wiremock-jre8 from 2.32.0 to 2.35.1 in /fluss-metrics/fluss-metrics-influxdb @app/dependabot dependencies, java
- #3322 [flink][spark] Fix partition pruning for non-string partition keys @fresh-borzoni
- #3319 [FIP-37] Add bitmap infrastructure: BitmapUtils, RoaringBitmapSerializer, AbstractRbAggFunction @Prajwal-banakar
- #3316 [lake/hudi] Introduce HudiBucketingFunction for bucket strategy @fhan688
✅ 关闭 Issue/PR
已关闭 Issues
- #3318 [spark] Use Fluss kv snapshot in lake-batch fallback @fresh-borzoni
- #2838 [metric] Support influxdb reporter @zhuyufeng0809
已合并 PRs
- #3317 [spark] Use kv snapshot in lake-batch fallback @fresh-borzoni
- #2839 [metric] Support influxdb reporter @zhuyufeng0809