[Umbrella] Rescale buckets for future partitions

未关闭
#4,288 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 2 天内回复

还没有人认领这个 Issue。

评估

难度
5/5
预计耗时
一周以上
新手友好度
20/100
Issue 类型
功能
描述清晰度
基本清楚
活跃度
停滞
技术栈
java

调研方向

先阅读 #3907 中链接的核心工作和已实现的 PR #3908,然后选择一个独立的子问题:#4280 用于聚合恢复测试,#4287 用于异步动态分区写入,或 #4289 用于功能文档。先阅读所选的子问题,并根据其中规定的恢复、路由或文档要求验证是否完成。

由索引模型根据 Issue 内容生成。

描述

Search before asking
  • I searched the existing issues and found no umbrella issue tracking this work.
Description

This umbrella issue tracks support for changing the bucket count of future partitions, together with the recovery coverage and client improvements needed around this feature.

Changing a partitioned table's bucket.num should affect only partitions created after the change. Existing partitions retain their original bucket counts, without data redistribution or lake-file rewrites. Reads, writes, and recovery must therefore work correctly when partitions of the same table have different bucket counts.

Sub-issues
  • #3907 — Core support for per-partition bucket rescaling, implemented in #3908: persist and propagate each partition's bucket count and use it for routing while allowing the table default to change for future partitions.
  • #4280 — Add aggregation recovery tests across partitions with different bucket counts. Cover both restoration from a real aggregation checkpoint and recovery through producer offsets without a checkpoint, including another failure during recovery. Exercise multiple operator subtasks and verify exact aggregate results with no duplicate application or data loss.
  • #4287 — Keep dynamic partition writes non-blocking with asynchronous partition creation and metadata refresh. Use fresh partition metadata when forming new batches and validate each queued batch's original routing at send time, failing incompatible batches rather than silently changing their routing.
  • #4289 — Add feature introduction and usage documentation, including runnable examples for increasing and decreasing future partitions' bucket counts, existing/pre-created partition semantics, and supported configurations and compatibility requirements.
Delivery scope

The core feature is tracked by #3907 and PR #3908. The recovery tests, asynchronous write-path improvement, and user documentation are tracked independently in the sub-issues above. In particular, #4287 is a follow-up to be addressed in a separate PR after #3908; it is not required to be implemented within #3908.

This umbrella remains open until all sub-issues are completed.

Are you willing to submit a PR?
  • I'm willing to submit a PR.
主要语言
Java
星标
2.2k
派生
668
平均合并
3 天 4 小时
30 天内合并 PR
146

环境准备

  • 没有 Dockerfile 或 Docker Compose 文件
  • 有 Pull Request 模板
  • 没有贡献指南

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

apache/fluss 的其他 Issue

查看 apache/fluss 的全部 Issue

相似的 Issue

更多 Java Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。