apache/hudi

Incremental Cleaning should not revert to brute force all-partition scanning in any cases

Open

#14,569 创建于 2025年11月30日

在 GitHub 查看
 (0 评论) (0 反应) (0 负责人)Java (4,823 star) (2,431 fork)batch import
component:cleaningfrom-jiragood first issuehelp-wantedpriority:hightype:feature

描述

After [https://github.com/apache/incubator-hudi/pull/1576] . Incremental Cleaning would still resort to full partition scan when  no previous clean operation was done in the dataset. This ticket is to design and implement a safe solution which would avoid full scanning in all cases.

JIRA info

Comments

09/Jul/20 13:10;309637554;i am happy to take this issue.;;;

贡献者指南

技术栈
java
领域
backenddatabaseperformance
议题类型
feature
难度面向新贡献者的预计实现难度，1 表示很小改动，5 表示专家级工作。
4
预计时间有经验贡献者完成调查、实现、测试并准备 pull request 的粗略时间范围。
3-5 days
活动状态议题当前的可参与程度：新鲜、活跃、陈旧、阻塞或等待维护者输入。
fresh
清晰度议题是否清楚说明期望改动、验收标准和下一步。
mostly clear
前置要求
Understanding of Hudi's cleaning mechanismJava development experience
新手友好度1-100 的估计分数，表示该议题对首次贡献者的友好程度。
20
研究方向
Review the linked PR #1576 to understand the current incremental cleaning behavior. Investigate the Hudi codebase to identify where full partition scans occur when no previous clean operation exists. Propose a design that safely avoids full scanning in all cases, possibly by maintaining metadata about cleaned partitions. The JIRA ticket HUDI 854 may contain additional requirements.