You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@hudi.apache.org by "Yuwei Xiao (Jira)" <ji...@apache.org> on 2022/03/23 11:25:00 UTC
[jira] [Assigned] (HUDI-3692) Write cannot see inflight compaction when using metadata table
[ https://issues.apache.org/jira/browse/HUDI-3692?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Yuwei Xiao reassigned HUDI-3692:
--------------------------------
Assignee: Yuwei Xiao
> Write cannot see inflight compaction when using metadata table
> --------------------------------------------------------------
>
> Key: HUDI-3692
> URL: https://issues.apache.org/jira/browse/HUDI-3692
> Project: Apache Hudi
> Issue Type: Bug
> Components: metadata, writer-core
> Reporter: Yuwei Xiao
> Assignee: Yuwei Xiao
> Priority: Major
> Attachments: image-2022-03-23-17-52-38-672.png
>
>
> The construction of metadata table accepts a timeline without compaction. So without further `sync()` called on the view, async in-flight compactions is not recognized by writers, leading to lost of updates.
> !image-2022-03-23-17-52-38-672.png!
> To reproduce:
> 1. insert 10 records to create file group A
> 2. update 5 records in file group A, produce log file 1 (make sure it is producing log file rather than re-write file group)
> 3. schedule compaction (do not execute)
> 4. delete 5 records in file group A, produce log file 2
> 5. insert the deleted records in file group A, produce log file 3
> 6. Read table, and the total number is 5 rather than the 10 records.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)