You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@hudi.apache.org by "sivabalan narayanan (Jira)" <ji...@apache.org> on 2021/11/24 02:27:00 UTC

[jira] [Updated] (HUDI-2842) AbstractLogRecordReader does not handle rollback with multi writers

     [ https://issues.apache.org/jira/browse/HUDI-2842?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

sivabalan narayanan updated HUDI-2842:
--------------------------------------
    Fix Version/s: 0.10.0

> AbstractLogRecordReader does not handle rollback with multi writers
> -------------------------------------------------------------------
>
>                 Key: HUDI-2842
>                 URL: https://issues.apache.org/jira/browse/HUDI-2842
>             Project: Apache Hudi
>          Issue Type: Bug
>    Affects Versions: 0.10.0
>            Reporter: sivabalan narayanan
>            Assignee: sivabalan narayanan
>            Priority: Blocker
>             Fix For: 0.10.0
>
>
> The way we handle rollback blocks in AbstractLogRecordReader is below:
> Once we encounter the rollback block, we walk back to previous log blocks to get hold of the blocks that needs to be rolledback. Whenever the instant time of a block differs from what the rollback block refers to, we wrap up the rollback block handling. This may not hold good with multi-writer scenario.  
>  
> DC1: logBlock1
> DC2: logBlock2 // failed mid way.
> DC3: logBlock3
> Rollback command block: rollback of DC2. 
>  
> When handling the rollback command block, with DC3 we bail since the instant time (DC3) does not match DC2 i.e. the commit being rolledback by DC4. 
>  



--
This message was sent by Atlassian Jira
(v8.20.1#820001)