You are viewing a plain text version of this content. The canonical link for it is here.

Posted to notifications@iotdb.apache.org by "Liuxuxin (Jira)" <ji...@apache.org> on 2022/11/02 09:20:00 UTC

[jira] [Commented] (IOTDB-3633) [ compaction & raft snapshot ] When recovery , the merged files without snapshot are restored to level-0 tsfiles

    [ https://issues.apache.org/jira/browse/IOTDB-3633?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17627558#comment-17627558 ] 

Liuxuxin commented on IOTDB-3633:
---------------------------------

One possible way to fix this problem would be to trigger a snapshot after a compaction ends and notify the consensus that a new snapshot has been created. However, this allows the consensus layer to couple to the compaction module, and it requires additional logic to handle as a standalone server without the consensus layer.

> [ compaction & raft snapshot ] When recovery , the merged files without snapshot are restored to level-0 tsfiles
> ----------------------------------------------------------------------------------------------------------------
>
>                 Key: IOTDB-3633
>                 URL: https://issues.apache.org/jira/browse/IOTDB-3633
>             Project: Apache IoTDB
>          Issue Type: Improvement
>          Components: mpp-cluster
>    Affects Versions: 0.14.0-SNAPSHOT
>            Reporter: 刘珍
>            Assignee: Liuxuxin
>            Priority: Major
>         Attachments: image-2022-06-23-16-55-05-097.png, image-2022-06-23-16-56-05-675.png
>
>
> 1. 停止datanode（kill -9），没有做snapshot，有合并文件
>  !image-2022-06-23-16-55-05-097.png! 
> 2. 重启恢复，恢复后的合并文件消解成合并发生前的0层文件
>  !image-2022-06-23-16-56-05-675.png! 
> 期望，raft log恢复后，已经合并的tsfile，还是合并状态。



--
This message was sent by Atlassian Jira
(v8.20.10#820010)