[jira] [Commented] (IOTDB-1693) IoTDB restart does not truncate broken ChunkGroup bug

["Liuxuxin (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/IOTDB-1693?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17416002#comment-17416002 ] 

Liuxuxin commented on IOTDB-1693:
---------------------------------

https://github.com/apache/iotdb/pull/3947

> IoTDB restart does not truncate broken ChunkGroup bug
> -----------------------------------------------------
>
>                 Key: IOTDB-1693
>                 URL: https://issues.apache.org/jira/browse/IOTDB-1693
>             Project: Apache IoTDB
>          Issue Type: Bug
>            Reporter: Liuxuxin
>            Priority: Major
>             Fix For: 0.12.2-SNAPSHOT
>
>         Attachments: image-2021-09-16-07-59-48-616.png
>
>
> h1. Bug Description
> When restoring a level 0 tsfile whose writing was interrupted, the wrong use of RestorableTsFileIOWriter will cause the file to have redundant ChunkGroup after Wal is written. These ChunkGroups will not be indexed by Metadata, so they will not cause query errors, but will cause redundant data in TsFile.
>  
> h1. Reproduce
> !image-2021-09-16-07-59-48-616.png!
> Add a System.exit(1) in MemTableFlushTask.ioTask, then run the IoTDB and insert some data. When flushing the memtable, the system will exit, and the writing tsfile is broken. Remove the System.exit(1) and restart the IoTDB, the tsfile will be restore during the recover process. Use TsFileSequenceRead to check the recovered tsfile, we will find out a broken ChunkGroup.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)