You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@kylin.apache.org by "Shaofeng SHI (JIRA)" <ji...@apache.org> on 2019/05/30 08:54:00 UTC
[jira] [Commented] (KYLIN-3994) StorageCleanupJob may delete data
of newly built segment because of cube cache in CubeManager
[ https://issues.apache.org/jira/browse/KYLIN-3994?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16851656#comment-16851656 ]
Shaofeng SHI commented on KYLIN-3994:
-------------------------------------
I agree with Zhengshuai; The current PR only narrow down the possibility of reading a dirty data, not solve the problem totally.
> StorageCleanupJob may delete data of newly built segment because of cube cache in CubeManager
> ---------------------------------------------------------------------------------------------
>
> Key: KYLIN-3994
> URL: https://issues.apache.org/jira/browse/KYLIN-3994
> Project: Kylin
> Issue Type: Bug
> Affects Versions: v2.5.2
> Reporter: Liu Shaohui
> Assignee: Liu Shaohui
> Priority: Major
> Fix For: v2.6.3
>
>
> In our production cluster, we found that the cube id data of a new-built segment is deleted by the StorageCleanupJob.
> After checking the code of cleanUnusedHdfsFiles in StorageCleanupJob, we found that there is a bug here: CubeManager read all cube meta in initiation and cache it for later
> listAllCubes operations, the metadata will be out of data after listing the hdfs working dir.
> So the working directory of a finished job may be deleted unexpectedly.
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)