[
https://issues.apache.org/jira/browse/KYLIN-968?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yerui Sun updated KYLIN-968:
----------------------------
Attachment: KYLIN-968-1.x-staging-v2.patch
Patch v2 for 1.x branch. Removed the unit test which didn't make sense.
BTW, I checked the 2.x branch code for this issue, and found that it has been
fixed. So I will not post a patch for 2.x branch anymore.
> CubeSegment.lastBuildJobID is null in new instance but used for rowkey_stats
> path
> ---------------------------------------------------------------------------------
>
> Key: KYLIN-968
> URL: https://issues.apache.org/jira/browse/KYLIN-968
> Project: Kylin
> Issue Type: Bug
> Components: Job Engine
> Affects Versions: v1.0, v0.7.2
> Reporter: Yerui Sun
> Assignee: Shaofeng SHI
> Fix For: v1.1
>
> Attachments: KYLIN-968-0.7-staging.patch,
> KYLIN-968-1.x-staging-v2.patch, KYLIN-968-1.x-staging.patch
>
>
> CubeSegment.lastBuildJobID is used to record the id of job which building
> this segment last time.
> When build or fresh a cube and append a new segment, the lastBuildingJobID
> remains as null until the segment build done and info updated. But in the
> segment building flow, lastBuildingJobID has been used for concat
> rowkey_stats path, in CubingJobBuilder.getRowkeyDistributionOutputPath,
> result in rowkey_stats path is like '/../kylin-null/../rowkey_stats', without
> job id in path.
> If more than one segments were building concurrently, the rowkey_stats path
> will be conflicts and result in wrong split keys for tables in HBase.
> Did anybody found this issue before? If not yet and we agreed with the
> analysis above, I'll apply a patch later.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
