[ 
https://issues.apache.org/jira/browse/SENTRY-1345?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Sravya Tirukkovalur updated SENTRY-1345:
----------------------------------------
    Description: 
This seems to be due to the fact that onAlterTableEvent is being trigged for 
this operation of "INSERT on a unpartitioned table" in Hive. And sentry- hive 
plugin thinks this is an alter table location command, and we add the new path 
and delete the old path for this Hive object. As add happens first and then the 
delete on the same path in this case, the path is lost resulting in no acls on 
this path. 
Workaround: This gets corrected at the next HMS restart. 

We can fix on Sentry side to not do add/delete if oldPath == newPath. But we 
also need to understand the Hive behavior.
Not exactly sure when this regressed as we dont seem to have coverage for it. 
Also adding a test case as part of the patch which can be used against older 
versions.

> ACLS on table folder disappear after insert for unpartitioned tables
> --------------------------------------------------------------------
>
>                 Key: SENTRY-1345
>                 URL: https://issues.apache.org/jira/browse/SENTRY-1345
>             Project: Sentry
>          Issue Type: Bug
>    Affects Versions: 1.7.0
>            Reporter: Sravya Tirukkovalur
>            Assignee: Sravya Tirukkovalur
>             Fix For: 1.8.0
>
>         Attachments: SENTRY-1345.patch
>
>
> This seems to be due to the fact that onAlterTableEvent is being trigged for 
> this operation of "INSERT on a unpartitioned table" in Hive. And sentry- hive 
> plugin thinks this is an alter table location command, and we add the new 
> path and delete the old path for this Hive object. As add happens first and 
> then the delete on the same path in this case, the path is lost resulting in 
> no acls on this path. 
> Workaround: This gets corrected at the next HMS restart. 
> We can fix on Sentry side to not do add/delete if oldPath == newPath. But we 
> also need to understand the Hive behavior.
> Not exactly sure when this regressed as we dont seem to have coverage for it. 
> Also adding a test case as part of the patch which can be used against older 
> versions.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to