[jira] [Commented] (SENTRY-939) Provide a configuration to be able to fall back to orginal HDFS user:group when NN has stale sentry permissions

["Sravya Tirukkovalur (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SENTRY-939?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14983742#comment-14983742 ] 

Sravya Tirukkovalur commented on SENTRY-939:
--------------------------------------------

Actually this is possible today by using the property "sentry.authorization-provider.include-hdfs-authz-as-acl" which would append original user:group:permissions as an extended acl, both in stale and non stale states.

> Provide a configuration to be able to fall back to orginal HDFS user:group when NN has stale sentry permissions
> ---------------------------------------------------------------------------------------------------------------
>
>                 Key: SENTRY-939
>                 URL: https://issues.apache.org/jira/browse/SENTRY-939
>             Project: Sentry
>          Issue Type: Bug
>    Affects Versions: 1.4.0
>            Reporter: Sravya Tirukkovalur
>            Assignee: Sravya Tirukkovalur
>             Fix For: 1.7.0
>
>
> Seems like some sentry users have a batch id for each application which creates the directory, ingests the data, then creates a table/partition over it. So the batch id is the the original owner of this directory.
> When NN has stale permissions during the time it cannot talk to Sentry (if sentry is down), falling back to hive:hive blocks everyone it would be helpful if at least this batch id has access otherwise ingest pipelines may break during this time. So would be helpful to provide a way to configure this behavior.
> - This does not entail ACLS.
> - We should add hive:hive as acl to make sure hive has access 



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