[ 
https://issues.apache.org/jira/browse/HIVE-13633?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Sergey Shelukhin updated HIVE-13633:
------------------------------------
    Description: 
>From the API discussion. We want to make sure that fields coming from external 
>clients that are used by LLAP as various kinds of IDs are handled correctly 
>and misbehaving clients do not cause bizarre issues. Some of the fields may 
>not be practically signable (e.g. attempt number, even we implemented some 
>signing scheme it would just be ugly to re-sign to retry), also someone could 
>just submit the same signed split twice. 
The fields are - fragment number (and various other parts of ID, presumably), 
attempt number, container id, maybe others.

  was:
>From the API discussion. We want to make sure that fields coming from external 
>clients that are used by LLAP as various kinds of IDs are handled correctly. 
>Some of the fields may not be practically signable (e.g. attempt number, even 
>we implemented some signing scheme it would just be ugly to re-sign to retry), 
>also someone could just submit the same signed split twice. 
The fields are - fragment number (and various other parts of ID, presumably), 
attempt number, container id, maybe others.


> LLAP: better handling of container id, attempt number, fragment number for 
> external clients
> -------------------------------------------------------------------------------------------
>
>                 Key: HIVE-13633
>                 URL: https://issues.apache.org/jira/browse/HIVE-13633
>             Project: Hive
>          Issue Type: Bug
>            Reporter: Sergey Shelukhin
>            Assignee: Siddharth Seth
>
> From the API discussion. We want to make sure that fields coming from 
> external clients that are used by LLAP as various kinds of IDs are handled 
> correctly and misbehaving clients do not cause bizarre issues. Some of the 
> fields may not be practically signable (e.g. attempt number, even we 
> implemented some signing scheme it would just be ugly to re-sign to retry), 
> also someone could just submit the same signed split twice. 
> The fields are - fragment number (and various other parts of ID, presumably), 
> attempt number, container id, maybe others.



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

Reply via email to