Li Lu updated YARN-2709:
    Attachment: YARN-2709-102014.patch

I've done a patch for this issue. In this patch, I refactored the retry logic 
in jersey retry filter and built a more generalized retry wrapper for timeline 
client. Both the jersey retry filter and the delegation token call can use this 
wrapper to retry, according to the retry settings (added in YARN-2673). 

To use the retry wrapper, the user only needs to implement a 
TimelineClientRetryOp, providing a) the operation that should be retried and b) 
a verifier to tell, on given exception e, whether a retry should happen. 

I've also added a unit test for retried on getting delegation token. 

> Add retry for timeline client getDelegationToken method
> -------------------------------------------------------
>                 Key: YARN-2709
>                 URL: https://issues.apache.org/jira/browse/YARN-2709
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Li Lu
>            Assignee: Li Lu
>         Attachments: YARN-2709-102014.patch
> As mentioned in YARN-2673, we need to add retry mechanism to timeline client 
> for secured clusters. This means if the timeline server is not available, a 
> timeline client needs to retry to get a delegation token. 

This message was sent by Atlassian JIRA

Reply via email to