[jira] [Commented] (AURORA-187) scheduler failover should never abort an update

["Maxim Khutornenko (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/AURORA-187?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14144112#comment-14144112 ] 

Maxim Khutornenko commented on AURORA-187:
------------------------------------------

This problem is wider than just the client updater and should be solvable via a dedicated error type exposed by the scheduler for something that can be retried. 

A possible solution here could be adding a new flag into StorageException (e.g. IsRetryable) that would be set in cases like this one. The LoggingInterceptor could check that flag and convert into a custom error type (e.g. ERROR_RETRYABLE) to help client dissect idempotent/retryable errors from all others. 

> scheduler failover should never abort an update
> -----------------------------------------------
>
>                 Key: AURORA-187
>                 URL: https://issues.apache.org/jira/browse/AURORA-187
>             Project: Aurora
>          Issue Type: Bug
>          Components: Client, Scheduler
>            Reporter: brian wickman
>            Priority: Critical
>
> User reported they had an aborted update
> {noformat}
>  INFO] Response from scheduler: ERROR (message: Aborting update without rollback! Fatal error: Storage is not READY)
> {noformat}
> This seems like a regression.  Users should never get aborted updates due to scheduler failovers.



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