[
https://issues.apache.org/jira/browse/TAJO-985?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14092353#comment-14092353
]
ASF GitHub Bot commented on TAJO-985:
-------------------------------------
Github user hyunsik commented on the pull request:
https://github.com/apache/tajo/pull/99#issuecomment-51736398
The main idea looks good to me. It will mitigate lock contention problem.
I have one suggestion. How about changing getState(boolean) and getState
signatures into different names? Since the uses of them may be sensitive, we
need to know what we invoke more apparently.
I would like to suggest getSynchronizedState() and getState(). It's just a
suggestion. The decision is up to you.
> Client API should be async
> --------------------------
>
> Key: TAJO-985
> URL: https://issues.apache.org/jira/browse/TAJO-985
> Project: Tajo
> Issue Type: Bug
> Components: query master
> Affects Versions: 0.9.0
> Reporter: Jinho Kim
> Assignee: Jinho Kim
>
> The query status synchronization on state machine always wait for time (event
> processing time) before status changes. If a lot of time are being processed
> to the event dispatcher, client api(getStatus, getProgress) will be waiting
> for changed event status. and query will be session timed out.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
