[
https://issues.apache.org/jira/browse/TAJO-1196?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jihoon Son updated TAJO-1196:
-----------------------------
Description:
I reported this problem in our dev mailing list before
(http://mail-archives.apache.org/mod_mbox/tajo-dev/201410.mbox/%3CCACZfFK743PCEJkz0=rdt0ltuw-ajatyopgunsmragprhfc7...@mail.gmail.com%3E),
and finally found where the problem is caused.
The problem is, when a task runner requests a task via
QueryMasterProtocolService.Interface.getTask(), the query master task is not
started yet, so QueryMasterManagerService sends 'stopTaskRunnerReq' as a reply
to that request.
This asynchronous problem may be prominent in quite slow machines.
I'm thinking solutions, but don't still have any ones.
Do you have any good idea?
was:
I reported this problem in our dev mailing list before
(http://mail-archives.apache.org/mod_mbox/tajo-dev/201410.mbox/%3CCACZfFK743PCEJkz0=rdt0ltuw-ajatyopgunsmragprhfc7...@mail.gmail.com%3E),
and finally found where the problem is caused.
The problem is, when a task runner requests a task via
QueryMasterProtocolService.Interface.getTask(), the query master is not started
yet, so QueryMasterManagerService sends 'stopTaskRunnerReq' as a reply to that
request.
This asynchronous problem may be prominent in quite slow machines.
I'm thinking solutions, but don't still have any ones.
Do you have any good idea?
> Unit test hangs occasionally and randomly
> -----------------------------------------
>
> Key: TAJO-1196
> URL: https://issues.apache.org/jira/browse/TAJO-1196
> Project: Tajo
> Issue Type: Bug
> Components: test
> Reporter: Jihoon Son
>
> I reported this problem in our dev mailing list before
> (http://mail-archives.apache.org/mod_mbox/tajo-dev/201410.mbox/%3CCACZfFK743PCEJkz0=rdt0ltuw-ajatyopgunsmragprhfc7...@mail.gmail.com%3E),
> and finally found where the problem is caused.
> The problem is, when a task runner requests a task via
> QueryMasterProtocolService.Interface.getTask(), the query master task is not
> started yet, so QueryMasterManagerService sends 'stopTaskRunnerReq' as a
> reply to that request.
> This asynchronous problem may be prominent in quite slow machines.
> I'm thinking solutions, but don't still have any ones.
> Do you have any good idea?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
