[
https://issues.apache.org/jira/browse/TEZ-2512?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jeff Zhang updated TEZ-2512:
----------------------------
Description:
Currently we only support for PREEMPTED/DISKS_FAILED, that means except for
these 2 cases, there would be no clear diagnostic message for other cases.
Although we can do that for other ContainerExitStatus, but it would cause
incompability issue ( ContainerExitStatus of hadoop-2.2 support less exit code
than that of hadoop-2.6).
So I think maybe could use tez's own ContainerExitStatus since all the memebers
are integer, so that we could keep updated with yarn's support of
ContainerExitStatus and also won't break the compatibility.
But this would require shimming to correctly handle more and more failure
scenarios reported by YARN across different versions. Create this ticket here
as backlog if it is necessary in the future.
> Create Tez's ContainerExitStatus to handle different container fail scenarios
> -----------------------------------------------------------------------------
>
> Key: TEZ-2512
> URL: https://issues.apache.org/jira/browse/TEZ-2512
> Project: Apache Tez
> Issue Type: Improvement
> Reporter: Jeff Zhang
>
> Currently we only support for PREEMPTED/DISKS_FAILED, that means except for
> these 2 cases, there would be no clear diagnostic message for other cases.
> Although we can do that for other ContainerExitStatus, but it would cause
> incompability issue ( ContainerExitStatus of hadoop-2.2 support less exit
> code than that of hadoop-2.6).
> So I think maybe could use tez's own ContainerExitStatus since all the
> memebers are integer, so that we could keep updated with yarn's support of
> ContainerExitStatus and also won't break the compatibility.
> But this would require shimming to correctly handle more and more failure
> scenarios reported by YARN across different versions. Create this ticket here
> as backlog if it is necessary in the future.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
