[
https://issues.apache.org/jira/browse/HADOOP-3696?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Arun C Murthy updated HADOOP-3696:
----------------------------------
Status: Open (was: Patch Available)
Overall, I agree with goals of this jira.
The fact that the TaskTracker (logically) depends on Hadoop Pipes (even after
the programmatic dependency is broken by Bill's patch) still bothers me.
I'd like to propose we add a new config parameter:
*mapred.task.debug.extra.opts* which are tacked to the end of the command line
for the debug script. Hadoop Pipes (or others like Hadoop Streaming) can then
use this seamlessly. Thoughts?
----
PS: I've opened HADOOP-3803 for further cleanup of handling the way task
debugging is done...
> Dependency cycle: Submitter and TaskTracker
> -------------------------------------------
>
> Key: HADOOP-3696
> URL: https://issues.apache.org/jira/browse/HADOOP-3696
> Project: Hadoop Core
> Issue Type: Improvement
> Components: mapred, pipes
> Affects Versions: 0.17.0
> Reporter: Bill de hOra
> Priority: Minor
> Attachments: HADOOP-3696.patch
>
>
> there's a cycle between org.apache.hadoop.mapred.pipes
> org.apache.hadoop.mapred; breaking the dependency between Submitter and
> TaskTracker fixes it.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
