[
https://issues.apache.org/jira/browse/HIVE-713?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12740207#action_12740207
]
Edward Capriolo commented on HIVE-713:
--------------------------------------
True. If we were able to implement JMX we could attach to a mapper/reduce and
profile it directly. IE real time how much memory a mapper is using, there are
a bunch of standard jmx objects that could be useful. Possibly a logging based
approach might be a better way to go. Another technique might be a job
completion URL. counters+job completion would be a nice combination.
JMX could be a tricky implementation due to the multi-jvm nature of the hadoop.
But if the framework is managing counters for us getting them into JMX should
be doable. I will review your code more.
> Integrate operator counters into query plan to view progress of queries
> -----------------------------------------------------------------------
>
> Key: HIVE-713
> URL: https://issues.apache.org/jira/browse/HIVE-713
> Project: Hadoop Hive
> Issue Type: New Feature
> Components: Logging
> Reporter: Raghotham Murthy
> Assignee: Raghotham Murthy
> Attachments: hive-713.1.donot.patch, hive-713.2.donot.patch,
> hive-713.2.patch, hive-713.3.patch, hive-713.4.donot.patch,
> hive-713.5.donot.patch, hive.713.1.good.patch, hive.713.patch.bad
>
>
> Annotate the QueryPlan - with multiple stages and each stage(Task) can have
> an Operator tree - with counters that each Operator/Task updates.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.