[
https://issues.apache.org/jira/browse/HIVE-17078?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yibing Shi updated HIVE-17078:
------------------------------
Attachment: HIVE-17078.5.PATCH
Thank you for the review, [~stakiar_impala_496e]!
I have modified the patch to adopt in point #1 and #3.
As for #2:
bq. Where does l4j print to?
It depends. By default the local task is run in a new process, and the log4j
appenders are not setup in the child process. As such, this l4j doesn't print
to anywhere.
But if {{hive.exec.submit.local.task.via.child}} is disabled, the Hive l4j
variable is used, and thus the information will be printed to Hive log.
> Add more logs to MapredLocalTask
> --------------------------------
>
> Key: HIVE-17078
> URL: https://issues.apache.org/jira/browse/HIVE-17078
> Project: Hive
> Issue Type: Improvement
> Reporter: Yibing Shi
> Assignee: Yibing Shi
> Priority: Minor
> Attachments: HIVE-17078.1.patch, HIVE-17078.2.patch,
> HIVE-17078.3.patch, HIVE-17078.4.PATCH, HIVE-17078.5.PATCH
>
>
> By default, {{MapredLocalTask}} is executed in a child process of Hive, in
> case the local task uses too much resources that may affect Hive. Currently,
> the stdout and stderr information of the child process is printed in Hive's
> stdout/stderr log, which doesn't have a timestamp information, and is
> separated from Hive service logs. This makes it hard to troubleshoot problems
> in MapredLocalTasks.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
