[
https://issues.apache.org/jira/browse/HIVE-5510?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13796325#comment-13796325
]
Eugene Koifman commented on HIVE-5510:
--------------------------------------
Comments on HIVE-5510-2.patch
build.xml changes concurrency from 5/5 to 1/1. Why?
TempletonControllerJob.Watcher.run()
This records all child IDs in the parent as a list. Makes sense.
This also creates a JobState for each child task. What is purpose of this?
We never (AFAIK) get any notifications from Hadoop when a subtask completes.
When is this JobState (for child task) ever updated? If not, then why create
it? Conversely, maybe it would be useful to track the state of child task, but
I think that would require more changes like registering a callback with
JobTracker for each child task that we find. Don't know whether this is really
useful.
DeleteDelegator
it uses System.err for logging. Why not log4j which will be in webhcat.log
with WARN log level
JobState:
getChildId() is never used
> [WebHCat] GET job/queue return wrong job information
> ----------------------------------------------------
>
> Key: HIVE-5510
> URL: https://issues.apache.org/jira/browse/HIVE-5510
> Project: Hive
> Issue Type: Bug
> Components: WebHCat
> Affects Versions: 0.12.0
> Reporter: Daniel Dai
> Assignee: Daniel Dai
> Fix For: 0.13.0
>
> Attachments: HIVE-5510-1.patch, HIVE-5510-2.patch,
> test_harnesss_1381798977
>
>
> GET job/queue of a TempletonController job return weird information. It is a
> mix of child job and itself. It should only pull the information of the
> controller job itself.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
