[
https://issues.apache.org/jira/browse/HIVE-4632?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13822200#comment-13822200
]
Ashutosh Chauhan commented on HIVE-4632:
----------------------------------------
Patch looks good. I think we should also change default (in HiveConf.java and
also in data/conf/hive-site.xml for tests) to use counter mechanism, instead of
current jdbc one. With the current default configs, stats cannot be collected,
so its useless anyway. Even after you configure it with custom settings, I have
found client takes forever to aggregate and publish stats once job ends, so
using counter will give us something which works out of box as well as
performing better.
> Use hadoop counter as a stat publisher
> --------------------------------------
>
> Key: HIVE-4632
> URL: https://issues.apache.org/jira/browse/HIVE-4632
> Project: Hive
> Issue Type: Improvement
> Components: Statistics
> Affects Versions: 0.12.0
> Reporter: Navis
> Assignee: Navis
> Priority: Minor
> Attachments: D11001.2.patch, HIVE-4632.D11001.1.patch
>
>
> Currently stats are all long/aggregation type and can be safely acquired by
> hadoop counter without other db or hbase.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
