[
https://issues.apache.org/jira/browse/HIVE-12558?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15178499#comment-15178499
]
Sergey Shelukhin commented on HIVE-12558:
-----------------------------------------
Test failures (llap and operation logging) look related. Spark timeouts are a
known condition.
> LLAP: output QueryFragmentCounters somewhere
> --------------------------------------------
>
> Key: HIVE-12558
> URL: https://issues.apache.org/jira/browse/HIVE-12558
> Project: Hive
> Issue Type: Bug
> Components: llap
> Reporter: Sergey Shelukhin
> Assignee: Prasanth Jayachandran
> Attachments: HIVE-12558.1.patch, HIVE-12558.2.patch,
> HIVE-12558.wip.patch, sample-output.png
>
>
> Right now, LLAP logs counters for every fragment; most of them are IO related
> and could be very useful, they also include table names so that things like
> cache hit ratio, etc., could be calculated for every table.
> We need to output them to some metrics system (preserving the breakdown by
> table, possibly also adding query ID or even stage) so that they'd be usable
> without grep/sed/awk.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
