[
https://issues.apache.org/jira/browse/HIVE-5589?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13805746#comment-13805746
]
Sergey Shelukhin commented on HIVE-5589:
----------------------------------------
Class name is pretty useless for aggregation, it's just smth like "Driver". I
am only replacing it for one statement.
The main problem with writing to database currently is that while doing what
this patch does would work for human reader of pseudo-xml, to write to database
you need to associate each perf logger statement with a particular workflow.
So you need some kind of shared ID for aggregation passed to every perfLogger
call to associate them with each other, or stored inside perfLogger in such
manner that different statements use different ones (for client, threadlocal
that is currently used might be sufficient to ensure only one query is running
at a time, so a pair of bounding calls can set/unset the query ID.
Let me file a JIRA for that...
> perflogger output is hard to associate with queries
> ---------------------------------------------------
>
> Key: HIVE-5589
> URL: https://issues.apache.org/jira/browse/HIVE-5589
> Project: Hive
> Issue Type: Improvement
> Reporter: Sergey Shelukhin
> Assignee: Sergey Shelukhin
> Priority: Minor
> Attachments: HIVE-5589.01.patch
>
>
> It would be nice to dump the query somewhere in output.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
