You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "Tao Li (JIRA)" <ji...@apache.org> on 2016/09/14 18:20:22 UTC
[jira] [Commented] (HIVE-14752) New metrics should be introduced to
more easily monitor the workload of HS2/HMS
[ https://issues.apache.org/jira/browse/HIVE-14752?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15491113#comment-15491113 ]
Tao Li commented on HIVE-14752:
-------------------------------
That sounds like a good idea. I think you are basically talking about adding service logs for HS2 and HMS.
I suggest we add a log entry for each API call against HS2/HMS, which can include information such as API operation name, user name/ID, start/end time, success/failure etc. We can run some scripts against the logs to get the metrics you mentioned, like the failure rate or throughput (request/sec).
We can also think about pushing the service metrics to a NoSQL db (e.g. HBase) to speed up queries.
> New metrics should be introduced to more easily monitor the workload of HS2/HMS
> -------------------------------------------------------------------------------
>
> Key: HIVE-14752
> URL: https://issues.apache.org/jira/browse/HIVE-14752
> Project: Hive
> Issue Type: Task
> Components: Hive, HiveServer2, Metastore
> Reporter: Barna Zsombor Klara
> Assignee: Barna Zsombor Klara
>
> Currently we have no way to easily track the number of sessions/queries in a HS2 instance.
> It would be great if we could tell the number of queries being run, which lifecycle phase they are in (submitted, compilation, execution), and how much time was spent executing/compiling them.
> This Jira will be an umbrella to track the subtasks for each potentially useful metric we can think of.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)