[
https://issues.apache.org/jira/browse/HIVE-12952?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Szehon Ho updated HIVE-12952:
-----------------------------
Attachment: HIVE-12952.patch
Attaching patch and Review board: [https://reviews.apache.org/r/43008/]
Implementation notes:
* The UI design choices are inspired from Impala, and HBase. This, like HBase
webui, uses Jamon, which is a superset of JSP and makes dynamic content a lot
easier. As such, brought in jamon dependency and also js bootstrap libraries
to support the dynamic tabs UI.
* On Hive side, refactored webui query logic into following classes:
SQLoperationDisplay (info captured from SQLOperation), QueryDisplay (info
captured from Driver).
TODO:
* Hard to get more MR job information for the stages including a job-tracking
url, due to MR JobSubmission being a separate process, need to think about it.
Same for Spark/tez.
* The explain plan might be a bit bulky and consume a bit of memory (though can
tune with "hive.server2.webui.max.historic.queries"). Perhaps in future we can
spill to local disk, and stream from there. This might also help with (1), if
we don't want to implement inter-process communciations.
> Show query sub-pages on webui
> -----------------------------
>
> Key: HIVE-12952
> URL: https://issues.apache.org/jira/browse/HIVE-12952
> Project: Hive
> Issue Type: Sub-task
> Components: Diagnosability
> Reporter: Szehon Ho
> Assignee: Szehon Ho
> Attachments: Error - Query Info Expired.png, HIVE-12952.patch, Query
> Drilldown link.png, Tab 3 - Stages.png, Tab 4 - Perf Logging.png, Tab1- Base
> Profile.png, Tab2 - Query Plan.png
>
>
> Today the queries showing in Running and Completed lists have some basic
> information like query string, elapsed time, state, user, etc.
> It may be helpful to have even more information like:
> * Job URL's, job status
> * Explain plan (configurable)
> * Error message (if failure)
> * Dynamic metrics, incl:
> ** Number of Tables/partitions fetched
> ** Time taken in each method via perf-logger.
> These should go in another page, so as not to clog the summary hiveserver.jsp
> page.
> This JIRA aims to tackle some of those.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
