[jira] (HIVE-26978) Stale "Runtime stats" causes poor query planning

["Zoltan Haindrich (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HIVE-26978 ]


    Zoltan Haindrich deleted comment on HIVE-26978:
    -----------------------------------------

was (Author: kgyrtkirk):
have you restarted the HS2?
the runtime stats are cached there; the metastore only stores them

> Stale "Runtime stats" causes poor query planning
> ------------------------------------------------
>
>                 Key: HIVE-26978
>                 URL: https://issues.apache.org/jira/browse/HIVE-26978
>             Project: Hive
>          Issue Type: Improvement
>          Components: HiveServer2
>            Reporter: Rajesh Balamohan
>            Priority: Major
>              Labels: performance
>         Attachments: Screenshot 2023-01-24 at 10.23.16 AM.png
>
>
> * Runtime stats can be stored in hiveserver or in metastore via "hive.query.reexecution.stats.persist.scope".
>  * Though the table is dropped and recreated, it ends up showing old stats via "RUNTIME" stats. Here is an example (note that the table is empty, but gets datasize and numRows from RUNTIME stats)
>  * This causes suboptimal plan for "MERGE INTO" queries by creating CUSTOM_EDGE instead of broadcast edge.
> !Screenshot 2023-01-24 at 10.23.16 AM.png|width=2053,height=753!
>  
>  



--
This message was sent by Atlassian Jira
(v8.20.10#820010)