[
https://issues.apache.org/jira/browse/HIVE-29203?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18020604#comment-18020604
]
Zhihua Deng commented on HIVE-29203:
------------------------------------
If we remove the batch processing from
[https://github.com/apache/hive/blob/4bb08099d91acbefee73a449a36abb1ecd2b5925/standalone-metastore/metastore-server/src/main/java/org/apache/hadoop/hive/metastore/MetaStoreDirectSql.java#L1882-L1891,]
perhaps we need to consider the memory usage when enableBitVector ||
enableKll, and the
restrictions on IN list size in aggrStatsUseDB.
> get_aggr_stats_for doesn't aggregate stats when direct sql batch retrieve is
> enabled
> ------------------------------------------------------------------------------------
>
> Key: HIVE-29203
> URL: https://issues.apache.org/jira/browse/HIVE-29203
> Project: Hive
> Issue Type: Bug
> Components: Standalone Metastore
> Reporter: Zhihua Deng
> Priority: Major
>
> In case of metastore.direct.sql.batch.size > 0, and number of partition names
> or columns in get_aggr_stats_for is bigger than the
> metastore.direct.sql.batch.size, then the
> AggrStats from the call get_aggr_stats_for might have un-merged stats for the
> same column, so the aggregated stats is not correct, which may make CBO
> generate an outdated execution plan.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
