You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Dongjoon Hyun (JIRA)" <ji...@apache.org> on 2019/07/28 22:44:00 UTC
[jira] [Assigned] (SPARK-25474) Support
`spark.sql.statistics.fallBackToHdfs` in data source tables
[ https://issues.apache.org/jira/browse/SPARK-25474?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Dongjoon Hyun reassigned SPARK-25474:
-------------------------------------
Assignee: shahid
> Support `spark.sql.statistics.fallBackToHdfs` in data source tables
> -------------------------------------------------------------------
>
> Key: SPARK-25474
> URL: https://issues.apache.org/jira/browse/SPARK-25474
> Project: Spark
> Issue Type: Bug
> Components: SQL
> Affects Versions: 2.3.1, 2.4.3
> Environment: Spark 2.3.1
> Hadoop 2.7.2
> Reporter: Ayush Anubhava
> Assignee: shahid
> Priority: Major
> Fix For: 3.0.0
>
>
> *Description :* Size in bytes of the query is coming in EB in case of parquet datasource. this would impact the performance , since join queries would always go as Sort Merge Join.
> *Precondition :* spark.sql.statistics.fallBackToHdfs = true
> Steps:
> {code:java}
> 0: jdbc:hive2://10.xx:23040/default> create table t1110 (a int, b string) using parquet PARTITIONED BY (b) ;
> +---------+--+
> | Result |
> +---------+--+
> +---------+--+
> 0: jdbc:hive2://10.1xx:23040/default> insert into t1110 values (2,'b');
> +---------+--+
> | Result |
> +---------+--+
> +---------+--+
> 0: jdbc:hive2://10.1xx:23040/default> insert into t1110 values (1,'a');
> +---------+--+
> | Result |
> +---------+--+
> +---------+--+
> 0: jdbc:hive2://10.xx.xx:23040/default> select * from t1110;
> +----+----+--+
> | a | b |
> +----+----+--+
> | 1 | a |
> | 2 | b |
> +----+----+--+
> {code}
> *{color:#d04437}Cost of the query shows sizeInBytes in EB{color}*
> {code:java}
> explain cost select * from t1110;
> | == Optimized Logical Plan ==
> Relation[a#23,b#24] parquet, Statistics(sizeInBytes=8.0 EB, hints=none)
> == Physical Plan ==
> *(1) FileScan parquet open.t1110[a#23,b#24] Batched: true, Format: Parquet, Location: CatalogFileIndex[hdfs://hacluster/user/sparkhive/warehouse/open.db/t1110], PartitionCount: 2, PartitionFilters: [], PushedFilters: [], ReadSchema: struct<a:int> |
> {code}
> *{color:#d04437}This would lead to Sort Merge Join in case of join query{color}*
> {code:java}
> 0: jdbc:hive2://10.xx.xx:23040/default> create table t110 (a int, b string) using parquet PARTITIONED BY (b) ;
> +---------+--+
> | Result |
> +---------+--+
> +---------+--+
> 0: jdbc:hive2://10.xx.xx:23040/default> insert into t110 values (1,'a');
> +---------+--+
> | Result |
> +---------+--+
> +---------+--+
> explain select * from t1110 t1 join t110 t2 on t1.a=t2.a;
> | == Physical Plan ==
> *(5) SortMergeJoin [a#23], [a#55], Inner
> :- *(2) Sort [a#23 ASC NULLS FIRST], false, 0
> : +- Exchange hashpartitioning(a#23, 200)
> : +- *(1) Project [a#23, b#24]
> : +- *(1) Filter isnotnull(a#23)
> : +- *(1) FileScan parquet open.t1110[a#23,b#24] Batched: true, Format: Parquet, Location: CatalogFileIndex[hdfs://hacluster/user/sparkhive/warehouse/open.db/t1110], PartitionCount: 2, PartitionFilters: [], PushedFilters: [IsNotNull(a)], ReadSchema: struct<a:int>
> +- *(4) Sort [a#55 ASC NULLS FIRST], false, 0
> +- Exchange hashpartitioning(a#55, 200)
> +- *(3) Project [a#55, b#56]
> +- *(3) Filter isnotnull(a#55)
> +- *(3) FileScan parquet open.t110[a#55,b#56] Batched: true, Format: Parquet, Location: CatalogFileIndex[hdfs://hacluster/user/sparkhive/warehouse/open.db/t110], PartitionCount: 1, PartitionFilters: [], PushedFilters: [IsNotNull(a)], ReadSchema: struct<a:int> |
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org