Open FANNG1 opened 6 days ago
There are known limitations with Hive Parquet reader implementation, you may want to try enabling spark.sql.parquet.writeLegacyFormat
when generating TPC-DS using Spark.
There are known limitations with Hive Parquet reader implementation, you may want to try enabling
spark.sql.parquet.writeLegacyFormat
when generating TPC-DS using Spark.
Yes, it works after setting spark.sql.parquet.writeLegacyFormat
to true
. Is there any other limitations about kyuubi spark hive connector ? and any plan to support new parquet format?
Limitations were listed on our first meeting. KSCH only supports using Hive SerDe to read/write Hive tables, consequently, it has the same limitation as Spark's built-in Hive implementation, specifically, Hive 2.3.9, e.g. poor performance because of non-vectorized, does not support new Parquet logical types.
This requires a mechanism to respect spark.sql.hive.convertMetastoreParquet
(or define a new dedicated configuration) to convert the Hive Parquet table reading to the Spark DataSource table.
Code of Conduct
Search before asking
Describe the bug
I follow the steps to generate tpcds data for spark https://github.com/yaooqinn/tpcds-for-spark/tree/master, when querying data get bellow exceptions: select * from catalog_sales limit 1;
select * from store_returns limit 1;
Affects Version(s)
1.8.2
Kyuubi Server Log Output
No response
Kyuubi Engine Log Output
No response
Kyuubi Server Configurations
No response
Kyuubi Engine Configurations
No response
Additional context
No response
Are you willing to submit PR?