spark-instrumented-optimizer/sql/hive-thriftserver
Kent Yao 8be16907c2 [SPARK-31170][SQL] Spark SQL Cli should respect hive-site.xml and spark.sql.warehouse.dir
### What changes were proposed in this pull request?

In Spark CLI, we create a hive `CliSessionState` and it does not load the `hive-site.xml`. So the configurations in `hive-site.xml` will not take effects like other spark-hive integration apps.

Also, the warehouse directory is not correctly picked. If the `default` database does not exist, the `CliSessionState` will create one during the first time it talks to the metastore. The `Location` of the default DB will be neither the value of `spark.sql.warehousr.dir` nor the user-specified value of `hive.metastore.warehourse.dir`, but the default value of `hive.metastore.warehourse.dir `which will always be `/user/hive/warehouse`.

This PR fixes CLiSuite failure with the hive-1.2 profile in https://github.com/apache/spark/pull/27933.

In https://github.com/apache/spark/pull/27933, we fix the issue in JIRA by deciding the warehouse dir using all properties from spark conf and Hadoop conf, but properties from `--hiveconf` is not included,  they will be applied to the `CliSessionState` instance after it initialized. When this command-line option key is `hive.metastore.warehouse.dir`, the actual warehouse dir is overridden. Because of the logic in Hive for creating the non-existing default database changed, that test passed with `Hive 2.3.6` but failed with `1.2`. So in this PR, Hadoop/Hive configurations are ordered by:
` spark.hive.xxx > spark.hadoop.xxx > --hiveconf xxx > hive-site.xml` througth `ShareState.loadHiveConfFile` before sessionState start

### Why are the changes needed?

Bugfix for Spark SQL CLI to pick right confs

### Does this PR introduce any user-facing change?

yes,
1. the non-exists default database will be created in the location specified by the users via `spark.sql.warehouse.dir` or `hive.metastore.warehouse.dir`, or the default value of `spark.sql.warehouse.dir` if none of them specified.

2. configurations from `hive-site.xml` will not override command-line options or the properties defined with `spark.hadoo(hive).` prefix in spark conf.

### How was this patch tested?

add cli ut

Closes #27969 from yaooqinn/SPARK-31170-2.

Authored-by: Kent Yao <yaooqinn@hotmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
2020-03-27 12:05:45 +08:00
..
src [SPARK-31170][SQL] Spark SQL Cli should respect hive-site.xml and spark.sql.warehouse.dir 2020-03-27 12:05:45 +08:00
v1.2 [SPARK-30904][SQL] Thrift RowBasedSet serialization throws NullPointerException on NULL BigDecimal 2020-02-21 21:39:35 -07:00
v2.3 [SPARK-30904][SQL] Thrift RowBasedSet serialization throws NullPointerException on NULL BigDecimal 2020-02-21 21:39:35 -07:00
pom.xml [SPARK-30950][BUILD] Setting version to 3.1.0-SNAPSHOT 2020-02-25 19:44:31 -08:00