Set up

Cloudera artifactory and HWC dependency

To pull the HWC dependency corresponding to a release, use the following artifactory:

    https://repository.cloudera.com/artifactory/cloudera-repos 
   

Use with Maven

To use HWC with maven, define the cloudera artifactory as a repository.

<repository>
  <id>cloudera</id>
  <name>cloudera</name>
  <url>https://repository.cloudera.com/artifactory/cloudera-repos</url>
 </repository>

In the pom.xml of the project, add the dependency as shown in the following example:

<dependency>
   <groupId>com.hortonworks.hive</groupId>
   <artifactId>hive-warehouse-connector_2.11</artifactId>
   <version>[***HWC VERSION***]</version>
   <scope>provided</scope>
</dependency>

Use with Sbt

Add the Cloudera and Shibboleth repositories, and the HWC dependency to the build sbt as follows:

resolvers += "Cloudera repo" at "https://repository.cloudera.com/artifactory/cloudera-repos",

resolvers += "opensaml Repository" at "https://build.shibboleth.net/nexus/content/repositories/releases",

libraryDependencies += "com.hortonworks.hive" % "hive-warehouse-connector_2.11" % "[***HWC VERSION***]" % "provided",

Dependency scope

Generally, you add HWC dependencies in provided scope unless there is a specific requirement to do otherwise. While running spark application, you can specify the HWC jar present in your distribution using the --jars option to spark-submit or spark-shell.

HWC binaries in CDP

HWC binaries are located within the appropriate directory in /opt/cloudera/parcels/CDH/lib/. The directory contains the HWC jar, a python zip, and the R package. Use these binaries to launch Spark applications in Scala, Java, Python, or R.

For Spark 2, the files are available in /opt/cloudera/parcels/CDH/lib/hive_warehouse_connector/.

• hive-warehouse-connector-assembly-<version>.jar
• pyspark_hwc-<version>.zip
• SparklyrHWC-<version>

HWC configurations

Learn about the configurations that are required by Spark when using HWC. Setting HWC configurations has been simplified. As a cluster administrator, you can specify the required configurations in Cloudera Manager and then enable HWC by setting spark.cloudera.useHWC=true. You can either add this property to spark-defaults.conf to enable HWC for all Spark jobs or enable HWC for a specific Spark job by using the --conf option in spark-shell or spark-submit.

Add the following configurations in the spark-defaults.conf file by going to Clusters > SPARK_ON_YARN-1 > Configuration in Cloudera Manager. This is a one-time activity that is performed by a cluster administrator.

• spark.sql.hive.hiveserver2.jdbc.url=jdbc:hive2://<jdbc-url>
• spark.datasource.hive.warehouse.read.mode=<mode>
• spark.security.credentials.hiveserver2.enabled=true
• spark.sql.hive.hiveserver2.jdbc.url.principal=<principal>/_HOST@ROOT.HWX.SITE
• spark.datasource.hive.warehouse.load.staging.dir=<hdfs://>/tmp/staging/hwc

Optional HWC configurations

Optionally, you can set the following properties:

• spark.datasource.hive.warehouse.write.path.strictColumnNamesMapping — Validates the mapping of columns against those in Hive to alert the user to input errors. Default = true.
• spark.sql.hive.conf.list — Propagates one or more configuration properties from the HWC to Hive. Set properties on the command line using the --conf option. For example:

  --conf spark.sql.hive.conf.list="hive.vectorized.execution.filesink.arrow.native.enabled=true;hive.vectorized.execution.enabled=true"

Do not attempt to set spark.sql.hive.conf.list programmatically.

Working with different languages

You use HWC APIs to perform basic read and write operations. You need to understand how to use HWC APIs with different languages. The following examples show basic capabilities that are covered in detail later in this documentation.

Use with Scala

import com.hortonworks.hwc.HiveWarehouseSession
import org.apache.spark.sql.{SaveMode, SparkSession}


object HWCApp {

  def main(args: Array[String]): Unit = {
    val spark = SparkSession.builder.appName("HWCApp").enableHiveSupport.getOrCreate
    val hwc = HiveWarehouseSession.session(spark).build
    // create sample data
    val tvSeries = createSampleDataDf(spark)
    val tableName = "tv_series"

    hwc.dropTable(tableName, true, true)

    println(s"=======Writing to hive table - $tableName via HWC=======")
    // write to hive table via HWC
    tvSeries.write.format(HiveWarehouseSession.HIVE_WAREHOUSE_CONNECTOR)
      .option("table", tableName)
      .mode(SaveMode.Append).save

    println(s"=======Reading hive table $tableName via HWC=======")
    // Read via HWC
    hwc.sql(s"select * from $tableName").show(truncate = false)

    hwc.close()
    spark.stop
  }

  private def createSampleDataDf(spark: SparkSession) = {
    spark.sql("drop table if exists tv_series_dataset")
    spark.sql("create table tv_series_dataset(id int, name string, genres string, rating double) using orc")
    spark.sql("insert into tv_series_dataset values " +
      "(1, 'Chernobyl', 'Drama|History|Tragedy|Science', 9.4), " +
      "(2, 'Westworld', 'Sci-fi', 8.6), (3, 'Sense8', 'Sci-fi', 8.3), " +
      "(4, 'Person of Interest', 'Drama|Sci-fi', 8.4), " +
      "(5, 'Its okay to not be okay', 'Drama', 8.7), " +
      "(6, 'Daredevil', 'Action|Sci-fi', 8.6), " +
      "(7, 'Money Heist', 'Drama|Thriller', 8.3), " +
      "(8, 'Breaking Bad', 'Crime|Drama', 9.5)")
    spark.sql("select * from tv_series_dataset")
  }

Use with Java

The following Java code is equivalent to the scala code above.

import com.hortonworks.hwc.HiveWarehouseSession
import org.apache.spark.sql.Dataset;
import org.apache.spark.sql.Row;
import org.apache.spark.sql.SaveMode;
import org.apache.spark.sql.SparkSession;

public class HWCApp {
  public static void main(String[] args) {
    SparkSession spark = SparkSession.builder().appName("HWCApp").enableHiveSupport().getOrCreate();
    // HiveWarehouseSession creation
    HiveWarehouseSession hwc = HiveWarehouseBuilder.session(spark).build();
    // create sample data
    Dataset<Row> tvSeries = createSampleDataDf(spark);
    String tableName = "tv_series";
    hwc.dropTable(tableName, true, true);
    System.out.println("=======Writing to hive table - " + tableName + " via HWC=======");
    // write data to hive table via HWC
    tvSeries.write().format(HiveWarehouseSession.HIVE_WAREHOUSE_CONNECTOR)
        .option("table", tableName)
        .mode(SaveMode.Append).save();

    System.out.println("=======Reading hive table - " + tableName + " via HWC=======");
    // read hive table as dataframe using HWC
    hwc.sql("select * from " + tableName).show(false);

    hwc.close();
    spark.stop();
  }
  private static Dataset<Row> createSampleDataDf(SparkSession spark) {
    spark.sql("drop table if exists tv_series_dataset");
    spark.sql("create table tv_series_dataset(id int, name string, genres string, rating double) using orc");
    spark.sql("insert into tv_series_dataset values " +
        "(1, 'Chernobyl', 'Drama|History|Tragedy|Science', 9.4), " +
        "(2, 'Westworld', 'Sci-fi', 8.6), (3, 'Sense8', 'Sci-fi', 8.3), " +
        "(4, 'Person of Interest', 'Drama|Sci-fi', 8.4), " +
        "(5, 'Its okay to not be okay', 'Drama', 8.7), " +
        "(6, 'Daredevil', 'Action|Sci-fi', 8.6), " +
        "(7, 'Money Heist', 'Drama|Thriller', 8.3), " +
        "(8, 'Breaking Bad', 'Crime|Drama', 9.5)");
    return spark.sql("select * from tv_series_dataset");
  }
}

Launching a Java or Scala app

After packaging the app in a jar, launch the app using standard Spark syntax for launching applications. Provide HWC jar from the distribution. The Spark application can be launched as follows:

spark-submit --jars /opt/cloudera/parcels/CDH/lib/hive_warehouse_connector/hive-warehouse-connector-assembly-<version>.jar \
--class com.cloudera.HWCApp \
...More spark/HWC confs...
...More spark/HWC confs...
/path-to-jar/hwc-app.jar 

Use with Python

from pyspark.sql import SparkSession
from pyspark_llap import HiveWarehouseSession

spark = SparkSession.builder.enableHiveSupport().appName("hwc-app").getOrCreate()
hwc = HiveWarehouseSession.session(spark).build()

tableName = "tv_series"
hwc.dropTable(tableName, True, True)

tvSeries = spark.createDataFrame([
    (1, "Chernobyl", "Drama|History|Tragedy|Science", 9.4),
    (2, "Westworld", "Sci-fi", 8.6),
    (3, "Sense8", "Sci-fi", 8.3),
    (4, "Person of Interest", "Drama|Sci-fi", 8.4),
    (5, "It's okay to not be okay", "Drama", 8.7),
    (6, "Daredevil", "Action|Sci-fi", 8.6),
    (7, "Money Heist", "Drama|Thriller", 8.3),
    (8, "Breaking Bad", "Crime|Drama", 9.5)
], ["id", "name", "genres", "rating"])

print("=======Writing to hive table - " + tableName + " via HWC=======")
# write to hive table via HWC
tvSeries.write.format(HiveWarehouseSession.HIVE_WAREHOUSE_CONNECTOR).option("table", tableName).mode("append").save()

print("=======Reading hive table - " + tableName + " via HWC=======")
# Read via HWC
hwc.sql("select * from " + tableName).show()

hwc.close()
spark.stop()

Launching a Python app

After getting the python code ready, launch it using spark-submit. Provide the HWC jar and HWC python zip as follows:

    spark-submit --jars /opt/cloudera/parcels/CDH/lib/hive_warehouse_connector/hive-warehouse-connector-assembly-<version>.jar \
    --py-files /opt/cloudera/parcels/CDH/lib/hive_warehouse_connector/pyspark_hwc-<version>.zip \
    ...More spark/HWC confs...
    ...More spark/HWC confs...
    /path-to-python-app/hwc-app.py

Use with Sparklyr

You can access Hive tables through R by loading the sparklyr library along with the SparklyrHWC package available in /opt/cloudera/parcels/CDH/lib/hive_warehouse_connector/, which can be used to trigger HWC APIs from R.

library(sparklyr)
library(SparklyrHWC, lib.loc = c(file.path(“<path to SparklyrHWC>")))

#Set env variables

Sys.setenv(SPARK_HOME = "/opt/cloudera/parcels/CDH/lib/spark/")
Sys.setenv(HADOOP_HOME = "/opt/cloudera/parcels/CDH/lib/hadoop")

#Configurations needed to use spark-acid and related configurations.

config <- spark_config()
config$spark.sql.hive.hiveserver2.jdbc.url="jdbc:hive2://<url>:10000/default"
config$spark.datasource.hive.warehouse.user.name="hive"
config$spark.hadoop.hive.metastore.uris="thrift://<url>:9083"
config$spark.sql.extensions="com.hortonworks.spark.sql.rule.Extensions"
config$spark.kryo.registrator="com.qubole.spark.hiveacid.util.HiveAcidKyroRegistrator"
config$spark.datasource.hive.warehouse.read.mode="DIRECT_READER_V2"

#Build HWC session
hs <- build(HiveWarehouseBuilder.session(sc))

#Use database
sparklyr::sdf_sql(sc,"use test")
#Reading a managed table using spark acid direct-reader
intDf <- sparklyr::spark_read_table(sc, 'emp_hwc')
#Converts SparkDataframe to R dataframe 
 sparklyr::sdf_collect(intDf1)  

#Writing into a managed table
#Read first table
intDf <- sparklyr::spark_read_table(sc, 'emp_hwc')
#read second table
intDf1 <-  sparklyr::spark_read_table(sc, 'emp_overwrite') 
#Commit transaction if read using spark-acid
commitTxn(hs) 
#Append the second table, to the first.
SparklyrHWC::spark_write_table('emp_hwc',intDf1,'append') 
#Overwrite the first table with the second table.
SparklyrHWC::spark_write_table('emp_hwc',intDf1,'overwrite') 

#Using HWC Api’s

#create a table from existing table
SparklyrHWC::executeUpdate(hs,"create table hwc1 as select * from 'emp_hwc'")
#Execute query
hwcDf <- SparklyrHWC::executeQuery(hs, "select * from hwc1")
#convert into R dataframe.
hwcSdf <- sparklyr::sdf_copy_to(sc, hwcDf)