How to read from hbase using spark

Question

The below code will read from the hbase, then convert it to json structure and the convert to schemaRDD , But the problem is that I am using List to store the json string then pass to javaRDD, for data of about 100 GB the master will be loaded with data in memory. What is the right way to load the data from hbase then perform manipulation,then convert to JavaRDD.

package hbase_reader;


import java.io.IOException;
import java.io.Serializable;
import java.util.ArrayList;
import java.util.List;

import org.apache.spark.api.java.JavaPairRDD;
import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.JavaSparkContext;
import org.apache.spark.rdd.RDD;
import org.apache.spark.sql.api.java.JavaSQLContext;
import org.apache.spark.sql.api.java.JavaSchemaRDD;
import org.apache.commons.cli.ParseException;
import org.apache.hadoop.hbase.HBaseConfiguration;
import org.apache.hadoop.hbase.KeyValue;
import org.apache.hadoop.hbase.client.HTable;
import org.apache.hadoop.hbase.client.Result;
import org.apache.hadoop.hbase.client.ResultScanner;
import org.apache.hadoop.hbase.client.Scan;
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
import org.apache.hadoop.hbase.mapreduce.TableInputFormat;
import org.apache.hadoop.hbase.util.Bytes;
import org.apache.hadoop.io.Text;
import org.apache.spark.SparkConf;

import scala.Function1;
import scala.Tuple2;
import scala.runtime.AbstractFunction1;

import com.google.common.collect.Lists;

public class hbase_reader {

    public static void main(String[] args) throws IOException, ParseException {

        List<String> jars = Lists.newArrayList("");

        SparkConf spconf = new SparkConf();
        spconf.setMaster("local[2]");
        spconf.setAppName("HBase");
        //spconf.setSparkHome("/opt/human/opt/spark-0.9.0-hdp1");
        spconf.setJars(jars.toArray(new String[jars.size()]));
        JavaSparkContext sc = new JavaSparkContext(spconf);
        //spconf.set("spark.executor.memory", "1g");

        JavaSQLContext jsql = new JavaSQLContext(sc);


        HBaseConfiguration conf = new HBaseConfiguration();
        String tableName = "HBase.CounData1_Raw_Min1";
        HTable table = new HTable(conf,tableName);
        try {

            ResultScanner scanner = table.getScanner(new Scan());
            List<String> jsonList = new ArrayList<String>();

            String json = null;

            for(Result rowResult:scanner) {
                json = "";
                String rowKey  = Bytes.toString(rowResult.getRow());
                for(byte[] s1:rowResult.getMap().keySet()) {
                    String s1_str = Bytes.toString(s1);

                    String jsonSame = "";
                    for(byte[] s2:rowResult.getMap().get(s1).keySet()) {
                        String s2_str = Bytes.toString(s2);
                        for(long s3:rowResult.getMap().get(s1).get(s2).keySet()) {
                            String s3_str = new String(rowResult.getMap().get(s1).get(s2).get(s3));
                            jsonSame += "\""+s2_str+"\":"+s3_str+",";
                        }
                    }
                    jsonSame = jsonSame.substring(0,jsonSame.length()-1);
                    json += "\""+s1_str+"\""+":{"+jsonSame+"}"+",";
                }
                json = json.substring(0,json.length()-1);
                json = "{\"RowKey\":\""+rowKey+"\","+json+"}";
                jsonList.add(json);
            }

            JavaRDD<String> jsonRDD = sc.parallelize(jsonList);

            JavaSchemaRDD schemaRDD = jsql.jsonRDD(jsonRDD);




            System.out.println(schemaRDD.take(2));

        } finally {
            table.close();
        }

    }

}

Answer 1

A Basic Example to Read the HBase data using Spark (Scala), You can also wrtie this in Java :

import org.apache.hadoop.hbase.client.{HBaseAdmin, Result}
import org.apache.hadoop.hbase.{ HBaseConfiguration, HTableDescriptor }
import org.apache.hadoop.hbase.mapreduce.TableInputFormat
import org.apache.hadoop.hbase.io.ImmutableBytesWritable

import org.apache.spark._

object HBaseRead {
  def main(args: Array[String]) {
    val sparkConf = new SparkConf().setAppName("HBaseRead").setMaster("local[2]")
    val sc = new SparkContext(sparkConf)
    val conf = HBaseConfiguration.create()
    val tableName = "table1"

    System.setProperty("user.name", "hdfs")
    System.setProperty("HADOOP_USER_NAME", "hdfs")
    conf.set("hbase.master", "localhost:60000")
    conf.setInt("timeout", 120000)
    conf.set("hbase.zookeeper.quorum", "localhost")
    conf.set("zookeeper.znode.parent", "/hbase-unsecure")
    conf.set(TableInputFormat.INPUT_TABLE, tableName)

    val admin = new HBaseAdmin(conf)
    if (!admin.isTableAvailable(tableName)) {
      val tableDesc = new HTableDescriptor(tableName)
      admin.createTable(tableDesc)
    }

    val hBaseRDD = sc.newAPIHadoopRDD(conf, classOf[TableInputFormat], classOf[ImmutableBytesWritable], classOf[Result])
    println("Number of Records found : " + hBaseRDD.count())
    sc.stop()
  }
}

UPDATED -2016

As of Spark 1.0.x+, Now you can use Spark-HBase Connector also :

Maven Dependency to Include :

<dependency>
  <groupId>it.nerdammer.bigdata</groupId>
  <artifactId>spark-hbase-connector_2.10</artifactId>
  <version>1.0.3</version> // Version can be changed as per your Spark version, I am using Spark 1.6.x
</dependency>

And find a below sample code for the same :

import org.apache.spark._
import it.nerdammer.spark.hbase._

object HBaseRead extends App {
    val sparkConf = new SparkConf().setAppName("Spark-HBase").setMaster("local[4]")
    sparkConf.set("spark.hbase.host", "<YourHostnameOnly>") //e.g. 192.168.1.1 or localhost or your hostanme
    val sc = new SparkContext(sparkConf)

    // For Example If you have an HBase Table as 'Document' with ColumnFamily 'SMPL' and qualifier as 'DocID, Title' then:

    val docRdd = sc.hbaseTable[(Option[String], Option[String])]("Document")
    .select("DocID", "Title").inColumnFamily("SMPL")

    println("Number of Records found : " + docRdd .count())
}

UPDATED - 2017

As of Spark 1.6.x+, Now you can use SHC Connector also (Hortonworks or HDP users) :

Maven Dependency to Include :

    <dependency>
        <groupId>com.hortonworks</groupId>
        <artifactId>shc</artifactId>
        <version>1.0.0-2.0-s_2.11</version> // Version depends on the Spark version and is supported upto Spark 2.x
    </dependency>

The Main advantage of using this connector is that it have flexibility in the Schema definition and doesn't need Hardcoded params just like in nerdammer/spark-hbase-connector. Also remember that it supports Spark 2.x so this connector is pretty much flexible and provides end-to-end support in Issues and PRs.

Find the below repository path for the latest readme and samples :

Hortonworks Spark HBase Connector

You can also convert this RDD's to DataFrames and run SQL over it or You can map these Dataset or DataFrames to user defined Java Pojo's or Case classes. It works brilliant.

Please comment below if you need anything else.

Answer 2

I prefer to read from hbase and do the json manipulation all in spark.
Spark provides JavaSparkContext.newAPIHadoopRDD function to read data from hadoop storage, including HBase. You will have to provide the HBase configuration, table name, and scan in the configuration parameter and table input format and it's key-value

You can use table input format class and it's job parameter to provide the table name and scan configuration

example:

conf.set(TableInputFormat.INPUT_TABLE, "tablename");
JavaPairRDD<ImmutableBytesWritable, Result> data = 
jsc.newAPIHadoopRDD(conf, TableInputFormat.class,ImmutableBytesWritable.class, Result.class);

then you can do the json manipulation in spark. Since spark can do recalculation when the memory is full, it will only load the data needed for the recalculation part (cmiiw) so you don't have to worry about the data size

Answer 3

just to add a comment on how to add scan:

TableInputFormat has the following attributes:

1. SCAN_ROW_START
2. SCAN_ROW_STOP

conf.set(TableInputFormat.SCAN_ROW_START, "startrowkey");
conf.set(TableInputFormat.SCAN_ROW_STOP, "stoprowkey");

Answer 4

Since the question is not new, there are a few other alternatives for now:

• hbase-spark, a module that is available directly in the HBase repo
• Spark-on-HBase by Hortonworks

I do not know much about the first project, but it looks like it does not support Spark 2.x. However, it has a rich support at the RDD level for Spark 1.6.x.

Spark-on-HBase, on the other hand, has branches for Spark 2.0 and upcoming Spark 2.1. This project is very promising since it is focused on Dataset/DataFrame APIs. Under the hood, it implements the standard Spark Datasource API and leverages the Spark Catalyst engine for query optimization. The developers claim here that it is capable of partition pruning, column pruning, predicate pushdown and achieving data locality.

A simple example, which uses the com.hortonworks:shc:1.0.0-2.0-s_2.11 artifact from this repo and Spark 2.0.2, is presented next:

case class Record(col0: Int, col1: Int, col2: Boolean)

val spark = SparkSession
  .builder()
  .appName("Spark HBase Example")
  .master("local[4]")
  .getOrCreate()

def catalog =
  s"""{
      |"table":{"namespace":"default", "name":"table1"},
      |"rowkey":"key",
      |"columns":{
      |"col0":{"cf":"rowkey", "col":"key", "type":"int"},
      |"col1":{"cf":"cf1", "col":"col1", "type":"int"},
      |"col2":{"cf":"cf2", "col":"col2", "type":"boolean"}
      |}
      |}""".stripMargin

val artificialData = (0 to 100).map(number => Record(number, number, number % 2 == 0))

// write
spark
  .createDataFrame(artificialData)
  .write
  .option(HBaseTableCatalog.tableCatalog, catalog)
  .option(HBaseTableCatalog.newTable, "5")
  .format("org.apache.spark.sql.execution.datasources.hbase")
  .save()

// read
val df = spark
  .read
  .option(HBaseTableCatalog.tableCatalog, catalog)
  .format("org.apache.spark.sql.execution.datasources.hbase")
  .load()

df.count()