Spark将HDFS数据导入到HBase

本程序运行环境：Spark+HDFS+HBase+Yarn
hadoop（HDFS+Yarn）集群搭建，参考：http://blog.csdn.net/korder/article/details/46909253
Spark on Yarn，参考：http://blog.csdn.net/korder/article/details/47422345
HBase集群搭建，参考：http://blog.csdn.net/korder/article/details/47423247

hbase表结构为：表名table，列族fam,列为col。

第一步：上代码
object inputHbase：

import org.apache.hadoop.hbase.client._
import org.apache.hadoop.hbase.util.Bytes
import org.apache.spark.{SparkContext, SparkConf}
import org.apache.hadoop.hbase._


/**
 * Created by Chensy on 15-8-10.
 */
object inputHbase {
  /**
   * hbase table:table col-family:fam col:col
   */

  def main(args: Array[String]) {
    val conf = new SparkConf()
    val sc = new SparkContext(conf)
    val readFile = sc.textFile(args(0)).map(x => x.split(","))
    val tableName = "table"

    readFile.foreachPartition{
        x=> {
          val myConf = HBaseConfiguration.create()
          myConf.set("hbase.zookeeper.quorum","172.23.27.45,172.23.27.46,172.23.27.47")
          myConf.set("hbase.zookeeper.property.clientPort","2181")
          myConf.set("hbase.defaults.for.version.skip","true")
          myConf.set("hbase.master","172.23.27.39:60000")
          myConf.set("hbase.cluster.distributed","true")
          myConf.set("hbase.rootdir","hdfs://cdh5-test/hbase")


          val myTable = new HTable(myConf,TableName.valueOf(tableName))
          //将自动提交关闭，如果不关闭，每写一条数据都会进行提交，是导入数据较慢的做主要因素。
          myTable.setAutoFlush(false,false)
          //设置缓存大小，当缓存大于设置值时，hbase会自动提交。此处可自己尝试大小，一般对大数据量，设置为5M即可。
          myTable.setWriteBufferSize(5*1024*1024)

          x.foreach{ y=> {
                val p = new Put(Bytes.toBytes("row"+y(0)))
                p.add("fam".getBytes,"col".getBytes,Bytes.toBytes("value"+y(0)))
                myTable.put(p)
            }

          }
          //每一个分片结束后都进行flushCommits()，如果不执行，当hbase最后缓存小于上面设定值时，不会进行提交，导致数据丢失。
          myTable.flushCommits()
        }
    }
    System.exit(0)
  }

}

第二步：打包，并传至HDFS

idea打包就不说了，inputHbase.jar
hadoop fs -put inputHbase.jar /xxx/spark/streaming 

第三步：添加相关jars
建个公共库，把需要用到的jar包存放一起，方便添加

第四步：编写执行脚本：submit-yarn-inputHbase.sh

[root@JXQ-23-27-38 streaming]# vim submit-yarn-inputHbase.sh 

cd $SPARK_HOME

#pwd

./bin/spark-submit --name inputHbase \
                   --class com.wylog.hbase.inputHbase \
                   --master yarn-cluster \
                   --num-executors 8 \
                   --executor-memory 4g  \
                   --executor-cores 8 \
                   --driver-memory 4g \
                   --driver-cores 4 \
                   --jars /root/spark/streaming/public_lib/hbase-client-0.98.6-cdh5.3.2.jar,
/root/spark/streaming/public_lib/hbase-server-0.98.6-cdh5.3.2.jar,
/root/spark/streaming/public_lib/hbase-protocol-0.98.6-cdh5.3.2.jar,
/root/spark/streaming/public_lib/htrace-core-2.04.jar  \
                   hdfs://cdh5-test/xxx/spark/streaming/inputHbase.jar \
                   hdfs://cdh5-test/data/notify-server/172.17.88.88/notify-server2_detail.log.*

参考资料：
http://www.cloudera.com/content/cloudera/zh-CN/documentation/core/v5-3-x/topics/admin_hbase_import.html#concept_asc_ctz_wp_unique_1