2

I am using the below code to write to hbase 

    jsonDStream.foreachRDD(new Function<JavaRDD<String>, Void>() {

        @Override
        public Void call(JavaRDD<String> rdd) throws Exception {

            DataFrame jsonFrame = sqlContext.jsonRDD(rdd);
            DataFrame selecteFieldFrame = jsonFrame.select("id_str","created_at","text");

            Configuration config = HBaseConfiguration.create();
            config.set("hbase.zookeeper.quorum", "d-9543");
            config.set("zookeeper.znode.parent","/hbase-unsecure");
            config.set("hbase.zookeeper.property.clientPort", "2181");
            final JobConf jobConfig=new JobConf(config,SveAsHadoopDataSetExample.class);

            jobConfig.setOutputFormat(TableOutputFormat.class);
            jobConfig.set(TableOutputFormat.OUTPUT_TABLE,"tableName");
             selecteFieldFrame.javaRDD().mapToPair(new PairFunction<Row, ImmutableBytesWritable, Put>() {

                @Override
                public Tuple2<ImmutableBytesWritable, Put> call(Row row) throws Exception {
                    // TODO Auto-generated method stub
                    return convertToPut(row);
                }
            }).saveAsHadoopDataset(jobConfig);


            return null;
        }
    });

But when i see zkDump in zookeeper the connections keeps on increasing

any suggestion/pointers will be of a great help!

Amit_Hora
  • 716
  • 1
  • 8
  • 27

1 Answers1

4

I have the same problem, it is a hbase bug, I fix it:

change org.apache.hadoop.hbase.mapred.TableOutputFormat to org.apache.hadoop.hbase.mapreduce.TableOutputFormat, and use org.apache.hadoop.mapreduce.Job, not org.apache.hadoop.mapred.JobConf

this is a sample:

import org.apache.hadoop.mapreduce.Job
import org.apache.hadoop.hbase.mapreduce.TableOutputFormat

val conf = HBaseConfiguration.create()
conf.set("hbase.zookeeper.quorum", zk_hosts)
conf.set("hbase.zookeeper.property.clientPort", zk_port)

conf.set(TableOutputFormat.OUTPUT_TABLE, "TABLE_NAME")
val job = Job.getInstance(conf)
job.setOutputFormatClass(classOf[TableOutputFormat[String]])

formatedLines.map{
  case (a,b, c) => {
    val row = Bytes.toBytes(a)

    val put = new Put(row)
    put.setDurability(Durability.SKIP_WAL)

    put.addColumn(Bytes.toBytes("cf"), Bytes.toBytes("node"), Bytes.toBytes(b))
    put.addColumn(Bytes.toBytes("cf"), Bytes.toBytes("topic"), Bytes.toBytes(c))

    (new ImmutableBytesWritable(row), put)
  }
}.saveAsNewAPIHadoopDataset(job.getConfiguration)

this may help you!

https://github.com/hortonworks-spark/shc/pull/20/commits/2074067c42c5a454fa4cdeec18c462b5367f23b9

leocook
  • 191
  • 1
  • 13
  • 2
    Whilst this may theoretically answer the question, [it would be preferable](//meta.stackoverflow.com/q/8259) to include the essential parts of the answer here, and provide the link for reference. – Bhargav Rao Oct 10 '16 at 12:15
  • Thank you man! Still most manuals have the old way which has the bug. I was expecting to have stg. job.close(). My problem was exactly same. – zorkaya Dec 22 '16 at 06:42