Hadoop系列学习–Partitioner内置分区与Partitioner自定义分区

Hadoop系列学习–Partitioner内置分区与Partitioner自定义分区

MapReduce的编程灵活性很高，其中Partitioner分区函数的作用也很重要。

Partitioner分区函数的作用：

根据业务的需求，灵活的根据业务输出多个日志

多个Reduce并发处理日志，提高工作运行的效率

使数据能够均匀分布在reduce上进行操作，避免产生热点区域。

**

默认的Partitioner分区

**

默认的partitioner是HashPartitioner,他对每条记录的键进行哈希操作以决定该记录应该属于哪个分区。每个分区对应一个reducer任务，所以分区数等于作业的reducer的个数。

public class HashPartitioner<K,V> extends Partitioner<K,V>{
@Override
public int getPartition(K k, V v, int numPartitions) {
return (k.hashCode() & Integer.MAX_VALUE) % numPartitions;
}
}

自定义Partitioner分区

默认分区具有限制性，由于它是根据HashCode的值去分区，而有一些业务，例如我们要将日志的11位数与不是11位数的输出日志分开。例如：

log1:
1111111111 aaa
1111111111 aaa1
1111111111 aaa14
1111111111 aaa2
11111111112 aaa3

log2:
222222 aaa
222222 aaa1
222222 aaa14
222222 aaa2
2222222 aaa3

我们需要将key是11位数的分到一个reduce任务，不是11位数的分到其他的reduce中。

/*Partition分区*/
public static class ParitionParition<K,V> extends Partitioner<K,V> {
@Override
public int getPartition(K k, V v, int i) {
int length = k.toString().length();
if(length == 11){
return 0;
} else {
return 1;
}
}
}

自定义Partitioner完整分区代码：

/**
* 分区
* 通过Partition使Maper的输出均匀分配到Reduce中
* 哪个key分配到哪个reduce中，
* Created with IntelliJ IDEA.
* User: Administrator
* Date: 15-5-14
* Time: 下午5:02
* To change this template use File | Settings | File Templates.
*/
public class PartitionerTest {
private static String keystr;
private static String valuestr;
/*Map*/
public static class PartitionMap extends Mapper<LongWritable, Text, LongWritable, Text> {
@Override
protected void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {
StringTokenizer stringTokenizer = new StringTokenizer(value.toString());
while (stringTokenizer.hasMoreTokens()){
keystr = stringTokenizer.nextToken();
valuestr = stringTokenizer.nextToken();
}
LongWritable keywritable = new LongWritable(Long.parseLong(keystr));
context.write(keywritable, new Text(valuestr));
}
}
/*Reduce*/
public static class PartitionReduce extends Reducer<LongWritable, Text, LongWritable, Text> {
@Override
protected void reduce(LongWritable key, Iterable<Text> values, Context context) throws IOException, InterruptedException {
for (Text value : values){
context.write(key, new Text("1"));
}
}
}

/*Partition分区*/
public static class ParitionParition<K,V> extends Partitioner<K,V> {
@Override
public int getPartition(K k, V v, int i) {
int length = k.toString().length();
if(length == 11){
return 0;
} else {
return 1;
}
}
}
public static void main(String[] args) throws IOException, ClassNotFoundException, InterruptedException {
Configuration configuration = new Configuration();
configuration.set("mapred.jar", "test1-1.0-SNAPSHOT-jar-with-dependencies.jar");
configuration.set("fs.default.name", "hdfs://127.0.0.1:9000");//namenode的地址
configuration.set("mapred.job.tracker", "127.0.0.1:9001");
String [] arg = new String[]{"/user/4xt0fuqksoqhq9p/partitioninput","/partitionoutput"};
String [] otherArgs = new GenericOptionsParser(configuration, arg).getRemainingArgs();
if(otherArgs.length < 2){
System.out.println("Use <Word> <Count>");
System.exit(2);
}

Job job = new Job(configuration, "ParitionerTest");
job.setJarByClass(PartitionerTest.class);
job.setMapperClass(PartitionMap.class);
job.setCombinerClass(PartitionReduce.class);
job.setReducerClass(PartitionReduce.class);
//Partition
job.setPartitionerClass(ParitionParition.class);
//
job.setNumReduceTasks(2);

job.setOutputKeyClass(LongWritable.class);
job.setOutputValueClass(Text.class);

FileInputFormat.addInputPath(job, new Path(otherArgs[0]));
FileOutputFormat.setOutputPath(job, new Path(otherArgs[1]));
System.exit(job.waitForCompletion(true) ?0:1);
}
}

得到的日志为