HBase协处理器统计表数据量

1.Java代码实现

import org.apache.hadoop.hbase.client.coprocessor.AggregationClient;
import org.apache.hadoop.hbase.client.coprocessor.LongColumnInterpreter;
import org.apache.hadoop.hbase.coprocessor.AggregateImplementation;

/**
* <p>
* 协处理器统计HBase表数据量
* </p>
* 
*/
public class HBaseRecordsCounter {

/**
* HBase API添加协处理器
* */
public static void addCoprocessor(Configuration conf, String tableName) {
try {

  byte[] tableNameBytes = Bytes.toBytes(tableName);
  HBaseAdmin hbaseAdmin = new HBaseAdmin(conf);
  HTableDescriptor htd = hbaseAdmin.getTableDescriptor(tableNameBytes);
  if (!htd.hasCoprocessor(AggregateImplementation.class.getName())) {
    hbaseAdmin.disableTable(tableNameBytes);
    htd.addCoprocessor(AggregateImplementation.class.getName());
    hbaseAdmin.modifyTable(tableNameBytes, htd);
    hbaseAdmin.enableTable(tableNameBytes);
  }

  hbaseAdmin.close();

} catch (MasterNotRunningException e) {
e.printStackTrace();
} catch (ZooKeeperConnectionException e) {
e.printStackTrace();
} catch (IOException e) {
e.printStackTrace();
}
}

/**
* 统计表数量
* 
*/
public static void exeCount(Configuration conf, String tableName, String family) {

try {
  // 使用hbase提供的聚合coprocessor
  AggregationClient aggregationClient = new AggregationClient(conf);
  Scan scan = new Scan();
  // 指定扫描列族，唯一值
  scan.addFamily(Bytes.toBytes(family));
  long start = System.currentTimeMillis();
  long rowCount = aggregationClient.rowCount(TableName.valueOf(tableName), new LongColumnInterpreter(), scan);
  System.out
  .println("Row count: " + rowCount + "; time cost: " + (System.currentTimeMillis() - start) + "ms");
} catch (Throwable e) {
  e.printStackTrace();
}
}

public static void main(String[] args) {

  String tableName = "test";
  Configuration conf = new Configuration();
  conf.set("hbase.zookeeper.quorum", "host1,host2,host3");
  conf.set("hbase.rootdir", "hdfs://host:8020/hbase");
  // 提高RPC通信时长
  conf.setLong("hbase.rpc.timeout", 600000);
  // 设置Scan缓存
  conf.setLong("hbase.client.scanner.caching", 1000);
  addCoprocessor(conf, tableName);
  exeCount(conf, tableName, "info");

}
}

2. 启用协处理器

启用协处理器方法1.

启动全局aggregation，能过操纵所有的表上的数据。通过修改hbase-site.xml这个文件来实现，只需要添加如下代码：

<property>
   <name>hbase.coprocessor.user.region.classes</name>
   <value>org.apache.hadoop.hbase.coprocessor.AggregateImplementation</value>
 </property>

启用协处理器方法2.

hbase shell添加coprocessor:

disable 'member'
alter 'member',METHOD => 'table_att','coprocessor' => 'hdfs://master24:9000/user/hadoop/jars/test.jar|mycoprocessor.SampleCoprocessor|1001|'
enable 'member'

hbase shell 删除coprocessor:

disable 'member'
alter 'member',METHOD => 'table_att_unset',NAME =>'coprocessor$1'
enable 'member'