关于线程池ConnectionPool的临时记录

hadoop中router就是自己实现线程池来维护响应线程，来响应请求的。

package org.apache.hadoop.hdfs.server.federation.router;
public class ConnectionPool {

  protected ConnectionContext getConnection() {

    this.lastActiveTime = Time.now();

    // Get a connection from the pool following round-robin
    ConnectionContext conn = null;
    List<ConnectionContext> tmpConnections = this.connections;
    int size = tmpConnections.size();
    // Inc and mask off sign bit, lookup index should be non-negative int
    int threadIndex = this.clientIndex.getAndIncrement() & 0x7FFFFFFF;
    for (int i=0; i<size; i++) {
      int index = (threadIndex + i) % size;
      conn = tmpConnections.get(index);
      if (conn != null && conn.isUsable()) {
        return conn;
      }
    }

    // We return a connection even if it's active
    return conn;
  }

}

connectionId是由ugi、addr、协议来确定的。

   this.connectionPoolId =
        new ConnectionPoolId(this.ugi, this.namenodeAddress, this.protocol);

该连接池的设计与《并发编程的艺术》第4章线程实例应用几乎一样，值得学习。