一次TNSPING速度极慢

9i单节点库，TNSPING高达15000ms。观察到客户端一直在不停的fork和释放监听子进程，但是listener.ora已经添加了

SUBSCRIBE_FOR_NODE_DOWN_EVENT_LISTENER=OFF。

客户端连接也不多，每秒6次。一般这种重启应该马上就好了。客户很多都是短连接，无法马上重启监听，刚到了下班

建议重启监听时，在不同的端口上多配置几个监听器

修改local_listener，使PMON向各个端口注册监听
alter system set local_listener='(DESCRIPTION=(ADDRESS_LIST=(ADDRESS=(PROTOCOL=TCP)(HOST=83.16.16.201)(PORT=1521))(ADDRESS=(PROTOCOL=TCP)(HOST=83.16.16.201)(PORT=1522))(ADDRESS=(PROTOCOL=TCP)(HOST=83.16.16.201)(PORT=1523))(ADDRESS=(PROTOCOL=TCP)(HOST=83.16.16.201)(PORT=1524))))'

使用如下TNS连接数据库
T11203 =
  (DESCRIPTION =
      (LOAD_BALANCE = ON)
      (ADDRESS = (PROTOCOL = TCP)(HOST = 83.16.16.201)(PORT = 1521))
      (ADDRESS = (PROTOCOL = TCP)(HOST = 83.16.16.201)(PORT = 1522))
      (ADDRESS = (PROTOCOL = TCP)(HOST = 83.16.16.201)(PORT = 1523))
      (ADDRESS = (PROTOCOL = TCP)(HOST = 83.16.16.201)(PORT = 1524))
      (CONNECT_DATA = (SERVER = DEDICATED)(SERVICE_NAME = o11203))
  )

这样可以随意重启有问题的监听，由于使用了客户端load_balance，客户端将随机选择一个监听连接

来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/8242091/viewspace-742298/，如需转载，请注明出处，否则将追究法律责任。

转载于:http://blog.itpub.net/8242091/viewspace-742298/