Oracle RAC 客户端连接负载均衡(Load Balance)

        实现负载均衡(Load Balance)是Oracle RAC最重要的特性之一，主要是把负载平均分配到集群中的各个节点，以提高系统的整体吞吐能力。通常情况下有两种方式来实现负载均衡，一个是基于客户端连接的负载均衡，一个是基于服务器端监听器(Listener)收集到的信息来将新的连接请求分配到连接数较少实例上的实现方式。本文主要讨论的是基于客户端连接的负载均衡，并给出演示。

        与负载均衡配置之前的监听配置请参考
              ORACLE RAC 监听配置 (listener.ora tnsnames.ora)
              ORACLE RAC 下非缺省端口监听配置(listener.ora tnsnames.ora)

 

一、客户端的负载均衡
         客户端的负载均衡主要是通过为tnsnames.ora增加load_balance=yes条目来实现，下面看看oracle(Note:226880.1)的解释

        The client load balancing feature enables clients to randomize connection requests among the listeners. Oracle Net progresses through the list of
        protocol addresses in a random sequence, balancing the load on the various listeners. Without client load balancing, Oracle Net progresses through the
        list of protocol addresses sequentially until one succeeds.  This normally is referred to connect-time load balance.

        从上面的描述中可以得知，如果未开启load_balance=yes时，Oracle Net会根据地址列表按顺序来选择一个进行连接，直到连接成功为止。  
        如果第一个host主机连接失败，在有多个地址的情形下，接下来选择第二个地址连接，依此类推，直到连接成功为止。
        当开启了load_balance=yes时，则Oracle Net会从多个地址中随机地选择一个地址进行连接，直到连接成功为止。
        注意，此连接方式仅根据地址列表随机选择，并不考虑到各个实例上当前真正连接数量的多少，也即是没有考虑各个节点真实的连接负载情况。

 

二、服务器与客户端的配置情况

1、服务器端监听器配置
oracle@bo2dbp:~> more $ORACLE_HOME/network/admin/listener.ora  --#节点bo2dbp上的listener
# listener.ora.bo2dbp Network Configuration File: /u01/oracle/db/network/admin/listener.ora.bo2dbp
# Generated by Oracle configuration tools.

LISTENER_NEW_BO2DBP =
  (DESCRIPTION_LIST =
    (DESCRIPTION =
      (ADDRESS = (PROTOCOL = TCP)(HOST = bo2dbp-vip.2gotrade.com)(PORT = 1314)(IP = FIRST))
      (ADDRESS = (PROTOCOL = TCP)(HOST = 192.168.7.51)(PORT = 1314)(IP = FIRST))
    )
  )
...........................--#其余部分省略，注意此处的配置是使用了1314的非缺省监听器端口号   
 
oracle@bo2dbs:/u01/oracle/db/network/admin> more listener.ora  --#节点bo2dbs上的listener
# listener.ora.bo2dbs Network Configuration File: /u01/oracle/db/network/admin/listener.ora.bo2dbs
# Generated by Oracle configuration tools.

LISTENER_NEW_BO2DBS =
  (DESCRIPTION_LIST =
    (DESCRIPTION =
      (ADDRESS = (PROTOCOL = TCP)(HOST = bo2dbs-vip.2gotrade.com)(PORT = 1314)(IP = FIRST))
      (ADDRESS = (PROTOCOL = TCP)(HOST = 192.168.7.52)(PORT = 1314)(IP = FIRST))
    )
  )
...............

2、参数配置
-->instrance ora10g1上的参数配置
SQL> show parameter instance_na

NAME                                 TYPE        VALUE
------------------------------------ ----------- ------------------------------
instance_name                        string      ora10g1

SQL> show parameter listener

NAME                                 TYPE        VALUE
------------------------------------ ----------- ------------------------------
local_listener                       string      local_lsnr_ora10g1
remote_listener                      string      remote_lsnr_ora10g

-->instrance ora10g2上的参数配置
SQL> show parameter instance_name

NAME                                 TYPE        VALUE
------------------------------------ ----------- ------------------------------
instance_name                        string      ora10g2

SQL> show parameter listener

NAME                                 TYPE        VALUE
------------------------------------ ----------- ------------------------------
local_listener                       string      local_lsnr_ora10g2
remote_listener                      string      remote_lsnr_ora10g

-->instrance ora10g2上的监听信息
SQL> ho ps -ef | grep lsnr
oracle   17372     1  0 11:00 ?        00:00:00 /u01/oracle/db/bin/tnslsnr LISTENER_NEW_BO2DBS -inherit
oracle   17502 24301  0 12:10 pts/0    00:00:00 /bin/bash -c ps -ef | grep lsnr
oracle   17504 17502  0 12:10 pts/0    00:00:00 grep lsnr

SQL> ho lsnrctl status LISTENER_NEW_BO2DBS
Listening Endpoints Summary...
  (DESCRIPTION=(ADDRESS=(PROTOCOL=tcp)(HOST=192.168.7.62)(PORT=1314)))
  (DESCRIPTION=(ADDRESS=(PROTOCOL=tcp)(HOST=192.168.7.52)(PORT=1314)))
Services Summary...
Service "ora10g" has 2 instance(s).
  Instance "ora10g1", status READY, has 1 handler(s) for this service...
  Instance "ora10g2", status READY, has 2 handler(s) for this service...
..........
  
-->由于只是测试基于客户端的load balance，因此移出remote_listener参数
SQL> alter system reset remote_listener scope=both sid='*';
alter system reset remote_listener scope=both sid='*'
*
ERROR at line 1:
ORA-32009: cannot reset the memory value for instance * from instance ora10g2

SQL> alter system reset remote_listener scope=spfile sid='*';

System altered.

SQL> ho srvctl stop database -d ora10g  -->关闭数据库ora10g

SQL> ho srvctl start database -d ora10g  -->启动数据库ora10g使得刚刚修改的remote_listener生效

SQL> ho lsnrctl status LISTENER_NEW_BO2DBS  -->此时可以看到只有ora10g2注册到监听器

Listening Endpoints Summary...
  (DESCRIPTION=(ADDRESS=(PROTOCOL=tcp)(HOST=192.168.7.62)(PORT=1314)))
  (DESCRIPTION=(ADDRESS=(PROTOCOL=tcp)(HOST=192.168.7.52)(PORT=1314)))
Services Summary...
Service "PLSExtProc" has 1 instance(s).
  Instance "PLSExtProc", status UNKNOWN, has 1 handler(s) for this service...
Service "ora10g" has 1 instance(s).
  Instance "ora10g2", status READY, has 1 handler(s) for this service...
..........  

oracle@bo2dbp:~> lsnrctl status LISTENER_NEW_BO2DBP  #同样在节点bo2dbp也只有ora10g1注册到监听器

Listening Endpoints Summary...
  (DESCRIPTION=(ADDRESS=(PROTOCOL=tcp)(HOST=192.168.7.61)(PORT=1314)))
  (DESCRIPTION=(ADDRESS=(PROTOCOL=tcp)(HOST=192.168.7.51)(PORT=1314)))
Services Summary...
Service "ora10g" has 1 instance(s).
  Instance "ora10g1", status READY, has 1 handler(s) for this service...
.........

3、客户端配置
SZDB:~ # ifconfig eth1 | grep "inet addr"|cut -d " " -f12|cut -d : -f2  #客户端主机的ip
192.168.7.2

SZDB:~ # su - oracle
oracle@SZDB:~> cat /etc/hosts     --#客户端主机添加了RAC上的两个虚拟节点的ip信息
192.168.7.61   bo2dbp-vip.2gotrade.com    bo2dbp-vip
192.168.7.62   bo2dbs-vip.2gotrade.com    bo2dbs-vip

oracle@SZDB:~> tail -12 $ORACLE_HOME/network/admin/tnsnames.ora

ORA10G =
  (DESCRIPTION =
    (ADDRESS = (PROTOCOL = TCP)(HOST = bo2dbp-vip.2gotrade.com)(PORT = 1314))
    (ADDRESS = (PROTOCOL = TCP)(HOST = bo2dbs-vip.2gotrade.com)(PORT = 1314))
    (LOAD_BALANCE = yes)
    (CONNECT_DATA =
      (SERVER = DEDICATED)
      (SERVICE_NAME = ora10g)
    )
  )

三、测试负载均衡(load balance)

1、启用load balance的测试
 
oracle@SZDB:~> more load_balance.sh 
#!/bin/bash
for i in {1..100}
do
echo $i
sqlplus -S system/oracle@ORA10G < ./load_balance.sh >load_bal.log
oracle@SZDB:~> head -20 load_bal.log
1

INSTANCE_NAME
----------------
ora10g2

2

INSTANCE_NAME
----------------
ora10g1

3

INSTANCE_NAME
----------------
ora10g2

4

oracle@SZDB:~> grep ora10g1 load_bal.log |wc -l
47
oracle@SZDB:~> grep ora10g2 load_bal.log |wc -l
53

从上面的log日志中可以看出启用客户端的负载均衡基本上使得从客户端发起连接的能够保持均衡。
  
2、未启用load balance的测试  
从客户端的tnsnames.ora中移出(LOAD_BALANCE = yes)选项，然后继续使用上面的脚本来测试
oracle@SZDB:~> grep ora10g1 no_load_bal.log |wc -l
100
oracle@SZDB:~> grep ora10g2 no_load_bal.log |wc -l
0

从上面的日志中可以看出当移出LOAD_BALANCE = yes项后，所有的用户连接请求都被定为到ora10g1，这是因为连接请求从tnsnames.ora中选择
列在ADDRESS项中排在第一行的位置。
下面我们关闭实例ora10g1，再来测试连接情形

oracle@bo2dbp:~> srvctl stop instance -d ora10g -i ora10g1

oracle@SZDB:~> ./load_balance.sh >no_load_bal_new.log
oracle@SZDB:~> grep ora10g1 no_load_bal_new.log |wc -l
0
oracle@SZDB:~> grep ora10g2 no_load_bal_new.log |wc -l
100  

由于实例ora10g1已经关闭，因此所有的连接请求都被分配到ora10g2。

 四、总结 
1、客户端的负载均衡配置较为简单，仅仅是在客户端的tnsnames.ora添加 LOAD_BALANCE = yes |on
2、其连接分配原则是根据tnsnames.ora中连接标识符下的ADDRESS列表随机选择来进行与服务器之间的连接
3、如果选择列表中的某个节点listener或instance不可用，则再从剩余的ADDRESS列表随机选择，直到成功为止
 
五、更多参考

有关基于用户管理的备份和备份恢复的概念请参考

    Oracle 冷备份

    Oracle 热备份

    Oracle 备份恢复概念

    Oracle 实例恢复

    Oracle 基于用户管理恢复的处理(详细描述了介质恢复及其处理)

    SYSTEM 表空间管理及备份恢复

    SYSAUX表空间管理及恢复

   Oracle 基于备份控制文件的恢复(unsing backup controlfile)

 

有关RMAN的备份恢复与管理请参考

    RMAN 概述及其体系结构

    RMAN 配置、监控与管理

    RMAN 备份详解

    RMAN 还原与恢复

    RMAN catalog 的创建和使用

    基于catalog 创建RMAN存储脚本

    基于catalog 的RMAN 备份与恢复

    RMAN 备份路径困惑

 

有关ORACLE体系结构请参考

    Oracle 表空间与数据文件

    Oracle 密码文件

    Oracle 参数文件

    Oracle 联机重做日志文件(ONLINE LOG FILE)

    Oracle 控制文件(CONTROLFILE)

    Oracle 归档日志

    Oracle 回滚(ROLLBACK)和撤销(UNDO)

    Oracle 数据库实例启动关闭过程

    Oracle 10g SGA 的自动化管理

    Oracle 实例和Oracle数据库(Oracle体系结构)