一次给用户配置RAC的LOAD BALANCE的经历。

最新推荐文章于 2024-10-17 18:58:27 发布

cnos50350

最新推荐文章于 2024-10-17 18:58:27 发布

阅读量86

点赞数

文章标签：数据库网络

配置RAC的LOAD BALANCE其实是一件很简单的工作，然而本来信心满满给用户配置，却引起用户数据库不能登陆的错误，事情经过到底是怎样的呢？[@more@]

一、配置过程：

1、先配置两个节点的tnsnames.ora文件，包含下面的内容：

failover =
(DESCRIPTION =
(enable=broken)
(LOAD_BALANCE = yes)
(ADDRESS = (PROTOCOL = TCP)(HOST = jxgs550a)(PORT = 1521))
(ADDRESS = (PROTOCOL = TCP)(HOST = jxgs550b)(PORT = 1521))
(CONNECT_DATA =
(SERVICE_NAME = jxdc.jxgs.gov)
(failover_mode=(type=select)(method=basic))
)
)

LISTENER_RACa =
(DESCRIPTION =
(ADDRESS = (PROTOCOL = TCP)(HOST = jxgs550a)(PORT = 1521))
)

LISTENER_RACb =
(DESCRIPTION =
(ADDRESS = (PROTOCOL = TCP)(HOST = jxgs550b)(PORT = 1521))
)

LISTENER_RAC =
(DESCRIPTION =
(ADDRESS = (PROTOCOL = TCP)(HOST = jxgs550a)(PORT = 1521))
(ADDRESS = (PROTOCOL = TCP)(HOST = jxgs550b)(PORT = 1521))
)

2、执行下面的命令，修改local_listener和remote_listener的值：
alter system set local_listener='LISTENER_RACa' scope=both sid='jxdc1';
alter system set remote_listener='LISTENER_RAC' scope=both sid='jxdc1';
alter system set local_listener='LISTENER_RACb' scope=both sid='jxdc2';
alter system set remote_listener='LISTENER_RAC' scope=both sid='jxdc2';
3、分别查看两边监听的状态：
在节点2：
$ lsnrctl service

LSNRCTL for IBM/AIX RISC System/6000: Version 9.2.0.8.0 - Production on 20-JUN-2007 12:19:10

Connecting to (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=jxgs550b)(PORT=1521)))
Services Summary...
Service "PLSExtProc" has 1 instance(s).
Instance "PLSExtProc", status UNKNOWN, has 1 handler(s) for this service...
Handler(s):
"DEDICATED" established:0 refused:0
LOCAL SERVER
Service "jxdc.jxgs.gov" has 2 instance(s).
Instance "jxdc1", status READY, has 1 handler(s) for this service...
Handler(s):
"DEDICATED" established:69 refused:0 state:ready
REMOTE SERVER
(ADDRESS=(PROTOCOL=TCP)(HOST=jxgs550a)(PORT=1521))
Instance "jxdc2", status READY, has 2 handler(s) for this service...
Handler(s):
"DEDICATED" established:0 refused:0 state:ready
REMOTE SERVER
(ADDRESS=(PROTOCOL=TCP)(HOST=jxgs550b)(PORT=1521))
"DEDICATED" established:43 refused:0 state:ready
LOCAL SERVER
在节点1：
$ lsnrctl service

LSNRCTL for IBM/AIX RISC System/6000: Version 9.2.0.8.0 - Production on 20-JUN-2007 12:18:02

Connecting to (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=jxgs550a)(PORT=1521)))
Services Summary...
Service "PLSExtProc" has 1 instance(s).
Instance "PLSExtProc", status UNKNOWN, has 1 handler(s) for this service...
Handler(s):
"DEDICATED" established:0 refused:0
LOCAL SERVER
Service "jxdc.jxgs.gov" has 2 instance(s).
Instance "jxdc1", status READY, has 2 handler(s) for this service...
Handler(s):
"DEDICATED" established:0 refused:0 state:ready
REMOTE SERVER
(ADDRESS=(PROTOCOL=TCP)(HOST=jxgs550a)(PORT=1521))
"DEDICATED" established:0 refused:0 state:ready
LOCAL SERVER
Instance "jxdc2", status READY, has 1 handler(s) for this service...
Handler(s):
"DEDICATED" established:4 refused:0 state:ready
REMOTE SERVER
(ADDRESS=(PROTOCOL=TCP)(HOST=jxgs550b)(PORT=1521))
4、从其它客户端开一个窗口进行测试，发现有时候可以连接数据库，有时候不能，不能的时候提示不能找到主机。
5、直接在服务器上进行连接没有任何问题。猜想是服务器上在/etc/hosts中定义了名称解析，在客户端则没有，于是将两接点上的tnsnames.ora中机器名称全部修改为IP地址，客户端有时能连接有时不能的问题得到了解决。
6、对这个问题，我个人的看法是当客户端试图连接某个节点的实例时候，监听根据负载均衡的原则决定客户端与哪个节点进行连接，并根据本机上tnsnames.ora将节点信息传递给客户端，客户端再根据相关信息进行连接，因为客户端无法解析机器名称，因此连接不上。
7、设置完毕后，通过语句select inst_id, count(*) from gv$session group by inst_id进行监测，发现负载是均衡的，一度达到实例1上130个会话，实例2上129个会话这样的均衡状态。