oracle RAC 9.2.0.4 for linux x86升级到9.2.0.8

oracle RAC 9.2.0.4升级到9.2.0.8
操作系统平台：Red Hat Enterprise Linux AS release 3
准备工作：
1. Check Tablespace Sizes and Set Parameter Values

2. Check SYSTEM Tablespace Size ensure that there is at least 10 MB of free
space allocated to the SYSTEM tablespace

3. Check XDB Tablespace Size For Oracle RAC installations, ensure that there
is at least 50 MB of free space allocated to the XDB tablespace

4. Set the SHARED_POOL_SIZE and JAVA_POOL_SIZE Initialization Parameters
SQL> ALTER SYSTEM SET SHARED_POOL_SIZE='150M' SCOPE=spfile;
SQL> ALTER SYSTEM SET JAVA_POOL_SIZE='150M' SCOPE=spfile;


为了数据的安全，做之前先备份数据：
1.1 逻辑导出数据库
ORACLE_HOME=/opt/oracle/product/9.2.0.4
ORACLE_SID=zfj1
NLS_LANG=american_america.zhs16gbk
export ORACLE_HOME ORACLE_SID NLS_LANG
$ORACLE_HOME/bin/exp rman/rmanzfj full=y buffer=4096000 compress=n file=/opt/ora
cle/backup/dmp/exp_zfj.dmp log=/opt/oracle/backup/dmp/exp_zfj.log

1.2 物理备份数据库：

[oracle@svr1 sh]$ rman target / catalog rman/rman@catadb（一个节点执行即可）

Recovery Manager: Release 9.2.0.8.0 - Production

Copyright (c) 1995, 2002, Oracle Corporation.All rights reserved.

connected to target database: ZFJ (DBID=3130691682)
connected to recovery catalog database
RMAN> run {
allocate channel c1 type disk;
backup
incremental level 0
skip inaccessible
tag zfj_level0
filesperset 15
format '/opt/oracle/dump/rmanzfj/zfj_DBLevel0%s_%p_%t'
release channel c1;
}
1.3 备份数据库安装文件（需要在两个节点都执行）
cd /opt/oracle
tar cvf zfj1_oracle.tar --preserve --exclude data_ocfs --exclude data_ocfs2
(备注：把 oralce安装目录文件打包备份，--preserve参数将保持文件属性不变
--exclude参数排除data_ocfs、data_ocfs2两个目录不打包到zfj1_oracle.tar）

1.4 配置linux操作系统启动到运行级别5，并可以用xmanger图形链接
1.4.1
vi /etc/inittab id:5:initdefault:
1.4.2
cd /etc/X11/gdm
vi /etc/gdm.conf
[xdmcp]
# Distributions: Ship with this off.It is never a safe thing to leave
# out on the net.Setting up /etc/hosts.allow and /etc/hosts.deny to

only
# allow local access is another alternative but not the safest.
# Firewalling port 177 is the safest if you wish to have xdmcp on.
# Read the manual for more notes on the security of XDMCP.
Enable=true

#PingIntervalSeconds=15
# The port.177 is the standard port so better keep it that way
Port=177
cd /etc/X11/xdm
vi Xaccess 取消如下行前面的#和*号
any host can get a login window
vixdm-config 在如下行前面加一个 ！
!DisplayManager.requestPort: 0

1.5 升级oracle 9i cluster manager(rac环境只需要在一个节点执行）
1.5.1 切换到oracle用户下 su - oracle

1.5.2 做如下准备工作，如果如下环境变量已经设置可以不予理睬
$ DISPLAY=local_host:0.0;export DISPLAY（xmanger 3.0好像自动会识别把图形切
换到用户终端，不需要设置）

1.5.3 查看oracle environment variable is set correctly

1.5.4 set the LD_LIBRARY_PATH environment variable to $ORACLE_HOME/lib
$ LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$ORACLE_HOME/lib; export
LD_LIBRARY_PATH

1.5.5 把oracle升级包上传到服务器上，并进入升级包目录Disk1 例如：/var/src/Disk1
./oracm/runInstaller(升级cluster manager)
备注：此时图形界面会出现，指定原始oracle home和path,此过程中需要输入
cluster节点和cluster私有节点（心跳）名，输入后next就安装成功了

1.5.6 $cd ORACLE_HOME/oui/bin(where oracle_home is the Oracle home where you
installed Oracle Universal Installer 10.1.0.5)
$./runInstaller -updateNodeList -noClusterEnabled -local ORACLE_home=
$ORACLE_HOME CLUSTER_NODES=svr1,svr2(svr1和svr2是rac的两个节点）

备注：RAC环境只需要在一个节点执行，升级程序自动会将另外一个节点的cluster

manager程序文件也进行升级

1.6 交互模式打oracle 9i数据升级补丁（此过程要确保cluster manager是启动的）


1.6.1 $cd ORACLE_HOME/oui/bin(where oracle_home is the Oracle home where you

installed Oracle Universal Installer 10.1.0.5)
$./runInstaller -updateNodeList -noClusterEnabled -local

ORACLE_HOME=$ORACLE_HOME CLUSTER_NODES=srv1,srv2

1.6.2 如果cluster manager没有启动，执行启动cluste manager
#/opt/oracle/product/9.2.0.4/oracm
#source ~oracle/.bash_profile (进入root用户下使用oracle环境变量，然后

执行cluster启动脚本）
#/opt/oracle/product/9.2.0.4/oracm/bin
#./ocmstart.sh

1.6.3 Enter following commands to start Oracle Universal Installer, where

patchset_directory is the directory you unzipped the patch set

software
$ patchset_directory/Disk1(patchset_directory实际上就是我上传补丁的

那个目录)
$ ./runInstaller
1.6.4 在oui图形界面中，指定oracle_home name(如果oracle_home名指定正确，自

动会找到以前oracle安装的路径)和路径，然后next,当提示用root用户运行

$ORACLE_HOME/root.sh脚本时，按照提示运行root.sh脚本，然后等安装完成

后退出oracle universal installer即可；

备注：在升级oracle 9i rac环境数据库的时候，一定要确认升级ok的oracle cluster

manage启动起来，否则会报错，不会启懂oracle universal installer图形界

面

1.7 upgrade the database
1.7.1 log in as the oracle software owner(typically oracle) su- oracle

1.7.2 start the Global Services Daemon(GSD) on each node of the cluster

as follows:(备注：在每个节点都要启动gsd服务，如下命令在每个节点执行
$gsdctl start
$lsnrctl start

1.7.3 start the database,where db_name is the database name:
$srvctl start database -d zfj(备注：如果gsd服务没有启动，srvctl就不

能使用，只能一个节点一个节点的使用sqlplus启动数据库，gsd服务的作用

就是一个命令就可以把所有节点的实例启动起来)、

1.7.4 Set the CLUSTER_DATABASE initialization parameter to FALSE: (作用就

是让clust manager暂时不起作用）
SQL> ALTER SYSTEM SET CLUSTER_DATABASE=FALSE SCOPE=spfile;

1.7.5 shutdown database
$ srvctl stop database -d zfj

1.8 Enter the following SQL*Plus commands:
$sqlplus "/ as sysdba"
SQL> STARTUP MIGRATE
SQL> SPOOL patch.log
SQL> @?/rdbms/admin/catpatch.sql
SQL> SPOOL OFF

1.9 查看patch.log文件是否有报错

2.0 restart the database;
SQL> SHUTDOWNimmediate;
SQL> STARTUP
2.1 Run the utlrp.sql script to recompile all invalid PL/SQL packages now

instead of when the packages are accessed for the first time. This step is

optional but recommended.

SQL> @?/rdbms/admin/utlrp.sql

2.2 Set the CLUSTER_DATABASE initialization parameter to TRUE:
SQL> ALTER SYSTEM SET CLUSTER_DATABASE=TRUE SCOPE=spfile; (1.7.4）

2.3 restart th database;
SQL> SHUTDOWNimmediate;
SQL> EXIT
$ srvctl start database -d zfj(两个节点的实例都启动)

2.4 If you are using the Oracle Recovery Manager catalog, enter the following

command:(由于我们有一个专门的catlog数据库，而且是oracle 10G的，版本比

9.2.0.8高，所以不需要升级，升级方法如下：)

$ rman catalog username/password@catadb(catadb是我的catalog数据库)
RMAN> UPGRADE CATALOG;

FROM:http://linux.chinaunix.net/bbs/archiver/tid-1038418.html