Oracle 10gR2 build 2-node RAC record

To better understand the differences between the 10gR2 RAC and the latest version, we will continue to build a 10gR2 RAC on the platform with the Environment configured in 11gR2.

Before installation, you must modify a Kernel Parameter of RHEL5.5 to allow clusterware to be installed on RHEL5. Otherwise, an error indicating that the OS Kernel does not meet the installation requirements and cannot be continued will be reported.

1. Modify RHEL5.5 Kernel Parameters

2. Install and test whether the packages required by RAC are correctly installed.

-------------- Split line --------------

Oracle11g R2 (11.2.0.1) is a relatively new version. The RAC management component is no longer clusterware but a brand new grid infrastructure. In addition, this grid is mainly used for Storage Management Using ASM, we do not recommend that you continue using bare devices. This makes the deployment of OCR and voting disks more flexible and automated.

The virtual platform is virtualbox4.0.4, the host operating system platform Ubuntu10.10, and the virtual operating system platform RHEL5.5.

1. download and install the required software packages, such as libaio, unixODBC, and sysstat.

2. Set User Groups on nodes 1 and 2: oinstall, dba, role, asmadmin, asmdba, and asmoper.

3. Create the software installation directory/opt/oracle,/opt/grid (custom), and set the directory permissions.

4. Set User Environment Variables ~ /. Bash_profle

5. Install the rlwrap and readline toolkit ~ /. Bashrc.

6. Set the memory and virtual memory of the virtual platform RHEL5.5, which shall not be less than 1.5 GB as required.

7. Configure the ssh communication trust of the node.

8. Configure the DNS server Service on each node. Forward and reverse interpretations.

9. Use vboxmanager to create a shared disk.

10. When creating an asm disk, you must create an asm disk group during the installation process. At least three disks are required.

11. Install grid infrastructure on any node.

12. After the installation is complete, test whether the crs service of the two nodes can be shut down normally.

13. Start to install the database software on a node. If you select rac mode, oracle will install the software on other nodes in Push mode without human intervention.

14. Use dbca or run the script to create a database on one of the nodes, test whether the database instance can run normally, and modify the parameter file.

15. When processes such as asm instance and crs, cssd, and cstd can be enabled normally, use dbca to create a database on one of the nodes.

16. 11gR2 manages each RAC node in two ways: 1. admin-manager 2. policy-manager. The former requires you to set ORACLE_SID for instances on each node, while the latter requires you to manage instances of all nodes in a 'pooled 'without setting a fixed ORACLE_SID, the pool server automatically generates a SID for each node. For example, if GDN is set to orcl, the SID of each node is set to something similar: orcl_1, orcl_2 ,........., automated and user-friendly. I think the first mode exists to be compatible with the usage habits of 9i 10g users, while the policy mode is a brand new attempt.

17. When 10 Gb of RAC is disabled, the ASM instance needs to be closed independently. In 11gR2 RAC, the ASM instance has been integrated with the crs, even if the database is closed, if you disable diskgroup, you cannot directly close the ASM instance. Using the-force option of srvctl will cause loss of asm_diskstring, so it is best not to use it. That is to say, to close the ASM instance, it should also be closed together with the crs, And the ASM instance will also be started when the crs is started. After countless attempts, we found that 11gR2 RAC is significantly different from 10g in many management concepts.

18. Execute the database and instance startup operations on the node. Instead of using sqlplus, use srvctl directly. If you use sqlplus/as sysdba to log on, use startup pfile = '..... the method will report the error of setting the instance_number parameter. It may be because I have used the policy-manager management mode and do not need to set this parameter (after all, it is a large pool at the beginning, when no application is accessed, there will be no concept of the number of instances). Instead, you can directly start the instance of the node by using srvctl start instance-d orcl-I orcl_1, the command to start the database has not changed much: srvctl start database-d orcl;

19. 11gR2 note the startup sequence: CRS | ASM instance-> database-> node instance

20. Close sequence: node instance> database> diskgroup> CRS | ASM INSTANCE

Important:

If the installation is interrupted or the disk is lost due to a problem during the installation process, use dd to format the asm disk and run $ ORACLE_CRS_HOME/crs/install/deinstall/rootcrs. after pl-deconfig-force-verbose is configured, run the root command again. sh for secondary configuration.