Planning to build an Oracle RAC system

Edit online

When planning to build an Oracle RAC cluster, you must first choose an Oracle RAC version, download it and then set up the home directories properly.

Versions of Oracle RAC

Oracle RAC 10g Release 2 version 10.2.0.2 is not certified by Oracle for Linux® on IBM® System z®. However, versions 10.2.0.3 and 10.2.0.4 are certified by Oracle for Linux on IBM System z. To install versions 10.2.0.3 or 10.2.0.4, it is necessary to begin by installing version 10.2.0.2 and then apply the upgrade as a patch.

The download of Oracle Database 10g Release 2 (10.2.0.2) for Linux on IBM System z is available from the Oracle Technical Network (OTN) at:

https://www.oracle.com/technetwork/database/enterprise-edition/downloads/index.html

Complete these steps to perform the download from this Web site:
  1. In the Downloads tab, select 11g Enterprise/Standard Editions.
  2. Scroll down to locate the 10g Release2 Standard and Enterprise Edition section.
  3. Make a selection for Linux running on IBM System z (z/Linux).
  4. Read and accept the License Agreement.
  5. Select Download the Complete Files.
  6. On the Download page, select these two files:
    • 10202_zlinux_database.zip
    • 10202_zlinux_clusterware.zip

The database package also contains ASM.

For the upgrade to version 10.2.0.4, download patchset p6810189_10204_LINUX-zSer/zip from a different location: Oracle's "My Oracle Support" formerly called Metalink. The patchset is named 10.2.0.4 Patch Set for Oracle Database Server for platform IBM:Linux on System z, and it contains clusterware and ASM as well as RDBMS, the database.

Oracle releases bundled patches (called CPUs - Quarterly Critical Patch Updates) on a regular schedule. They are roll up patches, which incorporate the prior updates to the release. For example, the fourth CPU to Oracle RAC for Linux on IBM System z takes Oracle RAC version 10.2.0.4.1 to version 10.2.0.4.4. CPUs for Linux on IBM System z can be applied with OPatch, a patch applicator that is part of Oracle RAC 10g Release 2.

When searching OTN and other Oracle sources for products applicable to IBM System z, it may be necessary to use search strings that are not the current IBM names, such as zSeries, z/Linux, s/390, or z390.

Note: Although you can download Oracle RAC and its patches and associated software at any time, do not install these products until after you have properly prepared the target Linux systems. See Preparing Linux for the installation of Oracle RAC.

Oracle HOME directories and multiple Oracle databases

The test system was built with all the Oracle components (ASM, CRS, and RDBMS) at the same release level. However, the Oracle RAC model can be built with additional databases at different release levels on the same server, as long as the clusterware (CRS) on any server is at the same or a later release level than the level of any of the other components, such as RDBMS or ASM. The additional databases can be single instance or in a cluster. Each database can be upgraded individually, while support continues for databases at an earlier level.

Operating in an Oracle environment with many possibilities is accomplished by maintaining sets of environmental variables such as $HOME, which would have a related $SID to identify a particular database, and $PATH, $LIBPATH and other environmental variables as needed. As an example, $HOME_CRS is set to /product/crs, which is the location in this installation of the Oracle RAC 10.2.0.4 CRS binaries, CRS libraries, CRS logs, and CRS parameters.

The database administrator's files, such as scripts and parameters for creating and maintaining databases, is located outside of the Oracle HOMEs. The actual data is located in the shared storage environment such as ASM.

In an Oracle clustered environment, each database can be upgraded and expanded individually while support continues for databases at earlier releases.

Oracle server processes – shared versus dedicated

After the clustered database is installed, one of the decisions to make is whether to use dedicated server processes or shared server processes. Dedicated server processes choose a one to one relationship between request and server. With the other choice, shared server processes, the requests are queued to match processes from any of the servers being served by dispatchers.

The efficiency of each approach is related to the type of activity that takes place. With the dedicated server setup, a client request to the database connects to a server process and holds it until the request is completed. In this study, this process provided very good throughput results. This configuration might use a large amount of memory with an increasing amount of client connections, because each server process requires a certain amount of memory. The shared server process reuses the same server process for multiple client processes, which goes the opposite way, reduced memory usage with possibly lower throughput.

When the workload is determined by many (1000 or more) connected users who spend significant time between actions, such as online shoppers, the shared server process setup might be more efficient. For workloads with a concentrated high utilization on each connection, for example when the Oracle database is the backend from an application server or behind a connection concentrator, dedicated server processes might be more appropriate.

Details on how to configure shared or dedicated servers are in the section Setting up the listeners.