Yes I know, it’s not supported and looks a bit crazy. But as I was using Linux 6 already with my RAC Database (see other blogs) and had to do some tests with an Oracle 10g database I tried to run it on the same environment. Another point is that I’m uncertain about the support for Linux 6. Due to the certification matrix it says: “Linux 5 Update 4 or greater” – but isn’t Linux 6 greater than Linux 5U4? Okay – nevertheless I’ll give it a trial:

To install any Oracle 10g database version on Linux 6 64 bit some additional 32 bit packages are required:

§ For the relink of some oracle executables:

glibc-devel-2.12-1.25.el6.i686.rpm

§ For the univeral installer to run properly:

libXau-1.0.5-1.el6.i686.rpm

libxcb-1.5-1.el6.i686.rpm

libX11-1.3-2.el6.i686.rpm

libXext-1.1-3.el6.i686.rpm

libXp-1.0.0-15.1.el6.i686.rpm

libICE-1.0.6-1.el6.i686.rpm

libuuid-2.17.2-12.el6.i686.rpm

libSM-1.1.0-7.1.el6.i686.rpm

libXt-1.0.7-1.el6.i686.rpm

libXi-1.3-3.el6.i686.rpm

libXtst-1.0.99.2-3.el6.i686.rpm

Luckily all those packages are available with the actual Linux 6 distribution.

Now we need to install 10.2.0.1 first because there is no full installable Oracle 10g version other than this. As expected the universal installer fails with the message that the OS is not supported. So it’s necessary to avoid any kind of preinstallation checks with the option “-IgnoreSysPrereqs”.

% runInstaller -IgnoreSysPrereqs

With this little trick the installation works very smoothly. Both nodes were found and the software was installed on both nodes automatically.

Now it’s time to install patch 10.2.0.5 and – oh wonder – the universal installer accepts the linux 6 and runs without any complains. And as the Oracle documentation says the installer accepts the 11g clusterware as well and all software is installed automatically on both nodes.

Create Database

Before we can start dbca to install the database there is one restriction which has to be resolved: to run a pre 11g database with grid infrastructure 11g the nodes have to be pinned to the crs (see “RAC: Frequently Asked Questions [ID 220970.1]”). The command has to be executed as root in the grid environment.

# olsnodes -t -n

asterix 1 Unpinned

obelix 2 Unpinned

# crsctl pin css -n asterix obelix

CRS-4664: Node asterix successfully pinned.

CRS-4664: Node obelix successfully pinned.

After pinning the nodes the database configuration assistant (dbca) recognizes that this is a RAC installation and asks for the nodes as well as for the ASM infrastructure. The whole configuration process runs without any problem.

Unfortunately it looks as if the scan listener cannot be used with Oracle 10g. So for the client connection the vip addresses seem to be the only way to use load balancing and failover (maybe someone can give me some hints how to make it work with the scan addresses).

The databases (10g and 11g) are both running fine in my test environment. If you want to share some experiences or give me some hints please comment with this blog.