Clusterware and Database Upgrades - Expert Oracle RAC 12c - page 395

Use the [-force] option in case if you encounter any issues with the preceding command due to the failed
upgrade.
The command will stop all currently running resources on the server and shut down the 12c cluster stack
subsequently.
After successfully executing the preceding command across all nodes in the cluster, execute the following
command on the first node:

$GRID_HOME/crs/install/rootcrs.pl -downgrade -lastnode

This process will downgrade the OCR and set to the previous release.
As GI installation owner, run the following commands from the 12c $GRID_HOME/oui/bin location on every
successfully upgraded node to complete the downgrade procedure:

./runInstaller -nowait -waitforcompletion -ignoreSysPrereqs -updateNodeList -silent CRS=false
ORACLE_HOME=/u01/app/12.1.0/grid

./runInstaller -nowait -waitforcompletion -ignoreSysPrereqs -updateNodeList -silent CRS=false
ORACLE_HOME=/u01/app/11.2.0/grid

■ You need to provide the 12c grid home location in the first command, and in the second command, you need to
provide the previous release grid home location.

Note

For Oracle version 11gR2 or later, after successfully executing the preceding steps, as the root user, bring up the
previous release cluster stack using the following command on each node in the cluster:

$/u01/app/11.2.0/grid/bin/crsctl start crs

For pre-Oracle 11gR2 versions, as root user, execute the root.sh script manually from the earlier Cluster release
on each node in sequence to bring up the previous cluster stack.
Oracle 12c GI supports downgrade option to Oracle 11gr2 and Oracle 11gR1.

■ Keep in mind that any pre-/post-upgrade configuration modifications performed will be removed and
unrecoverable after the downgrade.

Note

Forcing Upgrade—When Some Nodes Are Inaccessible
Pre-Oracle 12c, if some nodes are not accessible or become unreachable before or during the upgrade, then
the Clusterware upgrade won't be completed due to missing/inactive nodes in the cluster. In contrast, with the
introduction of the -force option with the rootupgrade.sh script, you can perform a forced upgrade despite
missing/inactive nodes in the cluster. Therefore, when you query active crs version to confirm the upgraded, you will
see that the crs active version is set to the new Cluster version in contrast to the pre-Oracle 12c releases.
Here is an example to run the rootupgrade.sh script with the -force option:

$ /u01/app/12.1.0/grid/rootupgrade -force