To begin, switch to the root user and change
directories to $ORA_CRS_HOME/install. Run the rootdelete.sh script
as shown in Figure G.1. It may appear to hang at a message that
reads in part ?Aborted? but allow it a couple of minutes to
complete. This script should be run from each node.
Next, run the rootdeinstall.sh script as shown
in Figure G.2. Allow this script to finish on one node before
beginning it on the next.
If you still have a copy of the old Oracle 9i
srvconfig file, you may optionally choose to use its configuration
as a starting point. To do so, follow the steps in Figure G.3. The
dd command, which requires a few minutes to complete, should be
run from only one node.
At this point, run $ORA_CRS_HOME/root.shfrom each node. Next, run $ORACLE_HOME/root.sh from
each node. Finally, run the 10g version of netca, as oracle, to
re-register the listeners with CRS.
10g Release 2
For 10g Release 2, the instructions are
similar, but be careful. Do not run the scripts $ORA_CRS_HOME/install/rootdelete.sh
first and then $ORA_CRS_HOME/install/rootdeinstall.sh second, as
was done for Release 1. These scripts cause a problem for vipca.
They make it so that vipca cannot create a new
vip, gsd, and ons! Figure G.4
shows the error I have experienced both on the command line and as
seen through the vipca tool. The only fix I have found for this
error is to revert to an image of Release 1 software, then to
reinstall Release 2.
To rebuild the 10g Release 2 CRS, run the
following scripts. Each script must be completed on each node
before continuing to the next script.
$db10g1/root.sh ### will launch vipca on first node
gedit /etc/oratab ### fix the entry for * to 10g-R2 home
$db10g2/bin/netca ### run as the oracle user, not root
Clearly, this is a hack (the files in the
release1 directory were copied from the previous release).
Although it works, it is doubtful that Oracle Support would agree