I saw people having difficulty in cleaning up failed CRS configuration. I just thought to write about it based on my practical experience.
Oracle, 10g RAC, provided its own cluster-ware stack called CRS. The main file components of CRS are the Oracle Cluster Repository (OCR) and the Voting Disk. Due to the nature of these files, they must be placed on shared devices; either on cluster filesystems or shared raw devices.
Step by Step Instructions on How to perform CLEAN removal of failed Oracle10gRAC R2 CRS installation on Linux systems
Follow these steps in the order given. Failure to follow may cause problems during subsequent CRS installation attempts.
1. Login as Root on node 1.
2. # /etc/init.d/init.crs disable
3. # /etc/init.d/init.crs stop
4. Remove *crs*, *css* and *evm* entries from /etc/inittab file.
5. # rm –rf /etc/init.d/*crs*
6. # rm –rf /etc/init.d/*css*
7. # rm –rf /etc/init.d/*evm*
8. rm –f /etc/oraInst.loc file.
9. #Rm –rf /etc/oracle direcotry.
10. Remove all oracle instance related entries from /etc/oratab file.
11. Remove oraInventory and CRS Home. Anything under Oracle_Base should be removed.
12. # chown oracle:dba /dev/raw/raw[n]
14. Re-initialize all raw devices if using ASM otherwise make sure OCFS came back up(/etc/fstab file should be updated for auto mount of OFCS). For raw devices use “ dd if=/dev/zero of=/dev/raw/raw[n] bs=1M count=10” to initialize.
Follow the same sequence on each node. Once you’re done rebooting all other nodes, you are at a clean, fresh starting point for a new CRS installation.