两个节点分别执行如下操作：

两个节点分别执行如下操作：
crsctl stop crs -f
cd /etc/oracle/
rm -rf scls_scr oprocd lastgasp o* setasmgid
vi  /etc/inittab（去掉ohas那一行[通常是最后一行]）
rm -f /etc/init.d/init.cssd 
rm -f /etc/init.d/init.crs 
rm -f /etc/init.d/init.crsd 
rm -f /etc/init.d/init.evmd 
rm -f /etc/rc2.d/K96init.crs 
rm -f /etc/rc2.d/S96init.crs 
rm -f /etc/rc3.d/K96init.crs 
rm -f /etc/rc3.d/S96init.crs 
rm -f /etc/rc5.d/K96init.crs 
rm -f /etc/rc5.d/S96init.crs 
rm -Rf /etc/oracle/scls_scr
rm -f /etc/inittab.crs 
cp /etc/inittab.orig /etc/inittab
rm -rf /etc/init.d/ohasd
rm -rf /etc/init.d/init.ohasd 
rm -rf /etc/oratab
rm -rf /etc/oraInst.loc 
rm -rf /var/tmp/.oracle
rm -rf /tmp/.oracle
rm -rf /u01/app
cd /tmp
rm -rf CVU_11.2.0.3.0_grid logs Logs OraInstall*
mkdir -p /u01/app/11.2.0.3/grid
mkdir -p /u01/app/grid
mkdir -p /u01/app/oracle
chown -R grid:oinstall /u01/app/11.2.0.3/grid
chown -R grid:oinstall /u01/app/grid
chown -R grid:oinstall /u01
mkdir -p /u01/app/oracle/product/11.2.0.3/dbhome_1
chown -R oracle:oinstall /u01/app/oracle
chown -R oracle:oinstall /u01/app/oracle/product/11.2.0.3/dbhome_1
---检查是否还有d.bin进程，如果kill掉：
ps -ef|grep d.bin
kill -9 pid
---检查ifconfig|grep 169.254如果有类似如下输出，需要重启eth1网卡：
eth1:1    Link encap:Ethernet  HWaddr 08:00:27:89:81:66  
          inet addr:169.254.159.3  Bcast:169.254.255.255  Mask:255.255.0.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1 
ifdown eth1
ifup eth1
ifconfig|grep 169.254
---两节点执行完后可以重新安装

PS: 文章来源于Lunar的oracle实验室

Purpose
This note provides steps to proceed from failed 11gR2 Grid Infrastructure installation. Similar procedure for 10gR1, 10gR2 or 11gR1 available in note 239998.1 and note 399482.1
If upgrade to 11.2 GI fails, refer to note 969254.1 for details. If 11.2 GI has been up and running and need to be deconfigured/deinstalled/reconfigured/reinstalled, refer to note 1377349.1 and note 1354258.1 
Scope
This document is intended for Oracle Clusterware/Oracle RAC Database Administrators and Oracle support engineers. 
Details
#################################################
A. OUI Failure
Case applies here is when runInstaller fails before user is prompted to execute rootupgrade.sh.
Refer to note 1056322.1 for solution details.
#################################################
B. GI Standalone Deconfigure and Reconfigure (Oracle Restart)
To deconfigure:
As root, execute “$GRID_HOME/crs/install/roothas.pl -deconfig -force -verbose”
If it fails, please disable GI, reboot the node and try the same command:
As root, execute “$GRID_HOME/bin/crsctl disable has”
As root, reboot the node; once the node comes backup, execute above deconfigure command again.
To reconfigure:
As root, execute “$GRID_HOME/root.sh”
#################################################
C. GI Cluster Deconfigure and Reconfigure
Identify cause of root.sh failure by reviewing logs in $GRID_HOME/cfgtoollogs/crsconfig and $GRID_HOME/log, once cause is identified and problem is fixed, deconfigure and reconfigure with steps below – keep in mind that you will need wait till each step finishes successfully before move to next one:
Step 0: For 11.2.0.2 and above, root.sh is restartable.
Once cause is identified and the problem is fixed, root.sh can be executed again on the failed node. If it succeeds, continue with your planned installation procedure; otherwise as root sequentially execute “$GRID_HOME/crs/install/rootcrs.pl -verbose -deconfig -force” and $GRID_HOME/root.sh on local node, if it succeeds, continue with your planned installation procedure, otherwise proceed to next step (Step 1) of the note.
Step 1: As root, run “$GRID_HOME/crs/install/rootcrs.pl -verbose -deconfig -force” on all nodes, except the last one.
Step 2: As root, run “$GRID_HOME/crs/install/rootcrs.pl -verbose -deconfig -force -lastnode” on last node. This command will zero out OCR, Voting Disk and the ASM diskgroup for OCR and Voting Disk
———————————————-
Note:
a. Step1 and 2 can be skipped on node(s) where root.sh haven’t been executed this time.
b. Step1 and 2 should remove checkpoint file. To verify:
ls -l $ORACLE_BASE/Clusterware/ckptGridHA_.xml
If it’s still there, please remove it manually with “rm” command on all nodes
c. If GPNP profile is different between nodes/setup, clean it up on all nodes as grid user
$ find <GRID_HOME>/gpnp/* -type f -exec rm -rf {} \;          
The profile needs to be cleaned up:
c1. If root.sh is executed concurrently – one should not execute root.sh on any other nodes before it finishes on first node.
c2. If network info, location of OCR or Voting Disk etc changed after Grid is installed – rare
———————————————-
tep 3: As root, run $GRID_HOME/root.sh on first node
Step 4: As root, run $GRID_HOME/root.sh on all other node(s), except last one.
Step 5: As root, run $GRID_HOME/root.sh on last node.
#################################################
D. Grid Infrastructure Deinstall
As grid user, execute “$GRID_HOME/deinstall/deinstall” and follow the instructions, for details, refer to the following documentation for your platform:
Oracle Grid Infrastructure Installation Guide
How to Modify or Deinstall Oracle Grid Infrastructure
If there’s any error, deconfig the failed GI with steps in Section B or C, and deinstall manually with note 1364419.1 
#################################################
E. OUI Plug-in Failure or OUI Window Closed While Running root.sh
Case applies here is OUI plug-in failure after root.sh has been executed successfully on all nodes or OUI window closed while running root.sh
Identify the cause of failed plug-in. Once root cause is identified and the issue is resolved, refer to note 1360798.1

PS: This article comes from MOS…

《Oracle 10g&11.1 Clusterware (CRS)安装失败如何卸载》请参考How to Proceed From a Failed 10g or 11.1 Oracle Clusterware(CRS)Installation