Oracle Server - Enterprise Edition - Version: 10.1.0.2 to 10.2.0.1
Information in this document applies to any platform.
Goal
Step by step approach to reinitialize the OCR and Vote Disks without performing a full reinstallation of the Oracle Clusterware
binaries for RAC on Windows platforms. These steps can be used in two basic scenarios: 1. to troubleshoot an installation
were has errored out OR 2. for a cluster that was running
fine, but the OCR and/or Vote disks have become damaged or unusable
Therefore,
please note that some of the listed steps (such as stopping the
clusterware services) may not be relevant to installation
troubleshooting.
Solution
1. Ensure all CRS resources are stopped using the "crs_stat
-t" utility to check the status and the "srvctl" utility to stop as
required. In particular before step 2 below make sure that you stop nodeapps.
etc. 2.
Stop all CRS services and set them to manual if required in Windows Services panel
Stop OracleCRService and set
to "Manual"on ALL nodes Stop OracleEVMService and set to "Manual"on ALL nodes Stop OracleCSService and
set to "Manual" on ALL nodes 3. Reformat partitions for OCR and voting disk files in preparation
for their recreation
- Go to a command prompt - navigate to CRS_HOME\bin run logpartformat /q for example if the raw partitions have the names 'votedsk1' and 'ocrcfg' :
Invoke
GuiOracleOBJManager.exe from
$CRS_HOME\bin and remove the assigned link names by placing a check
mark in the box preceding each link name then choosing 'Commit'
from the 'Options' menu.
5. Open a Windows command prompt and set TEMP and TMP to use c:\temp set
temp=c:\temp set tmp=c:\temp 6. From node1, find the crssetup.config.bat script, check that the contents
of this .bat file are correct for your environment, and execute this bat file eg, assuming your CRS_HOME is c:\oracle\product\10.2.0\crs cd
c:\oracle\product\10.2.0\crs\install
OR if you have already had 10.2.0.3 patchset in place:
c:\oracle\product\10.2.0\crs\install\prepatch10203\install dir
crssetup.config.bat
Note: At this point, you will have reinitialized
your OCR and Voting disk at version 10.2.0.1. If you are replacing a 10.2.0.1 OCR and Voting disk, proceed to step 7.
If you are replacing a patched version of your OCR and Voting disk, that is, you had previously a patchset to your CRS_HOME,
then you will need to run the patch102.bat script on each node before proceeding to step 8.
Patch102.bat can be found
in the folder $CRS_HOME\install. Please be aware that some of the commands that patch102.bat will run will be redundant, but
that this script needs to run to completion on each node to ensure your success herewith.
7. Take a copy of the OCR file as a starting point if required
ocrconfig -export c:\tmp\ocrdump_base.out
8. Ensure the "Oracle Process
Manager" service is set to "Automatic" on ALL nodes (Assuming that you have this implemented)
9.
Reboot all nodes in the cluster. Ensure that node1 restarts before the other nodes
10. Execute the vipca utility if it does not auto-execute Enter vip alias and IP as prompted in the GUI interface.
Once this is done, crs_stat -t should show vip/ons/gsd resource for all nodes.
11. Run oifcfg to configure public and cluster_interconnect eg: as oracle user,
srvctl add asm -n srvctl add database -d -o
srvctl add instance -d -i -n srvctl modify instance -d
-i -s <== repeat as needed to make the database instance resources dependent on their
corresponding ASM instance resources in the OCR An example of this last command: srvctl modify instance -d racdb -i racdb3 -s asm3
Please note, assuming that you are using ASM the asm_inst_name can only be +ASM1, +ASM2 and so on. Failure
in providing the correct name could cause an OCR corruption.
14. Once all resources are registered, start them via srvctl command eg: as oracle
srvctl start asm -n srvctl start instance -d -i
15. check crs_stat -t output, it should show all resources with ONLINE status on ALL nodes