Removing SF 5.0 for Oracle RAC
-
Stop VCS on the node on which you want to remove SF Oracle RAC.
# /opt/VRTSvcs/bin/hastop - local
-
Remove SF 5.0 for Oracle RAC.
See Removing SF Oracle RAC using the uninstallsfrac utility
-
Edit VCS configuration files on existing nodes.
See Editing VCS configuration files on existing nodes
Removing SF Oracle RAC using the uninstallsfrac utility
Run the uninstallsfrac script from any node in the cluster. Prior to invoking the uninstallsfrac script, you must take all service groups offline and shut down VCS.
The example procedure involves removing SF Oracle RAC from the node saturn.
To remove SF Oracle RAC using the uninstallsfrac utility
-
As superuser, start the uninstallation.
# cd /opt/VRTS/install
# ./uninstallsfrac
-
Decline to uninstall SF Oracle RAC from all systems.
VCS configuration files exist on this system with the following information:
Cluster Name: rac_cluster101
Cluster ID Number: 101
Systems: galaxy nebula saturn
Service Groups: cvm oradb1_grp
Do you want to uninstall SFRAC from these systems? [y,n,q] (y) n
Warning: Be sure to answer "n". Otherwise, the utility begins the process of uninstalling SF Oracle RAC from all systems.
-
Specify the name of the system on which you are uninstalling SF Oracle RAC.
Enter the system names separated by spaces on which to uninstall SFRAC:saturn
-
Review the output as the script checks for packages currently installed on your system. The utility also checks for dependencies between packages to determine the packages it can safely uninstall and in which order.
-
Confirm to uninstall SF Oracle RAC.
All SFRAC processes that are currently running will be stopped.
Are you sure you want to uninstall SFRAC? [y,n,q] (y)
-
Review the output as the script stops processes and drivers running on each node.
-
After the script uninstalls the packages, make a note of the location of the summary and log files that the uninstaller creates.
-
Reboot the node:
# /usr/sbin/shutdown -r now
Editing VCS configuration files on existing nodes
After running uninstallsfrac program, modify the configuration files on the existing nodes to remove references to the removed nodes.
Editing /etc/llthosts
On the each of the existing nodes, use vi or another editor to edit /etc/llthosts. Remove lines corresponding to the removed nodes; for example, if saturn is the node removed from the cluster, remove the line "2 saturn" from the file:
0 galaxy
1 nebula
2 saturn
Change to:
0 galaxy
1 nebula
Editing /etc/gabtab
In /etc/gabtab, change the command contained in the file to reflect the number of cluster nodes after the removal of a node:
/sbin/gabconfig -c -nN
N is the number of remaining nodes. For example, a file with two remaining nodes appears as:
/sbin/gabconfig -c -n2
Modifying the VCS configuration to remove the node
Modify the VCS configuration by editing /etc/VRTSvcs/conf/config/main.cf directly, using the VCS Cluster Manager, or using the command line, as illustrated in the following example. Refer to the Veritas Cluster Server User's Guide for details on configuring VCS.
Complete the procedure on one of the existing nodes.
To modify the VCS configuration
-
As superuser, enable write access to the configuration:
# haconf -makerw
-
Remove the node from the AutoStartList of the cvm service group by specifying the remaining nodes in the desired order:
# hagrp -modify cvm AutoStartList galaxy nebula
-
Remove the node from the SystemList of the cvm service group:
# hagrp -modify cvm SystemList -delete saturn
-
If you have the other service groups (such as the database service group or the ClusterService group) that have the removed node in their configuration, perform step 2 and step 3 for each of them.
-
Delete the node from the CVMCluster resource by removing it from the CVMNodeId attribute key list:
# hares -modify cvm_clus CVMNodeId -delete saturn
-
After deleting the removed node from all service groups in the configuration, delete the node from the cluster system list:
# hasys -delete saturn
-
Save the new configuration to disk:
# haconf -dump -makero
