Performing a rollback restore in an SFRAC environment
The following procedure describes the manual steps that are needed to restore volumes and file systems by using the snapshot rollback method in an SFRAC environment.
A typical host deployment for running NetBackup for Oracle in an SFRAC environment is as follows: Host A and Host B are in the cluster and Host C is used as an alternate client. The instant recovery snapshot is taken using the Oracle policy and the Hardware Snapshot FIM (frozen image method).
The application I/O stack is built upon the hardware array of VxVM (CVM) and VxFS (CFS).
The Veritas Cluster Server (VCS) controls the Oracle RAC database and other required essential resources such as shared storage. VCS defines and manages its resources as a single unit called a service group. A service group contains all the necessary components and resources of an application.
The following are entities in the VCS that monitor the application I/O stack:
CFSMount contains the mount points (cfs) where data files, archive logs, and control files are stored.
CVMVolDg contains all the Volume Groups (cvm) configured on top of the various array LUNs that participate in the hardware snapshot.
The Database Resource Group contains the database instance and assists in failover.
To perform a rollback restore in the SFRAC environment
- Ensure that you previously created a virtual IP for the clustered node. Point that virtual IP to the primary node of the cluster.
- On all the clustered nodes, take the VCS database service group (Oracle, CFSMount, and CVMVolDg resources) offline by using the following command:
# hagrp -offline <DB_Service_Group> -any
- Freeze the database service group.
# hagrp -freeze <DB_Service_Group>
- Mount the CFSMount points manually outside VCS on the primary node. This action helps when you start the database in mount state.
# mount -F vxfs -o cluster <mntPt>
- Start the database with mount option on the clustered primary node using one of these options:
Option 1:
# sqlplus /as sysdba # startup mount;
Option 2:
# sqlplus /as sysbackup # startup mount;
- Run the rollback restore operation from the RMAN script or the client GUI. A sample RMAN script file (
hot_database_backup_proc) is located in the following directory path:/usr/openv/netbackup/ext/db_ext/oracle/samples/rman
To perform PIT rollback, add the following parameter to the script:
NB_ORA_PC_RESTORE=rollback
- Unmount the CFS on the primary node.
# umount <mntPt>
- Unfreeze the VCS database service group.
# hagrp -unfreeze <DB_Service_Group>
- On all the clustered nodes, take the VCS database service group (Oracle, CFSMount, and CVMVolDg resources) back online:
# hagrp -online <DB_Service_Group> -any