Preparing the Standby Database Host : Oracle RAC One Node and Oracle Data Guard

Advertisements

As stated previously, here we are going to use a shared ACFS filesystem -- mounted on all nodes in the same cluster -- to store the diagnostic destination directory, the data pump directory, and other information. The directory already exists for the primary database but not the standby. As a first step the ASM Cluster File System has to be created on the DR cluster.

Begin by creating a new volume for the ASM Cluster File System as the grid infrastructure owner. (In environments without separation of duties this command would most likely be executed as the oracle user.) Note that the ORACLE_SID is set to the first ASM instance.

[grid@drnode1 ~] $ asmcmd volcreate -G ACFSDG -s 1G RONDG_V1

[grid@drnode1 ~] $ asmcmd volinfo -G ACFSDG RONDG_V1
Diskgroup Name: ACFSDG

         Volume Name: RONDG_V1
         Volume Device: /dev/asm/RONDG_v1-329
         State: ENABLED
         Size (MB): 1024
         Resize Unit (MB): 256
         Redundancy: UNPROT
         Stripe Columns: 4
         Stripe Width (K): 128
         Usage:
         Mountpath:

[grid@drnode1 ~] $

In the above example the ASM dynamic volume has been created on a dedicated disk group, called ACFSDG. Although convenient, it is not necessary to use a separate disk group for ADVM/ACFS. The reason in this case is simple:  while the database is located on "premium" classed storage, the ACFS volume is not, hence the use of a different disk group.

The above command created a 1GB ASM dynamic volume, called RONDG_V1 for later use by ACFS in disk group ACFSDG. Next the volume has to be formatted. You need to use the mkfs utility as root to do so, as shown in this example:

[root@drnode1 ~]# mkfs -t acfs /dev/asm/RONDG_v1-329
mkfs.acfs: version                   = 11.2.0.2.0
mkfs.acfs: on-disk version           = 39.0
mkfs.acfs: volume                    = /dev/asm/RONDG_v1-329
mkfs.acfs: volume size               = 1073741824
mkfs.acfs: Format complete.
[root@drnode1 ~]#

After which the volume has to be registered with the ACFS registry. (Alternatively it could have been created as a cluster resource with a dependency on the RONDG database.)

[root@drnode1 ~]# /sbin/acfsutil registry -a -n
> drnode1,drnode2 
> /dev/asm/RONDG_v1-329 /data/oracle/RONDG/
acfsutil registry: mount point /data/oracle/RONDG successfully added to Oracle Registry

Translated into English, we have instructed the clusterwide mount registry to add a new filesystem, called /dev/asm/RONDG_v1-329, and to mount it to /data/oracle/RONDG on all cluster nodes.

Think of the ACFS registry as an extended, portable, cluster wide /etc/fstab file. With this in use, it is not necessary or encouraged to add a mount point into each cluster node's fstab! You may have to mount the cluster filesystem manually once after its initial creation.

Let's walk through this with an example. First, create the mount point on all cluster nodes, then mount the cluster file system. (Again, the ACFS registry will perform this step at the next server reboot!) In the next step the filesystem permissions are relaxed to allow the oracle user to read, write, and change into the directory. The shell script shown below makes use of user equivalence and the fact that the oracle user has suitable entries in /etc/sudoers. You would not normally find this in production environments � in which case you need to ask your system administrator to perform this step for you.

[oracle@drnode1 tmp: RON]$ mkdir /data/oracle/RONDG/
[oracle@drnode1 tmp: RON]$ for i in 2 3 4
> do
> ssh drnode$i sudo mkdir �p /data/oracle/RONDG/
> done

[oracle@drnode1 tmp: RON]$ for i in 2 3 4 ; do
>  ssh drnode$i sudo mount -t acfs /dev/asm/RONDG_v1-329
>  /data/oracle/RONDG/;
>  mount;
>  done

[oracle@drnode1 oracle]# for i in 2 3 4
> do
> ssh drnode$i sudo chown -R oracle:oinstall /data/oracle/RONDG
> done

In a similar way I created the needed subdirectories for the administrative files under the mount point /data/oracle/RONDG. They were:

    * admin
    * admin/adump
    * admin/pfile
    * admin/scripts
    * admin/dpump