Donald K. Burleson Oracle RAC Tips

Configure using HACMP/ES

As discussed earlier, the HACMP method may involve the use of concurrent logical volume groups. The HACMP layer includes the concurrent logical volume manager (CLVM).

HACMP/ES installation has many steps, which include:

1. Installing and setting up the hardware.

2. Installing the AIX operating system.

3. Installing the latest maintenance and required patches.

4. Installing the HACMP/ES on each node and applying patches.

Configure the Cluster Topology

This step involves using smit hacmp menus to configure the cluster details. All configurations are done from one node and then synchronized to the other participating nodes. Adding a cluster definition process requires the information such as cluster id, cluster name, nodes, and adapter details to be supplied. After the topology is configured, it has to be synchronized with other nodes. Main steps for implementing an HACMP configuration are:

1. Configure the cluster: Select ‘Add a cluster definition’ - In this step the cluster must be given an identifier and a name.

2. Configure nodes of the cluster: Select ‘Add cluster nodes’. Here, the names of the nodes participating in the cluster are entered.

3. Configure the network: Select ‘Add an adapter’. Provide details such as Adapter IP label, network type, and network name, etc., for each node.

In the case of the implementation of a concurrent volume group, only a private network is needed. It’s not necessary to define boot, service, and standby addresses as it is in cascading or rotating HACMP configurations. In RAC environments, a high performance private network between machines participating with the RAC cluster will be needed. The network type of this interconnect can be HPS (High Performance Switch) or Ethernet (use a gigabit Ethernet, for example). Note: It is also advisable to have a non-IP network (like RS/232 or Target mode SSA).

Next step is to synchronize the cluster topology. The synchronization performs topology sanity checks, as well as pushing the configuration data to each of the nodes in the cluster configuration.

To synchronize, use the fast path command smit configchk.dialog.

Configure a Resource Group

In a RAC configuration, creation of a resource group is required. This resource group is a concurrent group for a shared volume group. The following steps add a resource group.

1. First, the Volume Group that is shared between the nodes must be created. The Oracle instances residing on multiple nodes share the volume group concurrently. The concurrent logical volume manager manages the volume group. The HACMP layer includes the CLVM, which enables the concurrent logical volume manager.

CLVM only works under the HACMP environment. Oracle will not allow VSD(s) and concurrent logical volumes (CLV) to be used on the same database. If PSSP services are being used, Oracle will report an error if the customer attempts to use CLV. If HACMP services are used, Oracle will report an error if the customer attempts to use VSD(s). When using CLVM or HACMP, the environment variable pgsd_subsys should be set to grpsvcs.

2. Using the smit vg and then selecting the add a volume group option creates the volume group. Provide the group name, disk details, major number, etc. The volume group should not be activated automatically.

3. Create the necessary volumes for use by the Oracle installation and database creation by using the command mklv.

4. Change permissions on the character devices so the software owner owns the volumes with the chown command.

5. Then, import the volume group to the other nodes. First, varyoff on the first node, then import the definition of the volume group using smit vg – select the ‘import a volume group’ option.

6. After importing the volume to each of the nodes in the cluster, change the ownership of the character devices to the software owner ‘oracle’.

7. Now, create a concurrent cluster resource group that manages the volume group created in the previous step.

8. Use the smit cm_add_grp with option add a resource group. Configure the node relation as concurrent, which indicates that the multiple nodes share the resource group. Specify the list of nodes that share this resource.

9. The next step is to synchronize resources by using the smit hacmp menu.

10. Now start the cluster services using the smit hacmp menu.

So far, we have assumed that the volume group has the only volumes and that they will be presented to Oracle as the raw devices. However, with the availability of the General Parallel File System (GPFS), we have the option of setting up the cluster file system for the use of Oracle database files. The next section explains the features of the GPFS.