Sample configuration
GLVM for a PowerHA® SystemMirror® cluster consists of two sites located at a geographic distance from one another. Both sites can support a combined total of 16 cluster nodes.
- Fallover to a Remote Backup Site. In this configuration, a local site serves as a production site and a remote site serves as a backup site. The backup site includes hardware, system and application software, and application data and files. It is ready to take over the current workload on the local site. In the event of a failure, the application from the local site automatically falls over to the remote site.
- Remote Mutual Takeover. Remote mutual takeover allows two sites to serve as backups for each other. If a failure occurs at a node on one site, the node at the other site automatically takes over the application. With mutual takeover, the initial local site may become the new remote site with the direction of mirroring reversed. Mutual takeover requires two separate resource groups.
Cluster configuration with sites and RPVs
The following figure shows a two-site cluster configuration with one node at each site:
In this example:
- Sites and nodes configuration.
- The PowerHA SystemMirror cluster contains two sites: site Ohio and site Indiana.
- Each site has one node.
- Each node has a local physical volume configured.
- Networks configuration.
- The TCP/IP XD_data network is used for data mirroring. This diagram uses just one network. You can configure up to four XD_data mirroring networks in a cluster with GLVM for PowerHA SystemMirror Enterprise Edition.
- RPV configuration. Two RPV client and server pairs are configured for each resource group. Note, this is true in this example because the resource group contains two disks. In general, however, an RPV client and server pair must be configured for each disk. The node on which the disk is local requires an RPV server, and the node on which the disk remote requires an RPV client.
- RPV client on Node A and RPV server on Node B. For a resource group residing on site Ohio, site Ohio is the local site and site Indiana is the remote site. Node A has an RPV client configured and Node B has a corresponding RPV server configured. The mirroring occurs in the direction from the RPV client on the local site to the RPV server on the remote site. This enables data mirroring for volume groups and applications residing on site Ohio (the local site for this resource group) to site Indiana (the remote site for this resource group).
- RPV client on Node B and RPV server on Node A. For a resource group residing on site Indiana, site Indiana is the local site, and site Ohio is the remote site. Node B has an RPV client configured. Node A has a corresponding RPV server configured. For this resource group, the applications that reside on site Indiana are mirrored to site Ohio. The mirroring occurs in the direction from the RPV client configured on Node B at site Indiana to the RPV server configured on Node A at site Ohio.
RPV client-server pairs required at both sites
The previous example illustrates that configuring RPV client-server pairs at both sites is necessary to ensure mutual takeover and recovery in PowerHA SystemMirror for resource groups that reside on nodes at each site.
For instance, a resource group resides on the local site and Node A is its home node. The goal of GLVM for PowerHA SystemMirror Enterprise Edition is to recover the resource group on the remote site, in case of a failure on the local site, and if there is no other node at the local site on which to recover the resource group.
For this purpose, an RPV client-server pair must be configured for the volume group in the resource group. An RPV client is configured on Node A (local site for this resource group) and the corresponding RPV server is configured on Node B (remote site for this resource group). Geographic mirroring occurs in the direction from the RPV client on the local site to the RPV server on the remote site. If a fallover between sites occurs, GLVM for PowerHA SystemMirror Enterprise Edition recovers the resource group on Node B and the application continues to run. GLVM for PowerHA SystemMirror Enterprise Edition activates the local volume group and enables access to the disk directly on the takeover node.
Similarly, for another resource group that initially resides on site Indiana (this is the local site for this resource group) and has Node B as its home node, an RPV client must be configured on Node B and the corresponding RPV server must be configured on Node A. If a fallover between sites occurs, GLVM for PowerHA SystemMirror Enterprise Edition recovers the resource group on Node A and the application continues to run. GLVM for PowerHA SystemMirror Enterprise Edition activates the local volume group and enables access to the disk directly on the takeover node at another site.
Cluster configuration with two nodes at each site
The following figure shows a sample two-site two-node cluster configuration with nodes that use geographically mirrored volume groups as shared disks.
In this example:
- The PowerHA SystemMirror cluster contains two sites: the local site and the remote site.
- Each site has two nodes.
- Node 2 on the local site has an RPV client configured. (This node also must have an RPV server configured, but it is not shown. See the following note).
- Node 3 on the remote site has an RPV server configured that communicates with the RPV client on Node 2 using at least one IP-based network (XD_data). A XD_ip networks is not shown. You can have up to four XD_data networks. The RPV client on Node 3 is also not shown, see the following note.
In this configuration, a resource group on Node 2 contains an application that accesses volume groups at the local site. All volume groups that are accessed by this application have both local PVs and RPVs. These volume groups are geographically mirrored with the GLVM mirroring function. The groups use an RPV as a layer that allows the node to access them.