Managing the shared storage configuration
The GDR solution manages disaster recovery across two sites based on storage replication across the sites. However, the GDR solution also supports a mode of deployment in which disks are shared across sites. In this case, the KSYS subsystem does not manage any storage subsystems. The disks can be shared across sites that are separated by short distances (0 - 100 km). The storage technologies (for example, Storwize® HyperSwap®) perform synchronous mirroring across sites and abstract the mirroring from the hosts. These storage technologies provide shared disk type of deployment for hosts across sites.
- Because the storage is shared, the N_Port ID virtualization (NPIV) and other similar ports are visible to VIOS on both sites. It might cause problems that are related to SAN login and disk validations. Hence, HMC and VIOS-related checks are not performed in the shared deployment mode. Therefore, the administrator must set up the sites appropriately, considering the storage, network, and so on, and must maintain the configuration settings. Any misconfiguration might result in errors during a disaster recovery.
- The DR failover rehearsal operation is not supported for shared storage deployments.
Shared storage without replication management
When the storage device in your environment is a single storage system that is split or separated by distance in two different sites as stretched systems, the storage replication management is hidden from the hosts and VIOS partitions. The storage subsystem is displayed as a single shared storage across the two sites. In this case, the storage recovery and replication are performed entirely by the storage platform. You do not need storage agents in the KSYS subsystem to interact with the storage devices. Therefore, you do not need to consider the disk pair and disk group mappings for the shared storage configuration.
When you move the virtual machines from the active site to the backup site, the KSYS subsystem considers the storage subsystem as unmirrored shared storage and starts the virtual machines on the backup site. If the disaster recovery operation is unplanned, the storage subsystem performs the entire storage recovery.
- The GDR solution supports this type of shared storage only for sites that are spanned across short distances. Also, the storage subsystem must provide shared storage characteristics.
- The GDR solution does not support heterogeneous storage systems for this type of shared mode configuration. You must deploy same type of storage systems across the sites in your environment.
The following figure shows an example of shared storage configuration with Storwize HyperSwap based technology:
Shared storage with SSP mirror management
VIOS can manage storage efficiently across multiple hosts in an environment by using the Shared Storage Pool (SSP) technology. Shared storage provides various features that support mirroring. Mirroring enables continuous storage availability across sites against storage failure in the site. If your environment is managed entirely by SSP for both system (for example, rootvg) and data disks, then you can spread the SSP cluster across two buildings and achieve protection against storage failures. This type of deployment is suited for short distances (less than 1 km).
SSP-based deployments belong to the shared storage category. That is, the KSYS subsystem does not manage the storage replication. Because the SSP-based storage exhibits the shared storage model, you do not need to consider the disk pair and disk group mappings for storage configuration.
When you move the virtual machines from the active site to the backup site, the KSYS subsystem does not perform any storage-specific preparation (unlike the conventional replicated storage systems) and starts the virtual machines on the backup site.
