Planning high availability

When you plan for replication for high availability, check that your environment supports it and that you understand the limits of configuration.

The following configuration limits apply to policy-based high availability:

  • A limit of one I/O group can be configured in the system.
  • A maximum of one partnership can be used for high availability.
  • A partnership cannot be used for replication for both high availability and disaster recovery.
  • SCSI hosts connected through Fibre Channel or iSCSI can be configured in a storage partition that is configured for high availability.
  • NVMe-TCP and NVMe-FC hosts can be configured in a storage partition that is configured for high availability.
  • Host clusters can be configured in a storage partition that is configured for high availability if both systems are running IBM Storage Virtualize 9.1.0 or later.
  • SCSI Persistent Reservations are supported for Fibre Channel hosts if both systems are running IBM Storage Virtualize 8.7.2 or later.
  • SCSI Persistent Reservations are supported for iSCSI hosts if both systems are running IBM Storage Virtualize 8.7.3 or later.
  • NVMe reservations are not supported for NVMe-FC and NVMe-TCP hosts.
  • Volumes that are configured for high availability cannot be mapped to other IBM Storage Virtualize systems.

Policy-based high availability requires a partnership to be configured between the systems by using either Fibre Channel or Ethernet (short distances that use RDMA) connectivity.

The systems should be sufficiently far apart such that anticipated disruptions affect only one location. The systems must be sufficiently close together to not cause the response time for I/O to exceed the maximum that an application can tolerate because of latency in replication link communications. For policy-based high availability, a maximum of 1 millisecond round-trip time (RTT) is supported for the inter-site link between systems.

IP connectivity is required between systems for management. Use of Network Address Translation (NAT) between systems is not supported. The following TCP ports are required to be accessible on the system management IP address:

  • Port 7443 for REST API access between systems.
  • Replication setup that uses the management GUI requires access to both systems from the host where the web browser is running. Ensure that any firewalls between the web browser and storage systems allow inbound traffic to port 7443 and 443 on the system management IP address.

In addition, both systems must have IP network connectivity to an external partnership IP quorum application to arbitrate if a site or system failure occurs. Multiple quorum applications can be deployed for redundancy.

Hosts must use an ALUA-compliant multipath policy. Policy-based high availability supports a reduced set of the operating systems supported by IBM Storage Virtualize systems. For more information, see Supported host operating systems for high availability.

Hosts can be configured with a non-uniform topology, where each host has connectivity to only one of the locations in the high availability configuration. To accurately report the high availability status of these hosts, specify a location for each host, and group those hosts in a host cluster with the uniform connectivity parameter set to No. Add the hosts and the host cluster to a storage partition with a replication policy that identifies the system at each location. Hosts should be defined in pairs, one for each location to provide redundant access by the host cluster in the case of an individual system outage.

The following restrictions apply to policy-based high availability:

  • When a storage partition is configured for high availability, names of partitions, hosts, volume groups, and volumes in the partition can be changed only if all systems are running IBM Storage Virtualize 9.1.0 or later. You can rename an object on the active management system and the changed object name applies to all copies.
  • Policy-based high availability cannot be used with volumes that are:
    • Image mode
    • Cache disabled
    • Configured to use Transparent Cloud Tiering (TCT)
  • A volume in a storage partition with a replication policy assigned cannot be:
    • Shrunk. Volume expansion is supported if all systems are running IBM Storage Virtualize 9.1.0 or later.
    • Migrated to image mode, or have an image mode copy added.
  • A volume can be moved between volume groups within a partition if all systems are running IBM Storage Virtualize 9.1.0 or later.