Overview
Policy-Based High Availability initially supported only FC SCSI and iSCSI hosts. Starting with 9.1.1 release, it now includes support for NVMe over Fabrics hosts.
This allows NVMe-FC or NVMe-TCP hosts to be configured in storage partitions. With this, the hosts using these protocols can take the advantage of IBM FlashSystem Grid features, along with data replication, online data migration and High availability data replication.
As shown in the diagram, a partition is configured with volumes that are mapped to a VMWare ESXi host (The host-object configuration can be over NVMe-FC or NVMe-TCP protocols) and can be part of host cluster optionally.
Essential prerequisites to consider
•User-defined portsets must be used for host objects on Active management site (AMS) and non-active management site.
•These portsets must be linked between the sites
•Hosts must have access to both the sites
•AMS and non-AMS must support NVMe host-attachment with supported adapters and FlashSystem code versions
The above picture depicts the overall configuration view. There are two FlashSystem clusters, Active management site and non-active management sites with partnership. Fiber Channel or RDMA based partnerships can be used for communication between two clusters for high availability configuration along with the requirement of an active IP quorum.
The AMS and non-AMS connect to VMware ESXi hosts through the fabric, and the host connection could be either NVMe-FC or NVMe-TCP hosts
On the host side, this can be a standalone ESXi or a VMware vSphere Metro Storage Cluster (vMSC) configuration, and both configuration work seamlessly
The hosts are accessible from both sites and host-objects and volumes mapped to the host object is reflected across the sites.
Redundant connectivity is established between all systems to ensure high availability and fault tolerance
Configuring a NVMe-FC or NVMe-TCP Host
So, how does the user configure/add NVMe-FC/NVMETCP host-object?
Inside a partition on the ‘Add Host' option, now user can see 2 more options, “Fibre Channel (NVMe)” and “TCP (NVMe)” listed under available options. You can select the appropriate host connection option and enter the required details — hostname, NQN, and linked portsets. Only portsets eligible for the selected protocol will be displayed. Ensure that these portsets are already linked to the partner system, which is one of the prerequisites covered above
Once the host object is configured on the AMS site, the protocol type NVMe-FC/NVMe-TCP and port definitions are reflected. Post this map volumes to these host objects.
Once complete same host object and protocol configuration details also reflected on the non-AMS site, ensuring consistency across both sites.
vSphere View — Mapped Volumes
On the host side, within the vSphere Client, users can view all the mapped volumes under Storage Devices.
· Volumes mapped to an NVMe-TCP host object appear as NVMe TCP disks.
· Volumes mapped to an NVMe-FC host object appear as NVMe Fibre Channel disks.
This makes it easy to identify which protocol each volume is using right from the vSphere interface.
Helpful References
Implementing a High-availability Solution: https://www.ibm.com/docs/en/flashsystem-9x00/9.1.1?topic=replication-high-availability
Host Configuration: https://www.ibm.com/docs/en/flashsystem-9x00/9.1.1?topic=concepts-hosts
Note: This blog provides a high-level overview of the feature and is not intended to replace the official documentation. For detailed guidance on limits, restrictions, and configuration best practices, please refer to the official IBM documentation before implementation.