The Highly Available NVMe-oF on RHEL 9 tech guide will instruct the reader on how to configure a Highly Available (HA) NVM Express over Fabrics (NVMe-oF) cluster using DRBD® 9 from LINBIT® and the Pacemaker cluster stack on Red Hat Enterprise Linux (RHEL) 9.
NVMe-oF is the concept of connecting to a remote NVM Express (NVMe) device over some network transport. Different connection media, or fabrics, sometimes have their own names such as NVMe over Fibre Channel (FC-NVMe) and NVMe over InfiniBand (NVMe/IB), but the concept can be generically referred to as NVMe-oF.
NVMe-oF enables low latency connections from client systems to storage servers with internal NVMe devices known as NVMe targets creating an end-to-end NVMe storage solution. However, using NVMe-oF to connect client systems to storage servers containing more traditional — non-NVMe — storage will also benefit from using the lower latency NVMe protocol over the wire.
Using LINBIT’s DRBD for synchronous replication of NVMe devices with the Pacemaker cluster stack for cluster resource management makes building active/passive HA NVMe target clusters possible using open source software and commercial off-the-shelf hardware.