Skip to main content
E-Series storage systems

Set up failover on the host in E-Series - Linux (NVMe over RoCE)

Contributors netapp-driley netapp-jolieg NetAppZacharyWambold
PDFs

To provide a redundant path to the storage array, you can configure the host to run failover.

Before you begin

Refer to the NetApp Interoperability Matrix Tool to ensure any required updates are installed, as important bug fixes may be required for versions of SLES or RHEL that are not available in the initial release.

About this task

For SLES 15, SLES 16, RHEL 9 and RHEL 10, Native NVMe Multipathing is enabled by default and no additional configuration is required.

Configure NVMe I/O policy

About this task

Linux NVMe native multipathing may default to the numa I/O policy. For E-Series/EF-Series NVMe-over-Fabrics configurations, the recommended I/O policy is round-robin, which distributes I/O evenly across all available paths.

Steps

  1. Verify the current NVMe I/O policy:

    # cat /sys/module/nvme_core/parameters/iopolicy
numa

  2. Set the round-robin I/O policy using one of the following methods:

    • Persistent (recommended) — Create a udev rules file:

      /etc/udev/rules.d/71-nvme-io-policy.rules
      ACTION=="add|change", SUBSYSTEM=="nvme-subsystem", ATTR{iopolicy}="round-robin"

      Then reload and apply the rules:

      udevadm control --reload
udevadm trigger --subsystem-match=nvme-subsystem

    • Global setting at boot — Add the following kernel parameter to your bootloader configuration:

      nvme_core.iopolicy=round-robin

    • Non-persistent (online, no I/O interruption) — Apply immediately without rebooting:

      echo "round-robin" > /sys/class/nvme-subsystem/nvme-subsys0/iopolicy
      Note This setting does not persist across reboots.
What's next?

With the native NVMe Multipathing enabled, access the NVMe volumes for an E-Series target from the host.