Networking Minerva User Guide

Assign Device Roles and Groups

Roles and groups give NVIDIA Minerva the network context it needs to interpret telemetry. Device roles identify a device's function, interface roles identify the direction and purpose of a connection, and groups place devices in the network hierarchy.

Device role assignment is a prerequisite for fabric validations; devices without a role are excluded from validation regardless of their operational state.

For repeatable or bulk device-role and group-membership changes, use the REST API. It does not document interface-role updates.

Automatic Discovery

For deployments that follow the NVIDIA Spectrum Reference Architecture, NVIDIA Minerva normally discovers standard switch roles and pod and Scalable Unit (SU) metadata automatically when the complete supported topology feed is delivered over OTLP with the expected RCP labels.

RCP label

How NVIDIA Minerva uses it

switch_type

Assigns the standard switch role, such as leaf, spine, or super_spine.

pod

Places leaf and spine switches in the corresponding pod group; for example, pod=0 maps to Pod-0.

su

Places leaf switches in the corresponding SU within a Pod; for example, pod=0 and su=2 map to Pod-0-SU-2.

LLDP telemetry received through OTLP supplies the observed devices, interfaces, and neighbor relationships. Companion switch-interface telemetry in the same OTLP feed supplies the pod and su labels used for group placement. LLDP alone does not provide pod or SU placement.

No separate RCP import is required when the complete supported OTLP feed already contains these labels. If links are discovered but roles or pod/SU assignments are absent, verify that the relevant labeled telemetry—not only LLDP—is reaching NVIDIA Minerva.

Automatic population is scoped to metadata present in the telemetry. Server roles, rack membership, custom roles, and custom groups can require another source or a manual or API update.

After telemetry begins, review the result under Topology > Provision > MetaData before making manual changes.

Choose an Update Method

Method

Use it for

Automatic OTLP discovery

The normal starting point for a supported Spectrum Reference Architecture deployment.

NVIDIA Minerva UI

Reviewing discovered values and making small or one-time corrections.

REST API

Repeatable, bulk, or automated updates to device roles and device group membership. See Manage Device Roles, Groups, and Administrative State with the REST API.

Manual changes update the current network model, but later labeled telemetry can update those values again. If an automatically populated value is consistently wrong, correct its source labels when possible and verify discovery again.

Update Roles and Groups

From the UI, go to Topology > Provision > MetaData.

If you need to assign a custom role or group, create it first on the Roles or Groups tab before following the steps below. To update a device role or group, follow these steps:

  1. From the Devices tab, select one or more devices.

  2. Select Actions > Edit device(s).

  3. Select an existing Role, Group Name, or both.

  4. Click OK, then confirm the change.

  5. Review the result notification or the Change Logs tab.

To update an interface role, follow these steps:

  1. Open the Interfaces tab, select one or more interfaces.

  2. Select Actions > Edit interface(s).

  3. Select an existing Role.

  4. Select OK, then confirm the change.

  5. Review the result notification or the Change Logs tab.

A device has one role value, so assigning a different role replaces its current role. Selecting a Group Name in Edit device(s) replaces the device's current group membership with the selected group — it does not add a membership. To add a device to a group without removing its existing memberships, use the REST API link operation instead. See Manage Device Roles, Groups, and Administrative State with the REST API.

Common Device Roles

Role

Purpose

Super spine (super_spine)

Upper fabric tier connected to spine switches, when present.

Spine (spine)

Fabric backbone connected to leaf switches.

Leaf (leaf)

Access tier connected to compute nodes.

Server (server)

Compute host.

Custom

A deployment-specific role created in NVIDIA Minerva.

Interface Roles

NVIDIA Minerva derives standard interface roles after both endpoint device roles and the discovered link are present in the network model. Standard pairs include:

  • Superspine to Spine and Spine to Superspine

  • Spine to Leaf and Leaf to Spine

  • Leaf to NIC and NIC to Leaf

Deployments may also have source-provided or customer-defined interface roles.

Common Groups

Group

Purpose

Example

Datacenter (datacenter)

Top-level physical grouping.

DC1

Pod (pod)

Logical grouping of network and compute resources.

Pod-0

Scalable Unit (su)

Repeatable building block within a pod.

Pod-0-SU-2

Rack (rack)

Optional compute grouping within an SU.

Rack-01

Super spine (superspine)

Datacenter-level grouping for super spine devices.

SuperSpine

Custom

A deployment-specific nested group.

Production


Last updated: