Understanding Virtual Port Groups (vPG)

In addition to vLE, Arista Networks’ NetVisor OS supports an additional and more flexible VXLAN-based technology called a Virtual Port Group (vPG). This powerful CLI construct can be used to define a set of ingress (or source) ports and of destination ports so that the hardware can transport traffic from the source ports to (multiple) desired destination ports, and (optionally) in the reverse direction too.

Forwarding is performed in hardware using vFlow policies. For more details on vFlow, please refer to the Configuring vFlow chapter.

When the vPG configuration forwards traffic from the source ports to the destination ports only, it is referred to as a unidirectional vPG topology (as the monitored traffic flows in one direction only).

The primary use case for this configuration is Arista’s Network Packet Broker. For more details on this use case, please refer to the Understanding and Configuring Network Packet Broker sections. These sections provide a full configuration walk-through to set up vPGs and vFlow policies to establish a secondary monitoring network for production traffic.

On the other hand, when the vPG configuration forwards traffic bidirectionally between source ports and destination ports, it is referred to as a bidirectional vPG topology: it can be useful to establish a sophisticated set of virtual wires between network nodes for bidirectional data exchange (for example, for lab automation use cases).

About Virtual Port Group State Tracking 

In bidirectional vPG topologies (for example, for lab automation purposes) it is often required to mirror the state of one end of the connection to the other end.

Starting from NetVisor OS release 7.0.1, the bidirectional vPG functionality leverages a BFD-based state tracking logic to achieve state tracking across VXLAN tunnels. In other words, when enabled this tracking logic makes sure that, if a port on a local vPG goes down, the corresponding remote vPG's port is also brought down, and vice versa. This tracking functionality can be particularly handy in lab automation use cases.

The common tracking logic is called the vBT (vLE/vPG BFD Tracking) module. For vPGs this logic is currently limited to intra-fabric tracking.

Whenever a local port (which is a part of a vPG with tracking enabled) goes down, the vBT logic uses BFD to propagate the change of state to the remote endpoint which then brings its corresponding port down and puts it into vbt-wait state. Subsequently, whenever that local port comes back up, the state change is again propagated and the remote endpoint is brought back up too and removed from the vbt-wait state.

Note that unidirectional vPGs don’t support this bidirectional tracking logic.

Moreover, only point-to-point bidirectional vPGs support state tracking: that means that each bidirectional vPG can have only one port in one node for tracking to be enabled.

Refer to the Configuring Virtual Port Group State Tracking section below for the command syntax details.