This describes how to use Bidirectional Forwarding Detection (BFD) in NetScaler.

Background

Bidirectional Forwarding Detection commonly referred as BFD is a simple hello protocol that provides fast failure detection mechanism between two routers/nodes. It detects failures on any bi-directional forwarding paths, such as direct physical links, tunnels, virtual links, multi-hop paths across network devices. BFD is a mechanism that is independent of media, routing protocol, and data protocol used and validates the operation of the forwarding plane. Since it is not tied to any routing specific protocols it can be used as a generic failure detection method between network devices.

Why BFD?

In Routing protocols, fault resolution happens in order of tens of seconds. Reducing the OSPF and BGP timers can bring it down to four to five seconds to detect failure, whereas BFD can do the trick in less than one second. This helps in faster network convergence, short application interruptions, and enhanced network reliability. Another advantage is that BFD is lightweight and can run completely on the data plane offloading the control plane CPU.

How does BFD work?

BFD session establishment happens in a series of steps. Let us assume the routing protocol used here is OSPF(BGP follows similar steps for other routing protocols).

1. OSPF discovers neighbours by sending Hello packets and establishes adjacencies with the neighbouring nodes.
2. OSPF notifies BFD of the neighbour to be monitored by including source and destination address.
3. BFD uses the information to establish session and sends control packets at regular intervals to the peer nodes.

There are 3 modes of BFD operation:

1. Asynchronous - Control packets flow both directions periodically. (Note: This is supported in NetScaler)
2. Demand - Periodic control packets are not sent. After BFD session establishment, a node can ask the other system to stop sending BFD Control packets, except when the node feels needs to verify connectivity explicitly. (Note: This is not supported in NetScaler)
3. Echo mode - a stream of Echo packets is transmitted in such a way that it loops back through its forwarding path to the same node to verify connectivity. Note: This is not supported in NetScaler.

Failure detection happens in a series of steps as mentioned below,

1. A link goes down between NetScaler and adjacent node
2. BFD sessions goes down immediately based on timer expiry
3. BFD notifies neighbour unreachability to OSPF
4. OSPF terminates the neighbour adjacency

---

Instructions

Enabling BFD for OSPF

The below mentioned configuration has to be done at VTYSH to enable BFD for OSPF.

Enable or Disable BFD for all interfaces at ospf router level
ZebOS(config-router)#bfd all-interfaces
ZebOS(config-router)#no bfd all-interfaces

Enable or disable BFD at interface level for OSPF
ZebOS(config-if)#ip ospf bfd
ZebOS(config-if)#ip ospf bfd disable

Enabling BFD for OSPFv3

Enable or Disable BFD for all interfaces at ospfv3 router level
ZebOS(config-router)#bfd all-interfaces
ZebOS(config-router)#no bfd all-interfaces 

Enable or disable BFD at interface level for OSPFv3
ZebOS(config-if)#ipv6 ospf bfd
ZebOS(config-if)#ipv6 ospf bfd disable

Enabling BFD for BGP

BGP IPv4 Singlehop-peer
ns(config-router)#neighbor <ipv4addr> fall-over bfd
ns(config-router)#no neighbor <ipv4addr> fall-over bfd

BGP IPv4 Multihop-peer
ns(config-router)#neighbor <mh-neighbor-ipv4addr> fall-over bfd multihop
ns(config-router)#no neighbor <mh-neighbor-ipv4addr> fall-over bfd multihop

BGP IPv6 Singlehop-peer
ns(config-router)#neighbor <ipv6addr> fall-over bfd
ns(config-router)#no neighbor <ipv6addr> fall-over bfd

BGP IPv6 Multihop-peer
ns(config-router)#neighbor <mh-neighbor-ipv6addr> fall-over bfd multihop
ns(config-router)#no neighbor <mh-neighbor-ipv6addr> fall-over bfd multihop

Configuring BFD Timers

BFD Single-hop Session Timer

Configure BFD single-hop sessions timer and reception interval in millisecond, and the Hello multiplier in interface mode
bfd singlehop-peer interval <100-30000> minrx <100-30000> multiplier <1-20>

Multiplier indicates the number of packets to be dropped before declaring the interface as down
Minrx is the value that is advertised to other nodes as the time interval for reception by NetScaler
Interval is the time interval with which NetScaler sends hello packets

To unset the timers:
no bfd singlehop-peer
Ex: ZebOS(config)#interface vlan10
ZebOS(config-if)#bfd singlehop-peer interval 100 minrx 100 multiplier 4

BFD Multihop Peer Timer

Configure BFD multihop-peer timer and reception intervals in milliseconds, and the Hello multiplier in config mode
bfd multihop-peer A.B.C.D interval <100-30000> minrx <100-30000> multiplier <1-20>

Unset the timers using no command:
no bfd multihop-peer A.B.C.D

Ex: ZebOS(config)#bfd multihop-peer 20.1.1.3 interval 100 minrx 100 multiplier 3

BFD IPV6 Multihop Peer Timer

bfd multihop-peer ipv6 X:X::X:X interval <100-30000> minrx <100-30000> multiplier <1-20>

Unset the timers using no command:
no bfd multihop-peer ipv6 X:X::X:X
Ex: ZebOS(config)#bfd multihop-peer ipv6 3001:20::4 interval 100 minrx 100 multiplier 3

BFD Passive mode

NetScaler will not initiate BFD control packets when passive mode is enabled. This is used where the device need not initiate control packets first. Once it receives a control packet, it triggers state change.

BFD Passive mode can be configured as follows in interface mode:
ZebOS(config)#interface <vlanid>
ZebOS(config-if)#bfd passive 

To unset the passive setting: 
ZebOS(config-if)#no bfd passive

BFD in NetScaler is supported for both IPv4 and IPv6 . It is supported for routing protocols in the context of admin-partitions, on spotted SNIP in the context of cluster and on SNIP-IP in HA_INC mode.

This describes how to use Bidirectional Forwarding Detection (BFD) in NetScaler.