12: GR and BFD

Overview of High Availability

Availability refers to the ability of a user group to access a desired system or resource. High availability is the ability to ensure a high degree of operational continuity to a resource for a user community. It is not the same as uptime. A system can be up, but unavailable.

High Availability Networks

When building networks for high availability, consider whether the base functionality of the network’s operations protected from hardware or software failure, and when a failure does occur, how the overall system or network will react and what kind of recovery can be expected.

High Availability Features

Junos supports a number of high availability features;

• Graceful restart
  • Enables uninterrupted packet forwarding with temprary suppression of routing protocol updates
• Graceful Routing Engine Switchover (GRES)
  • Provides switchover between routing engines when one fails
• Nonstop Routing (NSR)
  • Uses the same infrastructure as GRES, but also maintains routing information, so does not require neighbouring routers to restore routing information
  • NSR is a natural replacement for GRES, that depends on GRES
• Bidirectional Forwarding Detection (BFD)
  • A hello mechanism that detects link failures
• Virtual Router Redundancy Protocol (VRRP)
  • Enables first hop redundancy
  • No more than a static route is required on end devices
• Unified In-Service Software Upgrade (ISSU)
  • Enables Junos upgrades without interruption to traffic
  • Only supported on devices with dual routing engines

Graceful Restart

A router about to undergo a restart event informs its neighbours and requests a grace period. During the grace period, the router continues to forward traffic through existing paths. After the restart, the router synchronises with its neighbours, again without disrupting packet forwarding.

The neighbours of the restarting router hide the event from other routers, further reducing disruption. Graceful restart can only occur when the network topology is stable, the neighbour or peer cooperates, the restarting router is not already cooperating with another restart, and the grace period does not expire.

Graceful restart is supported by many standard protocols; OSPF, IS-IS, BGP, RIP, RSVP, LDP, MSDP and PIM.

Routers must have graceful restart enabled to support both graceful restart router modes; restarting router mode, and helper router mode. By default, Junos dedvices will act as helper routers, but not restarting routers.

Configuring Graceful Restart

Graceful restart helper mode can be disabled globally;

routing-options {
    graceful-restart {
        disable;
    }
}

It can also be disabled per protocol, or more granularly, depending on the protocol.

Graceful restarting restarting router mode can be configured globally, per protocol, or with more granularity. It is enabled globally with;

routing-options {
    graceful-restart;
}

Monitoring Graceful Restart

Verification methods vary between protocols. For BGP, use show bgp neighbor <PEER_ADDRESS>. The GracefulRestart options will appear if graceful restart helper router mode is enabled on the remote router.

For OSPF, enable traceoptions with the graceful-restart flag.

Bidirectional Forwarding Detection

All modern protocols have a mechanism to detect failures. The problem with these is that they are slow; for example, OSPF takes 40 seconds to detect failures with the default timers.

Bidirectional Forwarding Detection (BFD) is a simple protocol to rapidly detect link failures. It takes less than a second to detect a failure. When a failure is detected, the system is notified. With BFD, timers for other protocols can be left at their defaults. BFD provides a unified method for rapid link failure detection that works across a range of protocols. It also works for static routes, which would not otherwise have a link failure detection mechanism. BFD is supported for OSPF, IS-IS, RIP, BGP, RSVP, PIM and static routes.

Configuring BFD

bfd-liveness-detection enables BFD. minimum-interval sets the minimum transmit and receive interval in one step. The transmit and receive intervals can also be set separately. The value is set in milliseconds. It is not recommended to set a value lower than 300ms. The receive interval multiplier is 3; if a BFD notification is not received within 900ms, the link is considered to have failed.

If the minimum-interval values differ at each end, the higher value is used.

protocols {
    bgp {
        group my-group {
            bfd-liveness-detection {
                minimum-interval 300;
            }
            ...
        }
    }
}

protocols {
    ospf {
        area 0.0.0.0 {
            interface ge-0/0/0.0 {
                bfd-liveness-detection {
                    minimum-interval 300;
                }
            }
        }
    }
}

BFD is not useful for all media types. For instance, ATM and SONET both already include rapid link failure detection. Ethernet, however, benefits greatly from BFD, especially when intermediate switches are located between devices. Ethernet OAM is a close alternative to BFD, but operates at Layer 2.

Monitoring BFD

show bfd session lists BFD sessions with their interval times, which may not be the same as the locally configured intervals. show bgp neighbor <PEER_IP> shows the BfdEnabled option if BFD is supported and BFD: enabled, up if it is running. If BFD should be running, but is not, a common fault is that the firewall is not configured to permit BFD packets.