Issue/Introduction:
This article discusses VRRP behavior with 3 or more chassis in VRRP cluster.
Cause:
When configuring VRRP with 3 or more chassis, VRRP failover randomly occurs between the other chassis and not to the next-highest-priority one.
For example:
Chassis 1:
vrrp failover-group XXX failover-group-id X
priority 250
Chassis 2:
vrrp failover-group XXX failover-group-id X
priority 200
Chassis 3:
vrrp failover-group XXX failover-group-id X
priority 150
Failover from Chassis 1 will be to either Chassis 2 or Chassis 3.
Resolution:
VRRP failover uses a "first to answer" mechanism.
The chassis that answers firts will become "master" for the failover group after a failover.
To ensure that failover takes place by priority, enable preemption on chassis to ensure that the next priority chassis becomes master.
Even if another chassis answers first, preemption will ensure a failover to the highest priority chassis.
Chassis 1:
vrrp failover-group XXX failover-group-id X
preemption
priority 250 Chassis 2:
vrrp failover-group XXX failover-group-id X
preemption
priority 200 Chassis 3:
vrrp failover-group XXX failover-group-id X
preemption
priority 150
To reduce automatic failbacks, you can enable preemption on Chassis 2 only.
When Chassis 1 goes down, even if Chassis 3 is the first to answer, preemption will ensure that Chassis 2 becomes master.
When the problem that caused a failover from Chassis 1 is rectified, no failback will occur because preemption is not configured on Chassis 1.
However, in this case, failing back from Chassis 2 will be on "first to answer" mechanism.