Fixing VRRP state changes caused by "Receive advertisement timeout"

In some cases when load is high you may observe the following error in vrrp logs (docker logs vrrp):

Tue Nov 17 02:26:36 2020: (nas) Receive advertisement timeout
Tue Nov 17 02:26:36 2020: (nas) Entering MASTER STATE
Tue Nov 17 02:26:36 2020: (nas) setting VIPs.
Tue Nov 17 02:26:36 2020: (nas) Sending/queueing gratuitous ARPs on eth0 for 1.1.1.1

In this case an unintended VRRP state change is occurring for service nas because the instance in BACKUP state has detected that the MASTER instance is no longer responding. This situation may be caused by high load on the MASTER node and also connectivity issues between instances. The BACKUP instance waits 3 intervals to initiate switchover; by default this is 3×1 seconds — 3 seconds.

This timer value can be adjusted with configuration. To set the interval to 3 seconds, run the following:

ntk cfg set vrrp.services.nas.pulse_interval 3
ntk vrrp restart

Now the timeout for master failure detection will be 3×3 seconds — 9 seconds. This value should be the same on all nodes that serve this service using VRRP.

Remember to replace the service name (in this case nas) with your service name, if using custom services.