HA Pair with heartbeat only on Mgmt Interface

[Evan Chadwick1709164034]

Hi All, 

With HA, with no failover interface settings, or failsafe enabled. If an HA pair only using the mgmt interface for Heartbeat, has the primary ADC mgmt experience a loss of connection to the switch, will it give up primary status and the Secondary will become primary? And Active Active will not occur?
(i'm aiming to improve the heartbeat situation, but just wanting to accurately write up the current risk). 

[Evan Chadwick1709164034]

Anyone?

[CarlStalhood]

If all heartbeats are not making it between the nodes, then the two nodes will try to become primary (split brain) because they can't see the other node. This is why it's best to send heartbeats across all interfaces.

[Johannes Norz]

On 12/7/2023 at 9:57 AM, Carl Stalhood1709151912 said:

If all heartbeats are not making it between the nodes, then the two nodes will try to become primary (split brain) because they can't see the other node. This is why it's best to send heartbeats across all interfaces.

Well, it depends. Usually, this situation occurs, if a switch goes down. In this case, interfaces connected to this switch go down, so the node connected to this switch goes down as well, as it doesn't consider itself to be healthy anymore. So no split-brain situation will occur. If both NetScalers are connected to this failed switch, both of them will go down unless you set "maintain one node ...".

 

A split-brain situation will arise in case the connection between the two switches goes down. So both nodes consider themselves to be healthy, as all their interfaces are still up (but their respective partners are down) and get primary.

 

Admins may cause this split-brain situation by turning off HA monitoring for the heart-beat interface. In this case, the interface is not relevant for the health state and both NetScalers will consider themselves to be healthy but their partner to be down and get primary.

[Evan Chadwick1709164034]

A poor mans fix, until I can get the network improved, could just be to enable failsafe and set one ADC to take primary active role if isolation occurs and set the other to go secondary. Agree?