Forum Discussion

CANSTAYN569's avatar
CANSTAYN569
Icon for Nimbostratus rankNimbostratus
Aug 25, 2016

Hard-wired serial cable necessity

Hi all,

 

I have a customer who has a pair of 1600 Big-ip's running as HA. They have network fail-over configured. I am pretty sure about that. Although they have the serial cable plugged as well. For some reason they might need to physically distribute the devices within the datacenter. And i wonder if they really need a long hard-wire serial cable for HA.

 

If the network failover is working properly, would i have any problem after unplugging the device? Kind regards

 

  • The hardwire failover is not required if you have network failover running. Since BIG-IP (somewhere in 11.x) changed the Device Service Clustering method, network failover is required for proper status reporting among the devices as far as the next-active device and other status. Last I read, the hardwire failover does provide faster (sub-second?) response to a system failure but that need depends on your application - network failover timeout is 3 seconds by default. SOL2397 provides good information about the two methods, as well as the note that the distance limitation on the serial (hardwire) cable is 50 feet.

     

    If you have network failover properly configured I do not see a problem with removing the hardwire failover. I'm a very cautious person, so if this is production I'd be sure that hardwire failover is disabled in the config, and go as far as wait for a maintenance window to 'pull the plug'.

     

    SOL2397 notes that if network failover traffic is lost but the hardwire shows a good connection, hardwire takes precedence and the system does NOT fail over. I cannot recall if the opposite is true: if the hardwire fails but network failover shows a good connection, if the system remains in current state.

     

    Note: I just checked, and SOL13819 describes the process for disabling hardwire failover. Apparently this requires a restart of the sod process, which results in a system failover. Another reason to do this during a maintenance period.