Forum Discussion

Peter_Z's avatar
Icon for Cirrus rankCirrus
Aug 16, 2011

HW Failover - what's behind it?

I'm curious about How the HW (serial-based) failover really works.


In their documentation or training materials, F5 only states that there is merely a voltage (no special data) passed over the serial cable and used to detect the failed peer and initiate failover. I wonder how the failover with only a Serial cable (no LAN failover configured) works in situations like:



- manual failover (force standby)



- VLAN or gateway failsafe


- etc




Since there's no failed box in such situations and no special packets are exchanged over the serial cable to announce issues like failed VLAN the only conclusion I made is that the Active BIG-IP box stops sending a voltage across the cable for a short time so the Standby peer detects a failed peer and become the Active unit.




Is this assumption correct?






5 Replies

  • From my understanding you are correct. The Serial failover literally just sends an electrical signal. As long as its detected on the other unit, there is no failover. The instant that breaks it fails.
  • nathe's avatar
    Icon for Cirrocumulus rankCirrocumulus



    Is this assumption correct? That's not what I understand.



    If I understand it correctly there is both detection and failover triggers. The detection can be either serial or network and is just a signal to see if the other f5 is there, whether by voltage signal or network packet. If the one can't see the other then it will assume (or remain of course) an active state.



    The failover triggers are processes running on the individual f5s which are monitored (high availability table) and, if fail, result in a certain action e.g. restart service, failover, reboot system etc...So, this process monitoring could cause an Active to go Standby, and vice versa.



    So both these scenarios work together to ensure a working setup.



    As for manual failover scenario, the serial detection can still see the other appliance is alive and well but it doesn't need to suspend the voltage signal. It doesn't see it as a failed peer. It's working on a software level, rather than a hardware one.



    Hope my understanding is correct.






  • Even if it's at the software level, you have to have software-to-software link between the distant boxes (in a form of specially forged messages announcing the Active box presence or change of state etc). Otherwise you will end up with two Standby boxes.



    But you have no such connectivity with only serial cable - only a change at the physical layer can be detected.
  • Hi, I am new to this community.



    Could anyone help me with a good material on serial fail over configuration on F5 LTM.



  • Not sure if this helps. Also, I'm sure if you use Network & Serial failover, both must fail in order for a failover to occur but that may have changed;



    "Serial Failover is used to failover between two devices in an active/standby HA Pair or DSC Device Group with only two members; it cannot be used with VIPRION or an active/active HA Pair. It is the recommended failover method due to its speed; failing over in under 1s when a relevant event occurs and causes the loss of the ‘heartbeat’ on the cable. Except in the case of a power failure on the primary unit, failover will not actually occur between devices unless additional HA features such as Fail-safe or a HA Group are configured to trigger it.



    The maximum serial cable length is 50 feet; devices any further apart must use Network Failover.



    Where possible, Connection, Persistence and SNAT mirroring should be used to minimise the effects of failover and failback on client connections and traffic flow."