Forum Discussion

Davidfisher_345's avatar
Davidfisher_345
Icon for Altocumulus rankAltocumulus
Jun 26, 2018

When is the floating IP or anything used?

Hey all,

 

I created an ha pair in my lab on vm.

 

And i see the failover works even without a floating IP.

 

So when exactly do we need floating objects??

 

Thanks.

 

8 Replies

  • Floating self IP addresses are used for lots of things on a BIG-IP system, including as a possible default gateway for your pool members, as YossiV indicated. They are also used for source address translation addresses (SNAT) when Auto Map is specified.

     

    Why do you need floating self IP addresses? In an HA configuration, a floating self IP is shared between device group members, and will always point to the device that is active for its associated traffic group. Static (non-floating) self IP addresses are dedicated to a particular device and never point to anything but that device. For example, non-floating self IP addresses are used for monitor tests, as each device does its own monitoring, whether it's active or standby. Non-floating self IPs are also used for HA communication between devices in an HA device group. As a result, the MAC address of a non-floating self IP never changes. But you wouldn't want the default gateway on back-end servers (nodes) pointing to a non-floating self IP on a BIG-IP system as then traffic might get directed to a standby device. Same for SNATs. To support traffic processing through a BIG-IP HA configuration, the MAC address of a floating self IP address changes during failover to a MAC address on the active device. (This is why we say the IP address is "shared" between the device group members.)

     

    While you can initially set up an HA configuration without floating self IP addresses, you cannot successfully process traffic without floating self IP addresses. Defining an HA (sync-failover) device group simply involves identifying the non-floating self IP addresses that will be used by each device for HA communication. Eventually, you must define the floating self IP addresses that will be associated with traffic processing.

     

    • Hookah's avatar
      Hookah
      Icon for Nimbostratus rankNimbostratus

      Under the one-arm architecture , and ...

      1. I use SNAT AutoMap. The pool member's default gw is set to other L3 device.

      2. All http(s) traffic can re-establish connection when failover.

      3. In other words, I don't need to connection mirror, and I don't need the same Self-IP and MAC for SNAT AutoMap selection when failover occur.

      Based on the instructions above, is floating self-ip still necessary?

  • YossiV's avatar
    YossiV
    Icon for Nimbostratus rankNimbostratus

    floating ip is actully the VIP of the unit. and in some cases it also can be your pool members defult GW

     

  • Floats are used by virtual servers to talk to pool members (nodes). It is useful when you have have long running connections, so the mirrored sessions can persist after a failover. Otherwise, if you had to change IP addresses, session mirroring would be useless.

     

    Another situation is when you use BIG-IP as a network gateway, so you can have only one destination in you routing table being served by any active device in your cluster.

     

    Virtual addresses are always floating.

     

    Monitors use local Self IPs. Some management tasks can use local SelF IPs or the management port depending on your configuration.

     

    • Mayur_Sutare's avatar
      Mayur_Sutare
      Icon for MVP rankMVP

      Hi Kevin, yes you can! This shouldn't create any issue but just wanted to know what type of use case do you have for having such config?