For more information regarding the security incident at F5, the actions we are taking to address it, and our ongoing efforts to protect our customers, click here.

Forum Discussion

fubarSUSHI's avatar
fubarSUSHI
Icon for Altocumulus rankAltocumulus
Apr 06, 2014

Design question #2: Active-Standby pair in two different data centers

The question that got brought up is if a LTM pair (two devices paired) can be in two different data centers with a point to point 100mb line in between?

 

The request imho is "network-wise" not the most efficient but the plan is to have the Active side be in the PROD datacenter and the Standby side be in the DR datacenter. Can this actually be done? Obvious questions to me in which Im still gathering information:

 

  1. Are they bridging the vlans? Is the same l3 vlan being bridged over to the DR DC?
  2. Is it two different vlans? I think this will not work at all right? Just for the HA traffic alone requires broadcast traffic and it will not be able to go across a L3 domain. Is there a way to change all HA traffic to unicast traffic? Do I need to create a VRF?
  3. Has anyone else done this in two different datacenters or vlans for this matter? Im just unsure if this will work between two different vlans or even bridged vlans... isnt there a latency issue I need to worry about?
  4. Talking about latency... is there a latency maximum that I need to be concerned about?

Thanks for the help in advance...

 

application delivery
deployment
design
LTM

6 Replies

  • Luke_Lehman's avatar
    Luke_Lehman
    Icon for Employee rankEmployee
    Apr 06, 2014

    OTV is popular to discuss, but sometimes complex to run/maintain. Keeping different VLANs between locations is preferable for numerous reasons. Ease of support, prevention of "trombone-ing", etc.

     

    First off, what's the latency? If may be too high out of the gate which will make the rest of the questions moot.

     

  • kunjan's avatar
    kunjan
    Icon for Nimbostratus rankNimbostratus
    Apr 06, 2014

    The other option to extend L2 over L3 prior to OTV in cisco world is L2TP. While OTV helps to protect from broadcast storm going across WAN and bridging loops, L2tp might work for your case as it's a point-to-point 2DC setup. Again all depends on your full topology as in how users are connected to DR DC in the event of failure.

     

    While OTV kind of technologies works well in DR simulated environment when both DCs are up, in the real event of PROD DC going down, consider the OTV box in Prod DC might not be available which needs to be paired with DR DC OTV.

     

    Ideally you won't be having HA active and standby across Prod and DR(due to the above issues). Instead, IMHO, you will be having a GTM with LTM in the 2 DCs.. and GTM will do the job of directing the user traffic to appropriate LTM based on the scenario you have.

     

    • fubarSUSHI's avatar
      fubarSUSHI
      Icon for Altocumulus rankAltocumulus
      Apr 06, 2014
      L2TP and Bridging are the same things to me. OTV is a huge re-arch and Im not going there. lol This is a friendly thing... not a "let me step in and do it for you"
  • kunjan_118660's avatar
    kunjan_118660
    Icon for Cumulonimbus rankCumulonimbus
    Apr 06, 2014

    The other option to extend L2 over L3 prior to OTV in cisco world is L2TP. While OTV helps to protect from broadcast storm going across WAN and bridging loops, L2tp might work for your case as it's a point-to-point 2DC setup. Again all depends on your full topology as in how users are connected to DR DC in the event of failure.

     

    While OTV kind of technologies works well in DR simulated environment when both DCs are up, in the real event of PROD DC going down, consider the OTV box in Prod DC might not be available which needs to be paired with DR DC OTV.

     

    Ideally you won't be having HA active and standby across Prod and DR(due to the above issues). Instead, IMHO, you will be having a GTM with LTM in the 2 DCs.. and GTM will do the job of directing the user traffic to appropriate LTM based on the scenario you have.

     

    • fubarSUSHI's avatar
      fubarSUSHI
      Icon for Altocumulus rankAltocumulus
      Apr 06, 2014
      L2TP and Bridging are the same things to me. OTV is a huge re-arch and Im not going there. lol This is a friendly thing... not a "let me step in and do it for you"
  • What_Lies_Bene1's avatar
    What_Lies_Bene1
    Icon for Cirrostratus rankCirrostratus
    Apr 06, 2014

    A few things;

     

    1) I think the failover heartbeat is 1 per second and failover occurs if there are three 'lost' probes in a row. So, latency would need to be very high for this to be an issue. Network failover should work fine over very long distances.

     

    2) Don't forget VXLAN is supported on the box.

     

    3) Note, as far as I'm aware, failover doesn't require L2 adjacency, it's unicast UDP so should work fine if it's routed between devices.

     

    4) I'd agree it's not ideal to split an active and standby in this way.