Forum Discussion

Malcolm_Salmon1's avatar
Icon for Nimbostratus rankNimbostratus
Sep 03, 2013

LTM Active/Active vs Active/Standby?



I'm currently in the process of deploying several pairs of LTMs as HA pairs and am looking at the pros and cons of running an Active/Active vs Active/Standby configuration. I've had a read around and from what I understand the Active/Active mode of operation is based on associating VIPs with traffic groups and having one traffic group active per LTM.


F5 recommend running HA pairs as Active/Standby but I cant find anything that backs this up, i.e. there are no reasons given as to why this is. The only one I can really think of is that careful capacity management will need to be undertaken to ensure that each box doesnt run at over 50% of utilisation.


Are there any other technical or operational reasons why LTMs should not be run as an Active/Active pair?






7 Replies

  • if using active-active mode. can VIP-A in ltm-1 be confgured as active and then, VIp-A in ltm-2 as passive ? which means if VIP-A is down, the VIP-2 will be up in ltm-2...


    if can,where to configure this? thx u


  • Depend on your Upstream switching. If Upstream switches configured with vPC, active/active from LTM make sense. And most importantly you must configure SNAT properly, not "automap" use couple of IPs as a SNAT-Pool.


  • Malc,


    F5 don't recommend active/active on LTM's. Yes you need to implement Automap SNAT on the VIP's to ensure return path routing. To ensure the second unit responds to the ARP request to need to fix this in the configuration. To do so in old bigpipe: bigpipe virtual address x.x.x.x unit 2 or TMSH: tmsh modify ltm virtual-address x.x.x.x unit 2


    As I say F5 don't recommend this and it is sort of crude load balancing.






  • One arm is really just a routing thing, so just make sure you've got good floating IPs (used for Automap SNAT) in your traffic groups.


  • Thanks for the response. So it sounds like active/active isn't such a big thing now, it's just a case of keeping on top of the capacity management.


    Out of interest do you know if there are any additional considerations that need to be looked at when deploying Active/Active in one arm mode?




  • Prior to traffic groups, setting up an active/active HA pair was a little daunting. All of the VIPs/configs were identical on both boxes but some were disabled on each peer. You needed a failover event and mechanisms to both 1) enable the disabled VIPs, 2) ensure that VIPs weren't live on both boxes if the other unit came back up, and 3) ensure that a single box could handle all of the load. The recommendation to do active/standby was really motivated more by efficiency and simplicity of configuration. That of course has changed with 11.x with DSC (device services clustering) and traffic groups.


    You still have to manage capacity though.