Forum Discussion

Brian_Saunders1's avatar
Brian_Saunders1
Icon for Altostratus rankAltostratus
Nov 21, 2013

GTM Global Availability Load Balancing to Down Pool

Hey All,

 

I've been working on configuring Wide IPs utilizing the Global Availability setting. I have two LTMs configured as servers on the GTM with the bigip monitor applied, so the GTM knows the status of the LTMs VServers. I've created my Wide IP and configured two pools with Global Availability enabled (on one of the LTMs I configured a bogus health check so that the .dev pool would go down):

 

 

Within the pools I have the preferred method set to Global Availability, alternate set to Round Robin, and Fallback set to Return to DNS. This shouldn't really matter since I only have one server defined in each pool - I only care about the wide IP level at this point? When I do a test ping to the DNS name "test1.gtm.domain.lab" it responds with the server that is down.

 

 

Seems like it's bypassing the Wide IP setting altogether and not taking into account that the vserver is down. Anyone have any suggestions?

 

Thanks,

 

Brian

 

application delivery
availability
BIG-IP DNS
deployment
design
f5
gtm
  • Bhanu_9561's avatar
    Bhanu_9561
    Nov 21, 2013

    I would expect that to work.

     

    This is how I implement it. Can you instead create just one pool with the two Virtual Servers in it? Associate this one pool to the Wide IP. At the pool level have the settings of Global Availability, Round Robin and Return to DNS. This configuration should achieve the functionality.

     

  • Bhanu_9561's avatar
    Bhanu_9561
    Icon for Cirrus rankCirrus
    Nov 21, 2013

    I would expect that to work.

     

    This is how I implement it. Can you instead create just one pool with the two Virtual Servers in it? Associate this one pool to the Wide IP. At the pool level have the settings of Global Availability, Round Robin and Return to DNS. This configuration should achieve the functionality.

     

    • Brian_Saunders1's avatar
      Brian_Saunders1
      Icon for Altostratus rankAltostratus
      Nov 21, 2013
      Bhanu - thanks for your assistance. After re-configuring the wide ip as you suggested and adding the two virtuals to the pool it worked as expected!
  • bkchng_148728's avatar
    bkchng_148728
    Icon for Nimbostratus rankNimbostratus
    Apr 16, 2014

    Hi!

     

    May I ask if this is a bug? I'm currently facing the same problem here with 11.2.1 HF9.

     

  • gsharri's avatar
    gsharri
    Icon for Altostratus rankAltostratus
    Jul 01, 2014

    It is not a bug. This behavior is by design. When a wide IP is configured to load balance across multiple pools and the desired behavior is for GTM to go to the next pool if its first choice is unavailable then that first pool should have its fallback lb method set to "none".

     

    Example 1:

     

    WIP: www.foo.bar; lb method: Global Availability

     

    Europe pool: VS1, VS2; lb methods RR, RR, Return to DNS

     

    North America pool: VS3, VS4; lb methods RR, RR, Return to DNS

     

    In the above scenario the wip lb method global availability will select Europe, the first pool on the list. If both VS1 and VS2 are unavailable GTM will use the fallback method, return to DNS (or whatever method it is set to), in an attempt to resolve the query while sticking with the first choice pool. GTM will not use the North America pool even if the Europe pool is unavailable. This is the effect of having a fallback lb method set in a pool. If the desired behavior is to use the North America pool when Europe is unavailable then the Europe pool must have fallback lb method set to none like this

     

    Example 2:

     

    WIP: www.foo.bar; lb method: Global Availability

     

    Europe pool: VS1, VS2; lb methods RR, RR, none

     

    North America pool: VS3, VS4; lb methods RR, RR, Return to DNS

     

    Now when the Europe VS are unavailable GTM will move on to the next pool on the list.

     

    Scott

     

Related Content