Forum Discussion

Christy_Love's avatar
Christy_Love
Icon for Nimbostratus rankNimbostratus
May 31, 2017

Maintaining Internal GTM POOL Connections with Active/Active Configuration

I currently have an active/active GTM configuration that uses 2 GTM pools within a single wide-ip. The problem I am having is maintaining my session (or a user maintaining their session) within the defined GTM pool and its defined members at one datacenter. The traffic will redirect to the other GTM pool at the other datacenter to access a pool member when a redirect within the application occurs (the redirects may be a different application and are their own defined LTM VS within the GTM Pool). Cross datacenter traffic/access to the GTM pool member instances is happening and is causing timeout issues. If I disable one of the GTM pool members within the wideip, everything works properly and stays internal to its datacenter.

 

Each GTM pool resides at 2 different datacenters and each is configured exactly the same. The application setup I am front-ending with the F5 setup is a bit complex in that a single wide-ip/gtm pool redirects to various LTM VS's within a single session. Can someone please help figure out what I have configured incorrectly? Below is an example of how I have things setup in my environment:

 

1- wideip -(Round Robin LB Method)-"Persistence" is disabled b/c it breaks the application, 2 GTM pool members (DC1-GTM_POOL_A and DC1_GTM_POOL_B)

 

DC1-GTM_POOL_A (load balancing is Preferred: Global Availability, Alternate: None, Fallback: Fallback IP with fallback IPv4 IP defined--I set the load balancing method up within the gtm pool in this fashion versus using Round Robin, Return to DNS, etc b/c I was hoping that my session would stay internal to its GTM Pool and that the defined Fallback IP and would perhaps cache itself on the local clients system)

 

Pool Members- DC1-LTM VS1 using port 9943 (TYPE-Performance Layer 4) DC1-LTM VS2 using port 9932 (TYPE- Standard) DC1-LTM VS3 using port 9933 (TYPE-Standard) DC1-LTM VS4 using port 9934 (TYPE-Standard)

 

DC2-GTM_POOL_B (load balancing is Preferred: Global Availability, Alternate: None, Fallback: Fallback IP with fallback IPv4 IP defined)

 

Pool Members- DC2-LTM VS1 using port 9943 (TYPE-Performance Layer 4) DC2-LTM VS2 using port 9932 (TYPE- Standard) DC2-LTM VS3 using port 9933 (TYPE-Standard) DC2-LTM VS4 using port 9934 (TYPE-Standard)

 

What my end result should be is client_1's session who is directed to DC1-GTM_POOL_A at DC1 stay at DC1 and access only the VS members within its GTM Pool at DC1 until their session ends. Client_2's session that is directed to DC2-GTM_POOL_B at DC2 stay at DC2 and access only the VS members within its GTM Pool at DC2 until the session ends. And so on, and so on....Is this possible?

 

I would appreciate any feedback from anyone who may have some experience with this.

 

Thanks,

 

Christy

 

3 Replies

  • Ideally, a WIP should be serving a single application. When I look at the GTM pool members, they seem to be serving different applications based on the different port numbers utilized. May be you should create multiple WIPs such that each application has its own unique WIP and relevant GTM pools with corresponding pool members. Just a suggestion to see if this helps your scenario.

     

  • Vijay...thank you for responding back. This is one "big" application with multiple components/modules so to speak. Everything works flawlessly if one GTM pool member is disabled and all client sessions are going to the one GTM pool. Once I introduce the other GTM pool member the traffic will go between both GTM pools for one single client session when a redirect in the application occurs..... ie...opening a form for example, or signing into the application (SSO), resulting in cross datacenter traffic and timeout issues.

     

    I need for a single client session, when accessing the application, to stay within its initial GTM Pool during the duration of the session.

     

  • If it is one application, why the different VS ports ? May be I am not fully understanding the set up. Do you have access to an F5 Solution Engineer who might be of better use since you may have to white-board the traffic flow.