Big-IP DNS Question Regarding Connectivity Loss to LTM

NOTE: Due to issues posting this question last night I sent it via email to DevCentral support. They provided a detailed answer which is shown on the bottom of this post. I'm posting the question to see if there are other insights and/or in case it is helpful for someone else.

 

QUESTION:

 

I have significant experience with DNS and to a much, much lesser degree, Big-IP DNS (hereto referred to as GTM.. sorry). I'm trying to understand how GTM works in the scenario below. The question originated from a disagreement with another fellow whom I have great respect but think is wrong :)

 

There are two GTMs (Wide IP Only configuration) somewhere on the network and a single LTM at each of two data centers. Each LTM has a virtual server which is up and running throughout this scenario. The GTMs are configured with a WIP named xyz.co.com and are configured to respond robin between the two LTMs assuming they are healthy. Please assume everything is configured exactly as it should be and working as expected.

 

GTM1 / GTM2 hosts WIP xyz.co.com

 

Site A - LTM1 hosts VS 10.1.1.1

 

Site B - LTM2 hosts VS 10.2.2.2

 

Queries for xyz.co.com sent to either GTM receives a response alternating between 10.1.1.1 and 10.2.2.2 and life is good!

 

The scenario is a network problem occurs:

 

• It causes a complete loss of connectivity between GTM2 and both LTMs however GTM2 has full connectivity with GTM1.

   

• GTM1 is not affected and has full connectivity with both LTMs as well as GTM2.

   

Question: At this juncture, what would be the result of a query sent to GMT2 for xyz.co.com?

 

I'm thinking it will respond exactly as it did prior to the network issue because GTM2 is synced with GTM1 which knows that the VSs on both LTMs are up and available. My colleague is thinking GTM2 will not resolve the query because it can't see the LTMs.

 

Answer from Jason Rahm from DevCentral:

 

To answer your question, first, a GTM within a sync group is selected to be responsible for probing a path or server, then the responsible GTM must delegate a big3d to carry out the actual monitoring. It could choose itself, or it could choose another big3d such as the LTM in that datacenter. But…that is not a permanent choice, the monitoring is delegated to the big3d with the least iQuery queue workload. If GTM2 loses connectivity to all LTMs, then the responsible GTM would for sure be GTM1, and it would then hand out monitoring to itself or either LTM at this point, and with them having connectivity and assuming all servers OK the monitor being successful, that message would then be communicated over to GTM2, and it would respond successfully with GTM1 as configured. Now, there might be a brief moment of time where the GTMs are reassigning monitors where the actual monitor thresholds are exceeded, so there is a possibility of a fallback answer or timeout on the wip during transition.