Nimbostratus
Employee
If your desire is to have the DNS requests routed to a specific GTM then that will pose a difficulty.
If, however, you want to have the resources (the actual website, for example) to which the DNS records point distributed by topology with the possibility of failover to the other side in the event of failover then you will have exactly the same wideip at both sites.
For example if I have two datacenters with the same website, one in the UK (200.200.200.200) and one in the US (100.100.100.100) and wanted to distribute them by topology so that EMEA users go to the UK and all others go to the US as long as both sites are up, but fail to the other site in the event of failover then I would use the following configuration:
wideip: www.mysite.com
(set up the topology record for EMEA following the documentation)
pool:
primary lb method: topology
secondary method: global availability
virtual server 100.100.100.100 <-- the US virtual server will be used by all who do not match the EMEA topo record
virtual server 200.200.200.200
This wideip would be exactly the same on both GTMs.
Now, when a user makes a DNS request they will be directed to the the UK resource if they are from EMEA but all others will use the US virtual. If for some reason either virtual server is unavailable, all users will be directed to the remaining available site.
This DOES have the implication that US users may make their DNS requests against the UK GTM. That is unavoidable. But it is a necessary evil if one considers that if the US GTM is down they need to be able to resolve against the one in the UK or your site will be entirely unavailable.