HA Configuration (One in primary and One in DR)

i suggest dont extend the subnet over separate Layer2 (vlans etc.).
it will complicate l3 ip routing configurations.

adding to Paulius answer:
use different ip range for each DC then use dns/gtm to do dns based load balancing.
you still can make both bigip into SYNC-ONLY device group (not SYNC-FAILOVER), so each bigip has online backup of other's config.

To do that on prem you would be better off using them in standalone mode along with the module GTM and ip address unique to each DC for the various connected subsets and VIPs. Then if you really needed stuff to sync between them on the LTM level like for common non-unique stuff such as certificates, you could setup a partition that sync's between them. Though partitions have their own set of issues.

You certainly could get this to work if you are able to extend all the necessary L2 networks between the DCs for the F5, with say OTV, MPLS and like technologies at the switching/router level. Though that adds a lot more complexity to the surrounding network environment that would have to be contended with to ensure traffic routed and flows correctly.

I have though, seen this done in a cloud environment like AWS with the F5 Cloud Fail-over Extension (CFE) because you can use the built orchestration from those cloud providers with CFE to change where ip address are attached to at a given point in time. 
https://clouddocs.f5.com/products/extensions/f5-cloud-failover/latest/userguide/overview.html

Hello JR_TX , 

Interesting deployment. 

I believe it can be done but it some how complex. 

Please have a look in this article : https://my.f5.com/manage/s/article/K45921315

and let me know if this worked with you or not

If they each are on different subnets then those wouldn't be HA for LTM but you could make them HA for GTM.