Unique FQDNs for each server port when using GTM?
We are using GTM to perform global site load balancing across multiple datacenters to LTMs at each data center. Each LTM is responsible for load balancing requests for multiple applications across 2 servers. Each application is listening on different port.
At a high level, this is how I see the GTM/LTM works:
1) GTM has a WideIP that maps to a pool of virtual servers (the LTMs)
2) Eash new request will have its domain name resolved, by the GTM, to an ARecord of the selected LTM
3) the LTM will in turn selects an application server(final destination) registered in its pool
4) the consumer will communicate with the application server directly for subsequent requests until the DNS TTL expires.
Question:
if I have multiple application services hosted on a single server, each listening on different port. This server is replicated across all my data centers for high availability. to access each applicaition, using F5 GTM/LTM
a) is it mandatory to define multiple FQDNs, GTM and LTM pools for the solution to work?
for example to
reach each app I would use
www.myapp1.mycompany.com
www.myapp2.mycompany.com
www.myapp3.mycompany.com
b) or I can define 1 FQDN that maps to 1 GTM pool --> N LTM pools --> a selected server in a data center
For example to reach the same app I would use
www.myapps.mycompany.com:2001
www.myapps.mycompany.com:2002
www.myapps.mycompany.com:2003
The above (b) is what we are doing today with just the LTMs.
There are obviously gaps in my understanding of how GTM and LTM work but I hope you can understand my question and able to provide some comments on the 2 approaches.
TIA