Open nielsduelund opened 5 years ago
Interesting that AWS is dead silent on your question, I feel their architecture is incomplete.
Based on their somewhat vague descriptions for post-deployment tasks and sample of connecting the client to a RDGW instance, I think their solution is to use manual load-balancing.
That is, assign a unique DNS name to each EIP (for example RDGW1.example.com
in the QS descriptions) and every time a client connects, let the client choose a name to connect to. This is of course a very kludgey and almost silly approach.
I have successfully deployed the quickstart. So far so good. My question is how to make "clients" choose a RDGW. As I see it, you passed this problem in the architecture drawing, by simply dragging a line from the client to both RDGW servers (one in each AZ). How does the auto scaling group (suddenly making more IP-adresses available) work in practice with regard to this. I haven't found any good solutions. My initial approach would be to use round robin DNS, but this solution seems not to be functional with Windows server 2012 onwards? I thought about using a load balancer, but as I understand it, UDP traffic is not supported by ELB. Am I simply overthinking it ? If I'm leaving out information, please ask me. Thanks, Niels