I have 2 servers set up in a simple DAG setup with my one witness server. Mail1 - Primary, Mail2 - Secondary, and FS1 - Witness server. 

Of course on the first server mail1 all of the external settings are populated for OWA, OAB, and so on and works fine. So what is the proper way to have external OWA, OAB, and so work also on mail2?  Lets say when I take the mail1 offline for or when the primary database moves to the 2nd server mail2?

You need load balancing for the client access services. That will provide high availability.

Effectively what happens is that there is a virtual DNS name used to represent the services. That name is configured in all of the client access URLs instead of the actual server names.

Load balancing is used to have clients connect to either server. If you have the money for it, you can use a hardware load balancer such as Kemp. Otherwise, you can use some free software load balancers such as ARR for IIS or haproxy. Finally, I've successfully used DNS round robin in several environments.

Hi,

There are two choices for multi-datacenter environment:

1. Deploy a unified namespace for the site resilient datacenter pair (unbound model). 
2. Deploy a dedicated namespace for each datacenter in the site resilient pair (bound model).

Generally, it is suggested to a unified namespace (mail.domain.com) by deploy Load Balance for two datacenters. In an unbound model, you have a single DAG deployed across the datacenter pair. This DAG has Mailbox servers in each datacenter typically all Mailbox servers are active and host active database copies, however you could deploy all active copies in a single datacenter. In this model, clients can connect to both datacenters in the event there is a WAN failure. All services (OWA, OAB etc.) can use mail.domain.com in two datacenters.

For more information about this, please refer to:

http://blogs.technet.com/b/exchange/archive/2014/02/28/namespace-planning-in-exchange-2013.aspx

Regards,

It looks like I might go with the hardware based load balancing device from Kemp. We are not that large of a company. Maybe 150 active users at a time for Exchange and the virtual Kemp device is about $1800 which is not a bad price, but everyone uses phones and Web App so being able to use both servers is a plus.

Right now I have both VM's of Exchange in the same data center on different hardware. Had one in another data center paid like $4000-$5000 to set it up took weeks to replicate 1TB and it broke a few months later with the witness server not working right if my one Internet connection went down that had the VPN to replicate the data all of the databases on both servers went down for some reason until the Internet was restored to reach the other server.

So we are moving to another disaster plan that will move our main VM's offsite if we have a failure that would happen our main servers will go active in less that 15 minutes with no more that 5 minutes of data lose and here is the best part you can test the backup live with out any data lose switching between both and you turn this on with a click of a button. It doesn't get much better than that. In this plan only 1 Exchange server the main one will move and the other one will go off

To be transparent - this does not sound like something that Microsoft directly supports.

"moving to another disaster plan that will move our main VM's offsite if we have a failure that would happen our main servers will go active in less that 15 minutes with no more that 5 minutes of data lose and here is the best part you can test the backup live with out any data lose switching between both and you turn this on with a click of a button

"

One other option to avoid the CAPEX for new LB and DC setup is to move to Exchange Online in Office 365.  Has that been a consideration?

No the disaster plan I am moving to is done at the VM level for VMware hyper visors only. The main issues with most peoples DR plan is getting the data back to your original data center after the switch. Their system does it this for you with no data lose and its very easy to test it. You run on your offsite VM's until the data is copied & merged back over to your data center and than switch back to your data center.

Exchange Online would cost us way to much money in monthly fees. I have a lot of old mail boxes I keep if we need data for some reason. We just renewed our licenses this year.

I am not found of full hosted services. Everyone wants a monthly fee for something. You don't own anything when you give up your network to the cloud. The only reason I like the DR plan we are going to is I have my own personal encrypted key to servers. If I forget no one can access my servers. 

Still the same thing.  That is not native Exchange replication, and as such not within our direct control. Thus not tested nor supported by Microsoft.  If you run into issues, then you call the people who built that solution.

I've seen this offered by other hosters over the years, trying to do this from Exchange 2000/2003 onwards. 

I am using this to replicate everything on my MS servers that run SQL, RDP, IIS and that. This is a total DR solution for everything not just Exchange. This is not to take the place of native Exchange replication. This is to take the place if my data center failed or lost power. All this solution is that it making an exact copy of your servers and keeping them running/staged and up to date at an another data center. MS would not even know if was happening, but you have to be running VMWare to do it not Hyper-V. I have seen this work and talk to many companies that use it. All of them love the solution.

MS hosted servers on MS cloud has to do something close to this to backup the clients servers to other data centers if MS has an outage or loses of a data center.