DAG, high availability, disaster recovery

Question

we have 2 exchange servers server0.domain.com and server1.domain.com we have DAG with witness server on site. active database is on server0 Both servers are exchange 2016 on server 2016 However, i tried to model a disaster for one server being down. I put server1 offline, took the network off. Some computers (we have 120 clients.) say server not available. despite server0 is primary server and well computers try to access server1 and fail. Some access server0 and are OK. what to do?

problem only happens internally. Externally all is good.

Is there a way when one server being down and having DAG things be OK?

Otherwise what sense has a DAG?

please advise.

Answer 1

Well, if the load balancer is down, then yes you are dead until you fix it :)

The load balancer is a single name space that all the clients can connect to and you can mark down a server down manually or the load balancer will mark it down if its not available:

Managed Availability includes an offline responder. When the offline responder is invoked, the affected protocol (or server) is removed from service.
To ensure that load balancers do not route traffic to a Mailbox server that Managed Availability has marked as offline, load balancer health probes must be configured to check <virtualdirectory>/healthcheck.htm, for example, <https://mail.contoso.com/owa/healthcheck.htm>.

I recommend you read through the link I posted. If you decide you dont need a load balancer, then you will see the issues you describe and clients may lose connectivity to Exchange when one server is down. There are many options out there for 3rd party load balancers: F5, Kemp etc...
You can even use round robin DNS:

Third party vendors
Example:
https://kemptechnologies.com/microsoft-load-balancing/load-balancing-microsoft-exchange

Answer 2

DAGs provide high availability for mailbox databases, not client connections.

Hopefully you are using a load balancer for client connections

https://learn.microsoft.com/en-us/exchange/architecture/client-access/load-balancing?view=exchserver-2019

Answer 3

let me make clear.
having 2 exchange servers on site automatically needs load-balancer, is that what you mean?

Otherwise if any of them is down clients go here and there (either one server, or the other) so some outlooks always will jive that "server is not available"?

then we go to another problem - what if load balancer is down? then we are dead?

another inconvenience:

Lets say office network is 192.168.15.0/24

Outlook clients are all on this network and so is load balancer.

and Exchange servers both are on different network?

So load balancer acts basically as router?

This is quite complex.

I would prefer all devices being on single network