Question on client connectivity with network outage at one data center

Question

We will be moving from Exchange 2010 to 2016 shortly and moving to a design that will span both data centers. Our DCs are connected by a P2P primary connection, and they are both on our Meraki network as a secondary connection. Our field sites are all on Meraki. The plan is to have 4 copies of each database, and to have active databases in each data center. The question came up - what happens to clients connecting to an Active database where the Meraki is down in DC1 where the active database lives? As the DCs are connected by the P2P, there should not be any failovers triggered. If the client connects to a mailbox server in DC2, does it proxy the connection to the active database in DC1 over the P2P?

We have our existing 2010 environment in one DC behind an F5 for the CAS. Given the above, does it make sense to stay with leveraging the F5 or revert back to round-robin DNS for client access?

Answer 1

We have the F5 spanning both data centers. But if the P2P is up, won't the health check made for the F5 members always come up Healthy since the Meraki being down in DC1 won't affect connectivity between DC1 and DC2?

Answer 2

Hardware load balancers are usually robust solutions that provide the most capacity and can be configured to load balance in many ways. So agree with AD-7937 that it's recommended to try staying with F5.

Understanding Load Balancing in Exchange 2010

As mentioned by AD-7937 and also indicated in the article above, "most hardware load balancer vendors have detailed documentation about how their product works with Exchange 2010", you could look for some documentation from F5 side for more information.

Answer 3

So if the F5 spans to both Data Centers, then you are in good shape.
So if the Meraki is down to DC1, but still active to DC2, then the clients will never notice. They will simply connect to the F5 in DC2 since its all part of the same namespace, the F5 will then route to an available Client Access Service running on any Exchange Server it can connect to and that will be proxied to the mail server where the mailbox lives.
https://techcommunity.microsoft.com/t5/exchange-team-blog/client-connectivity-in-an-exchange-2016-coexistence-environment/ba-p/603945

You just need ensure you have health checking enabled on the F5 to mark any server down in the pool:
https://learn.microsoft.com/en-us/exchange/architecture/client-access/load-balancing?view=exchserver-2019
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 would also review F5 documentation. They have alot in regards to Exchange

Answer 4

FSW will be at a 3rd site, connecting to both sites of the Meraki. The F5 does span both DCs.

What's the best way to direct clients to DC2? Would this mean something done on the F5?

Answer 5

1. Where is the File Share Witness?
2. for the client connectivity, I would continue to use the F5. I assume its spanning the data centers? ( i.e the namespace)

If the connectivity is good between the data centers and Exchange Servers, then yes, the connection will proxy from DC2 to DC1 - assuming you have accounted for how to ensure the clients now go through DC2 and not attempt to use DC1. Round Robin can do this, but its a poor substitute for a real load balancer like an F5.