DAG failover causes users Outlook to disconnect from exchange

ideally the connectivity breaks and once the databases are mounted on the second node of the DAG, outlook should reconnect. if it does not then there is an issue.

you could try ExBpa to start with to find whther there are any misconfig in DAG

I have had microsoft check the DAG setup before for another issue and they said it was correct..
So I am not sure why Outlook gets disconnected when it fails over. Anyone else have any ideas?

Also have ran the EXBPA in the past and it does not find anything wrong, other than some older drivers, which is no big deal.

let me ask this, once the databases are mounted on node 2. do other clients like OWA , Async etc are able to connect ?

what abt the CAS server, is it in a load balanced if yes is it NLB or a hardware Load balancer?

Outlook clients are connecting to CAS Array which is composed of One or More Client Access Servers.

Therefore if your DAG is not running CAS roles as well then failover is not visible to Outlook clients and they remain connected to casarray.

I have one question to start with

1. Are your DAG Servers also running CAS Role and are you using those as CASArray?

~ Singh

The DAG has 2 servers with no hardware load balancer. So I thought the DAG created the software load balancer or is that wrong?

yes they are running the CAS role and not sure about the CAS array. How do I check that?

Run Get-ClientAccessArray on Exchange Mgmt Shell.

Do you have any Load balancer? If yes, is it Windows NLB or Hardware Load Balancer?

~ Singh

yes when I run that it shows the active server as the rpcclientaccess server.

There are 3 databases and they all three point to the same server.

This is what your problem is.

Outlook Clients will connect to the RPC Client Access Server set on the Database which is a CAS Server or CAS Array. CAS Array is nothing but a combination of 2 or more CAS servers in Load balancer.

DAG doesn't create any software load balancing for clients.

When you are taking your Main Server down, Outlook clients are unable to connect to this server and hence remain disconnected.

This is the solution for you.

 Add a Network Load balancer in Front of DAG/CAS Server, Create a CAS Array and resolve the IP of CAS Array to Load balancer. Therefore whenever you are taking any of your DAG server is going down, Load balancer will detect it and send request only to server which is Up.

OR

Everytime you take your Main Server down, you need to change RPC Client Access Server on all the Databases to point to other DAG/CAS Server

OR

Install a New Client Access Server and Set this New CAS Server as RPC client Access Server on all the Databases. Therefore your users will remain connected to New CAS Server and this server will automatically know which DAG server is having Database mounted through Active Manager.

~ Singh

The load balancer you speak of, can this be software or hardware and is there any good software load balancers out there that you know of?
 The option of installing the new CAS, do I need to just spin up a VM and then install the CAS role onto it?

software Load balancer is Windows Load Balancer but that's not going to help you because your DAG Servers are your CAS Servers.

There are a couple of Hardware load balancer like Cisco ACE, F5 or Big5 and list goes on.

Yes, you can install the CAS on VM and that will act as your RPC End point for all the Outlook Clients.

Only problem is, if your new CAS Server goes down, all client will disconnect even if DAG is working fine because New CAS Server is where your Outlook clients are connected.

~ Singh

Like windows network load balancing

Softwate Load balancer will help if you have CAS Servers on different server than Mailbox Servers

~ Singh

CAS servers and Mailbox servers are on both exchange servers in the DAG

This is what I said above. Software Load balancer is not going to help you in this case.

You need to have different CAS if you want to utilize software load balancing OR follow one of the above steps that I mentioned.

~ Singh

solved