I've setup a co-exist environment with Exchange 2010 and Exchange 2016.
At the moment mail-flow seems to work with no problems what so ever, and I have migrated two test-users from 2010 to 2016.
SMTP traffic and HTTPS is proxied from the new Exchange 2016 installation - and works fine for the 2010 users behind.
My problem is that when the users are migrated, outlook 2016 is having issues connecting to the new server.
When I open outlook, it still uses RPC/HTTP towards the old server.
If I delete the old profile, and re-create it using autodiscover, it works fine.
I then see MAPI traffic and it hits the new server as expected.
This would cause a problem if I manually have to re-create profiles for all the users in our company..
Does anyone have any pointers?
Edited with more information:
I migrated my own user, and I had my outlook open. I got the message that my box was migrated and that I needed to restart outlook; and so I did! My mobile-phone and OWA works like intended, and it's just my outlook clients acting up. It happened on two of my computers connected to the same mailbox and same user (home and office computer).
When opening my outlook client I get the certificate warning message with the name of my old servers internal name (like: Exch2010.domain.local), and the certificate in use is for our FQDN as mail.company.com.
Second Edit:
I just migrated a test-user from 2010 with outlook closed and tried connecting from an external network. It asked me the question if I wanted to allow https://mail.company.com/autodiscover/autodiscover.xml to edit my settings. I clicked "allow", and nothing happened. The mailbox stayed in a disconnected state and the connection status had one connection which had the status "established". The connection was proxied through mail.company.com and to Exch2010.company.local. I restarted outlook and the same thing happened.
Then I moved the client from the external network, and put it in our internal network instead. Now it gives me the same message "allow this website to configure migration.test@company.com server settings?" but with the local name of the NEW exchange server in an URL format like "https://exch2016.company.local/autodiscover/autodiscover.xml". I allowed and accepted the certificate warning (as it said it was using the local name again - exch216.company.local; which doesn't match the SSL cert).
Outlook is still not operational and is in a "disconnected" state. My own mailbox is "established" but doesn't update.
Out of curiosity I checked the servers by running:
Get-AutodiscoverVirtualDirectory | fl
InternalURL and ExternalURL for all three are blank. I'm not experienced enough to tell if this is correct or not.
For some reason, it seems like the servers are internally announcing their local names instead of the correct "mail.company.com".
I also checked the servers by running:
Get-ClientAccessServer -Identity SERVER | fl
All of them got the AutoDiscoverServiceInternalUri set to "https://mail.company.com/autodiscover/autodiscover.xml".
FQDN for all of them is set as their local hostnames (servername.company.local).
I have no idea what to try next..
Edit number three; as a reply to @Sembee: Hello @Sembee; thanks for your reply. I left them blank and did no changes. I checked the InternalURI for clientaccessserver on all three servers, and they are identical. The Autodiscover-tests complete without problem and have done so since the beginning. Re-configuring a users outlook makes the outlook work again (fresh data from autodiscover). All traffic are passing through the 2016 server for now (afaik), and proxies OK to the 2010 server. None have mentioned any problems. Outlook just isn't working after migrating to 2016 without re-creating the profile and running a new autodiscover.
Fourth edit: When I came in to work today, my own laptop (did not work yesterday/sunday) AND the laptop (did not work on saturday) I use for testing both worked fine. It could be some sort of delayed sync doing this to me? I'm currently setting up another test, to see if it behaves in the same way - and if its possible to time it somehow. Any pointers would be much appreciated.