We have a 3-node Exchange 2013 CU8 DAG.

Two servers are located in site A

One server is located in site B

The two servers in site A have active databases, and site B just contains passive copies for failover purposes.  Over the weekend, I ran the commands below and patched the server in site B.  That went successfully.  I then went to site A, and on one of the servers, prior to running the commands below, activated each database individually on the server in site B so I could get all active databases off during patching.  

As I activated each database through ECP on the server in site B, it reported successful.  However, after a period of time (just a few minutes) I started noticing that each database was failing back over to it's preferred copy.  

I wasn't able to locate much in the logs.  On the passive server where I was activating the database copies, I did notice some errors with "no best copy found".  However, after failing back, all passive copies and active copies are healthy.  

I can't determine why Exchange would have automatically failed the database back over.  Any ideas where I should be looking?

Commands:

1. Verify the powershell execution policy is unrestricted.  Use set-executionpolicy executionpolicy unrestricted, if necessary
2. Verify SERVERNAME is not the PAM (primaryactivemanager).  Use get-databaseavailabilitygroup status identity DAG1|fl name,primaryactivemanager to determine this.
3. If necessary, move the PAM.  Use move-clustergroup cluster DAG1 name cluster group node:OTHERSERVER to move it.
4. Set-servercomponentstate SERVERNAME component hubtransport state draining requester maintenance
5. Restart-service msexchangetransport
6. Restart-service msexchangefrontendtransport
7. Redirect-message server SERVERNAME target OTHERSERVERNAME.FQDN
8. Suspend-clusternode SERVERNAME
9. Set-mailboxserver SERVERNAME databasecopyactivationdisabledandmovenow $TRUE
10. Verify the active databases on this server are moved to another server.  If not, use the ECP to perform a server switch over.
11. Set-mailboxserver SERVERNAME databasecopyautoactivationpolicy blocked
12. Set-servercomponentstate SERVERNAME component serverwideoffline state inactive requester maintenance
13. Install CU8
14. Set-servercomponentstate SERVERNAME component serverwideoffline state active requester maintenance
15. Resume-clusternode SERVERNAME
16. Set-mailboxserver SERVERNAME databasecopyactivationdisabledandmovenow $FALSE
17. Set-mailboxserver SERVERNAME databasecopyautoactivationpolicy unrestricted
18. Set-servercomponentstate SERVERNAME component hubtransport state active requester maintenance
19. Restart-service msexchangetransport
20. Restart-service msexchangefrontendtransport

• Edited by Frentic 17 hours 36 minutes ago

Hello

Did you check if the B site server is set to lagged copies.

Get-MailboxDatabaseCopyStatus <DatabaseCopyName> | Format-List