1. To minimize downtime, we recommend Start the rolling upgrade by updating any mirroring partner that is currently the mirror server in all its mirroring sessions. You might have to update multiple server instances at this point.
2. For each mirroring session whose mirror server instance has just been upgraded, wait for the session to synchronize. Then, connect to the principal server instance, and manually fail over the session. On failover, the upgraded server instance becomes the principal server for that session, and the former principal server becomes the mirror server.
3. After you fail over, we recommend that you run the DBCC CHECKDB command on the principal database.
4. Upgrade each server instance that is now the mirror server in all mirroring sessions in which it is a partner. You might have to update multiple servers at this point.
5. Resume the mirroring session.
6. Upgrade any remaining server instance that is the witness in all its mirroring sessions. After an upgraded witness rejoins a mirroring session, automatic failover becomes possible again. You might have to update multiple servers at this point.
How to: Minimize Downtime for Mirrored Databases When Upgrading Server Instances