What is mirroring?

Mirroring constantly synchronizes the transaction log of a mirror server with the transaction log of a principal server. The synchronization can be either synchronal to enable high availability, or asynchronal to enable high performance.

If set to high availability, a witness server is required. this can be either the principal server itself, or a 3rd server. If a 3rd server is chosen, this ensures more security in case of a failure.

To configure mirroring, the mirror server must restore a backup of the principal with NORECOVERY mode. Thus you cannot access the mirror database directly, since it remains in recovery mode. But you can create a snapshot of the mirror to read data. Also, the database must set to full recovery mode. (Note: if you want to have a full working copy of a database, then log shipping or replication is your choice over mirroring)

Mirroring requires endpoints (TCP or HTTP) for the principal, the witness and the mirror server. The endpoints are configured automatically, when you create mirroring in the properties options of the database.

The main purpose for mirroring however is to have a high available backup in case of an error of the principal server.  In High Availability mode, a failover is automatically performed to the mirror, when a failure occurs on the principal. Otherwise you need to launch a failover manually:

ALTER DATABASE <mirror_database> SET PARTNER FORCE_SERVICE_ALLOW_DATA_LOSS