The steps for recovering from primary server failure depend on the synchronization mode you are using for your database mirroring system.

If you are running in synchronous mode, then all of the transactions that are present on the primary server are also guaranteed to be committed on the mirror server. The mirror server can take over as the new primary server without any user intervention.

In asynchronous or asyncfullpage mode, failover from the primary server to the mirror server is not automatic because the mirror server may not have all committed transactions that were applied on the primary server. Unless you specified that autofailover should take place, when using one of the asynchronous modes, a mirror server, by default, cannot take ownership of a database when the primary fails. When the failed server is restarted, it detects whether transactions were lost. If transactions were lost, it writes a message to the database server message log and shuts down the database.

When starting the original mirror server as the new primary server, you have two options for getting the database files on both servers into the same state: