In many situations a publication server can be automatically recovered. Because the distribution server retains a pointer to the last distributed transaction within the publication server's transaction log, the publication server can easily recover from off-line situations caused by events such as loss of power, loss of network connection, or minor hardware failure.
When the publication server returns online and the database is again available, the distribution log reader process locates the last distributed transaction and proceeds forward from that point, transferring replicated transactions into the distribution database. It is not necessary to drop and re-create publications.
However, if it becomes necessary to rebuild a published database on the publication server using an older backup tape, the distribution server may no longer contain a valid pointer to a publication server's last distributed transaction. If this situation occurs, the log reader task will log an error and reschedule itself in retry mode. (For information about retry mode, see Distribution Server Recovery., next.)