Media recovery not required in case of instance failure during hot backup
Posted by Pavan DBA on April 25, 2012
If an instance crashes or is aborted while one of the tablespaces is in backup mode, the tablespace will appear to need media recovery upon startup. In normal instance failure situations, automatic crash recovery using the online redologs will be automatically performed, with no intervention required.
Because hot backup mode artificially freezes the checkpoint SCN in the datafile header, it will appear on startup after a crash to be a file restored from backup. Consequently, Oracle will suggest that you apply archived redologs from the time that the tablespace was placed in backup mode.
As I explained in my earlier post related to hot backup (https://pavandba.com/2012/04/23/what-happens-during-hot-backup/) the only difference between a datafile in backup mode and not in backup mode is the checkpoint SCN in the header. Because there is no data missing from the datafile, there is no need to apply recovery to it. For this reason, Oracle provides a way to clear the hot backup flags in the datafile header before you open the database.
If an instance will not start up automatically after a failure, and the reason is that a datafile needs recovery,then the v$backup view may be queried from mount mode. This view will show which datafiles are in backup mode. Based on this information, the following command may be used to take the files out of backup mode:
alter database datafile <filenum> end backup;