Ansicht
Dokumentation

Recovery after missing "end backup"
Tip-No.: 2138 ( INFO2138 )

Short text

Recovery after missing "end backup"
Tip-No.: 2138

Symptom

DB reports: ORA1113
file needs media revovery starting at log sequence ...

Cause

1st possibility: abend after online backup
2nd possibility: hardware error

Solution

1st possibility: Abend after online backup
------------------------------------------

During online backup, the tablespace that was just saved is marked.
The tool brbackup issues the command "Alter tablespace ... begin backup"
to the DB to do this. From now on, the current redo-log-number in the
header of the data files of this tablespace is frozen (in order to
ensure data consistency together with offline-redo-logs). After backing
up the files in thsi tablespace, the tool brbackup issues a "Alter
tablespace ... end backup".
If the tool brbackup is interrupted suddenly, before issuing the command
to end backup of the tablespace, then the redo-log-number remains
frozen.
The "... begin backup" and " ... end backup" are logged in trace file
/usr/ora//saptrace/background/alert_.log . After an interrup-e
ed data backup, the alert_.log must be checked for a possible
missing "... end backup".


a. Missing End-Backup is recognised before startup of a DB:
-----------------------------------------------------------

Simply issue "... end backup" in sqlplus:

sqlplus system/manager
SQL> alter tablespace PSAP... end backup;

(This command can be called repeatedly without harmful side-effect).


b. The error is only noticed when starting the DB:
--------------------------------------------------

ORACLE requires a recovery in this case, merely in order to update the
frozen up number.

sqldba
SQLDBA> shutdown abort
SQLDBA> startup dba mount exclusive
SQLDBA> recover database

(A "recover database" is preferrable here if other files are affected).

The DB now requires the file names (complete with path) of the
offline-redo-logs (archived redo-logs).
Import the required redo-logs from tape into sufficiently large file
system (see description of brbackup, brarchive in System Manual).
In this exceptional case, the recovery does not run from the time of
the last data backup, but rather from redo-log-file (online or
offline/archived) that was the currently written redo-log at the time
of the "Start backup...".

For every recovered offline-redo-log, the DB reports:
"log applied"
and awaits the specification of the next file.

After all necessary redo-logs have been re-run, the DB reports:
"recovery completed".

The database must now be re-started again:
SQLDBA> shutdown immediate
SQLDBA> startup


2nd possibility: hardware error
-------------------------------

-Check the affected file system (checkfs ...)
- Import the affected files from the last data backup. Created the
necessary symbolic links (soft links) for the directories, so that
the data files are not imported into the defective file system.
- Start a "recover database" as described for ORACLE.