Posts Tagged ‘Oracle’

Recover an Oracle Database with Missing Archived Logs

December 22, 2012 Leave a comment

Scenario:  To recover an Oracle database from a backup with missing archived logs. The recovery process will stop at some point asking for archive logs when we try to recover from such a state.

The assumption here is that we have exhausted all possible locations to find another good and valid copy or backup of the archivelog that we are looking for.

If the archivelog is not found in any of the locations, then the approach and strategy on how to recover and open the database depends on the SCN (System Change Number) of the datafiles, as well as, whether the logsequence# required for the recovery is still available in the online redo logs.

For the SCN of the datafiles, it is important to know the mode of the database when the datafiles are backed up. That is whether the database is open, mounted or shutdown (normally) when the backup is taken.

If the datafiles are restored from an online or hot backup, which means that the database is open when the backup istaken, then we must apply at least the archivelog(s) or redolog(s) whose log sequence# are generated from the beginning and until the completion of the said backup that was used to restore the datafiles.

However, if the datafiles are restored from an offline or cold backup, and the database is cleanly shutdown before thebackup is taken, that means that the database is either not open, is in nomount mode or mounted when the backup is taken, then the datafiles are already synchronized in terms of their SCN. In this situation, we can immediately open the database without even applying archivelogs, because the datafiles are already in a consistent state, except if there is a requirement to roll the database forward to a point-in-time after the said backup is taken.

The critical key thing here is to ensure that all of the online datafiles are synchronized in terms of their SCN before we can normally open the database. So, run the following SQL statement, as shown below, to determine whether the datafiles aresynchronized or not. Take note that we query the V$DATAFILE_HEADER, because we want to know the SCN recorded inthe header of the physical datafile, and not the V$DATAFILE, which derives the information from the controlfile.

select status, checkpoint_change#,  to_char(checkpoint_time,

‘DD-MON-YYYY HH24:MI:SS’) as checkpoint_time, count(*)

from v$datafile_header

group by status, checkpoint_change#, checkpoint_time

order by status, checkpoint_change#, checkpoint_time;

The results of the above query must return one and only one row for the online datafiles, which means that they are already synchronized in terms of their SCN. Otherwise, if the results return more than one row for the online datafiles, then the datafiles are still not synchronized yet. In this case, we need to apply archivelog(s) or redolog(s) to synchronize all of the online datafiles. By the way, take note of the CHECKPOINT_TIME in the V$DATAFILE_HEADER,which indicates the date and time how far the datafiles have been recovered.

The results of the query above may return some offline datafiles. So, ensure that all of the required datafiles are online,because we may not be able to recover later the offline datafile once we open the database in resetlogs. Even though wecan recover the database beyond resetlogs for the Oracle database starting from 10g and later versions due to the introduction of the format “%R” in the LOG_ARCHIVE_FORMAT, it is recommended that you online the required datafilesnow than after the database is open in resetlogs to avoid any possible problems. However, in some cases, we intentionally offline the datafile(s), because we are doing a partial database restore, or perhaps we don’t need thecontents of the said datafile.

You may run the following query to determine the offline datafiles:

select file#, name from


where file# in (select file#

from v$datafile_header

where status=’OFFLINE’);

You may issue the following SQL statement to change the status of the required datafile(s) from “OFFLINE” to “ONLINE”:

alter database datafile <file#> online;

If we are lucky that the required log sequence# is still available in the online redologs and the corresponding redologmember is still physically existing on disk, then we may apply them instead of the archivelog. To confirm, issue the following query, as shown below, that is to determine the redolog member(s) that you can apply to recover the database:

set echo on feedback on pagesize 100 numwidth 16
alter session set nls_date_format = ‘DD-MON-YYYY HH24:MI:SS’;

select LF.member,, L.thread#, L.sequence#, L.status,L.first_change#, L.first_time, DF.min_checkpoint_change#
from v$log L, v$logfile LF,
(select min(checkpoint_change#) min_checkpoint_change#
from v$datafile_headerwhere status=’ONLINE’) DF
where =
and L.first_change# >= DF.min_checkpoint_change#;

If the above query returns no rows, because the V$DATABASE.CONTROLFILE_TYPE has a value of “BACKUP”, then try to apply each of the redolog members one at a time during the recovery. You may run the following query to determine theredolog members:

select * from v$logfile;

If you have tried to apply all of the online redolog members instead of an archivelog during the recovery, but you always received the ORA-00310 error, as shown in the example below, then the log sequence# required for recovery is no longer available in the online redolog.

ORA-00279: change 189189555 generated at 11/03/2007 09:27:46 needed for thread 1

ORA-00289: suggestion : +BACKUP

ORA-00280: change 189189555 for thread 1 is in sequence #428

Specify log: {<RET>=suggested | filename | AUTO | CANCEL}


ORA-00310: archived log contains sequence 503; sequence 428 required

ORA-00334: archived log: ‘+BACKUP/prmy/onlinelog/group_2.258.603422107’

After trying all of the possible solutions mentioned above, but you still cannot open the database, because the archivelog required for recovery is either missing, lost or corrupted, or the corresponding log sequence# is no longer available in theonline redolog, since they are already overwritten during the redolog switches, then we cannot normally open the database, since the datafiles are in an inconsistent state. So, the following are the 3 options available to allow you to open the database:


Force open the database by setting some hidden parameters in the init.ora. Note that you can only do this under the guidance of Oracle Support with a service request. As per Oracle Metalink,  there is no 100% guarantee that this will open the database. However, once the database is opened, then we must immediately rebuild the database.

Database rebuild means doing the following, namely:

(1) perform a full-database export

(2) create a brand new and separate database

(3) import the recent export dump.

This option can be tedious and time consuming, but once we successfullyopen the new database, then we expect minimal or perhaps no data loss at all. Before you try this option, ensure that you have a good and valid backup of the current database.

When recovery process is initiated using backup controlfile, it will output recovery succeeded but inorder to open the database the datafiles should be in consistent state.

SQL> recover database until cancel using backup controlfile;
ORA-00279: change 9867098396261 generated at 03/21/2008 13:37:44 needed for
thread 1
ORA-00289: suggestion : /arcredo/XSCLFY/log1_648355446_2093.arc
ORA-00280: change 9867098396261 for thread 1 is in sequence #2093Specify log: {=suggested | filename | AUTO | CANCEL}
ORA-01547: warning: RECOVER succeeded but OPEN RESETLOGS would get error below
ORA-01195: online backup of file 1 needs more recovery to be consistent
ORA-01110: data file 1: ‘/u100/oradata/XSCLFY/SYSTEM01_SCLFY.dbf’
ORA-01112: media recovery not startedSQL> alter database open resetlogs;
alter database open resetlogs
ERROR at line 1:
ORA-01195: online backup of file 1 needs more recovery to be consistent
ORA-01110: data file 1: ‘/u100/oradata/XSCLFY/SYSTEM01_SCLFY.dbf’

Now, a hidden parameter _ALLOW_RESETLOGS_CORRUPTION=TRUE will allow us to open database even though it’s not properly recovered.

Force open the database by setting the _ALLOW_RESETLOGS_CORRUPTION = TRUE. It allows us to open database but instance may crash immediately due to undo tablespace corruption. Check alert log file to view details of the issue.To resolve undo corruption issue, change undo_management to “Manual” in init.ora. Now the database will open successfully. Once database is up and running, create a new undo tablespace and drop the old corrupted undo tablespace. Also chang back the undo_management to “Auto” and undo_tablespace to “NewUndoTablespace” in init.ora._ALLOW_RESETLOGS_CORRUPTION=TRUE allows database to open without consistency checks. This may result in a corrupted database. The database should be recreated.



If you have a good and valid backup of the database, then restore the database from the said backup,and recover the database by applying up to the last available archivelog. In this option, we will only recover the databaseup to the last archivelog that is applied, and any data after that are lost. If no archivelogs are applied at all, then we can only recover the database from the backup that is restored. However, if we restored from an online or hot backup, then we may not be able to open the database, because we still need to apply the archivelogs generated during the said backup inorder to synchronize the SCN of the datafiles before we can normally open the database.


Manually extract the data using the Oracle’s Data Unloader (DUL), which is performed by Oracle Field Support at the customer site on the next business day and for an extra charge. If the customer wants to pursue this approach, we need the complete name, phone# and email address of the person who has the authority to sign the work order in behalf of the customer.

Source / Reference:

How to recover and open the database if the archivelog required for recovery is either missing, lost or corrupted?

How to recover and open the database if the archive log required for recovery is missing.


Recovering an Oracle Database with Missing Archived Logs

Resolving missing archive log gap at Standby Database

Standby Database

May 21, 2012 Leave a comment

Data Guard promises:

  • Disaster recovery
  • High availability
  • Data protection
  • Flexible balancing between data protection and performance requirements
  •  Simple management through the data guard broker.
  • Gap detection


1. Very low failure rate
All system components are duplicated. The primary and standby instances can run on different hosts. They can also have separate locations depending on the safety requirements.

2. Very short downtime

If an error occurs in the primary database system and you have to recover the database, you can perform the recovery very quickly on the standby host. You can avoid the time-consuming datafile restore, since these files are already located on the standby host.

The only thing you need to do is to import the last entries from the redo log files. Therefore, the standby instance can take over the tasks of the primary instance very quickly.

3. Significant decrease of the load on the production host

The database backup requires considerable resources and time for large databases. Since the backup can run on the standby host, the load on the primary instance is reduced significantly.
Therefore, the resources on the production host are fully available for production operation, and you do not need to interrupt or restrict database operation for a backup.

4. Consistency

Applying redo log files to the standby database immediately verifies their consistency. No other tool can achieve this level of verification.


1. High costs

For a standby database scenario, all system components need to be available in duplicate. In particular, duplicate hardware resources (CPU, hard disks, and so on) are expensive.

2. High system administration expense

You need to set up the standby host. If structural changes are made on the primary database system, you must make the required resources are available on the standby host. When the standby instance has taken over production operation – a “takeover” – you must set up a replacement standby database.

3. High requirements for switchover software

So that the standby instance can take over production operation, the appropriate switchover software is required. The user and the suppliers of the hardware and system software are responsible for selecting this software and making sure that it functions correctly.

Step-by-step instructions on how to create a Physical Standby Database on Windows and UNIX servers, and maintenance tips on the databases in a Data Guard Environment:

(Article by Hailie Jiao )

In this example the database version is The Primary database and Standby database are located on different machines at different sites. The Primary database is called PRIM and the Standby database is called STAN. I use Flash Recovery Area, and OMF.

a) Setup the environment

1. Make sure the operating system and platform architecture on the primary and standby systems are the same.
2. Install Oracle database software without the starter database on the standby server and patch it if necessary. Make sure the same Oracle software release is used on the Primary and Standby databases, and Oracle home paths are identical.
3. Test the Standby Database creation on a test environment first before working on the Production database.

b) On the Primary Database Side

1. Enable forced logging on your primary database:


2. Create a password file if it doesn’t exist.
1) To check if a password file already exists, run the following command:

SQL> select * from v$pwfile_users;

2) If it doesn’t exist, use the following command to create one:

On Windows:
 $cd %ORACLE_HOME%\database
 $orapwd file=pwdPRIM.ora password=xxxxxxxx force=y
 (Note: Replace xxxxxxxxx with the password for the SYS user.)
 $Orapwd file=pwdPRIM.ora password=xxxxxxxx force=y
 (Note: Replace xxxxxxxxx with your actual password for the SYS user.)

3. Configure a Standby Redo log.

1)  The size of the standby redo log files should match the size of the current Primary database online redo log files. To find out the size of your online redo log files:

SQL> select bytes from v$log;


2) Use the following command to determine your current log file groups:

SQL> select group#, member from v$logfile;

3) Create standby Redo log groups.
Primary database had 3 log file groups originally and 3 standby redo log groups are created now using the following commands:


4) To verify the results of the standby redo log groups creation, run the following query:

SQL>select * from v$standby_log;

4. Enable Archiving on Primary.
If your primary database is not already in Archive Log mode, enable the archive log mode:

SQL>shutdown immediate;
SQL>startup mount;
SQL>alter database archivelog;
SQL>alter database open;
SQL>archive log list;

5. Set Primary Database Initialization Parameters

Create a text initialization parameter file (PFILE) from the server parameter file (SPFILE), to add the new primary role parameters.

1) Create pfile from spfile for the primary database:

On Windows:
SQL>create pfile=’\database\pfilePRIM.ora’ from spfile;
(Note- specify your Oracle home path to replace ‘’).
SQL>create pfile=’/dbs/pfilePRIM.ora’ from spfile;
(Note- specify your Oracle home path to replace ‘’).
2) Edit pfilePRIM.ora to add the new primary and standby role parameters: (Here the file paths are from a windows system. For UNIX system, specify the path accordingly)

# Specify the location of the standby DB datafiles followed by the primary location;
# Specify the location of the standby DB online redo log files followed by the primary location 

6. Create spfile from pfile, and restart primary database using the new spfile.

Data Guard must use SPFILE. Create the SPFILE and restart database.

On windows:
SQL> shutdown immediate;
SQL> startup nomount pfile=’\database\pfilePRIM.ora’;
SQL>create spfile from pfile=’\database\pfilePRIM.ora’;
-- Restart the Primary database using the newly created SPFILE.
SQL>shutdown immediate;
(Note- specify your Oracle home path to replace ‘’).
SQL> shutdown immediate;
SQL> startup nomount pfile=’/dbs/pfilePRIM.ora’;
SQL>create spfile from pfile=’/dbs/pfilePRIM.ora’;
-- Restart the Primary database using the newly created SPFILE.
SQL>shutdown immediate;

c) On the Standby Database Site:

1. Create a copy of Primary database data files on the Standby Server:
On Primary DB:

SQL>shutdown immediate;

On Standby Server (While the Primary database is shut down):
1) Create directory for data files, for example, on windows,
On UNIX, create the directory accordingly.

2) Copy the data files and temp files over.

3) Create directory (multiplexing) for online logs, for example, on Windows,
E:\oracle\product\10.2.0\oradata\STAN\ONLINELOG and
On UNIX, create the directories accordingly.

4) Copy the online logs over.

2. Create a Control File for the standby database:
On Primary DB, create a control file for the standby to use:

SQL>startup mount;
SQL>alter database create standby controlfile as ‘STAN.ctl;

3. Copy the Primary DB pfile to Standby server and rename/edit the file.

1) Copy pfilePRIM.ora from Primary server to Standby server, to database folder on Windows or dbs folder on UNIX under the Oracle home path.
2) Rename it to pfileSTAN.ora, and modify the file as follows. : (Here the file paths are from a windows system. For UNIX system, specify the path accordingly)

 # Specify the location of the primary DB datafiles followed by the standby location
 # Specify the location of the primary DB online redo log files followed by the standby location
(Note: Not all the parameter entries are listed here.)

4. On Standby server, create all required directories for dump and archived log destination:
Create directories adump, bdump, cdump, udump, and archived log destinations for the standby database.

5. Copy the standby control file ‘STAN.ctl’ from primary to standby destinations

6. Copy the Primary password file to standby and rename it to pwdSTAN.ora.
On Windows copy it to \database folder, and on UNIX copy it to /dbs directory. And then rename the password file.

7. For Windows, create a Windows-based services (optional):
$oradim –NEW –SID STAN –STARTMODE manual

8. Configure listeners for the primary and standby databases.

1) On Primary system: use Oracle Net Manager to configure a listener for PRIM and STAN. Then restart the listener.

$lsnrctl stop
$lsnrctl start

2) On Standby server: use Net Manager to configure a listener for PRIM and STAN. Then restart the listener.

$lsnrctl stop
$lsnrctl start

9. Create Oracle Net service names.
1) On Primary system: use Oracle Net Manager to create network service names for PRIM and STAN. Check tnsping to both services:

$tnsping PRIM
$tnsping STAN

2) On Standby system: use Oracle Net Manager to create network service names for PRIM and STAN. Check tnsping to both services:
$tnsping PRIM
$tnsping STAN

10. On Standby server, setup the environment variables to point to the Standby database.


11. Start up nomount the standby database and generate a spfile.

On Windows:
 SQL>startup nomount pfile=’\database\pfileSTAN.ora’;
 SQL>create spfile from pfile=’\database\pfileSTAN.ora’;
 -- Restart the Standby database using the newly created SPFILE.
 SQL>shutdown immediate;
 SQL>startup mount;
 SQL>startup nomount pfile=’/dbs/pfileSTAN.ora’;
 SQL>create spfile from pfile=’/dbs/pfileSTAN.ora’;
 -- Restart the Standby database using the newly created SPFILE.
 SQL>shutdown immediate;
 SQL>startup mount;
 (Note- specify your Oracle home path to replace ‘’).

12. Start Redo apply

1) On the standby database, to start redo apply:

SQL>alter database recover managed standby database disconnect from session;

If you ever need to stop log apply services:

SQL> alter database recover managed standby database cancel;

13. Verify the standby database is performing properly:

1) On Standby perform a query:

SQL>select sequence#, first_time, next_time from v$archived_log;

2) On Primary, force a logfile switch:

SQL>alter system switch logfile;

3) On Standby, verify the archived redo log files were applied:

SQL>select sequence#, applied from v$archived_log order by sequence#;

14. If you want the redo data to be applied as it is received without waiting for the current standby redo log file to be archived, enable the real-time apply.

To start real-time apply:

SQL> alter database recover managed standby database using current logfile disconnect;

15. To create multiple standby databases, repeat this procedure.

d) Maintenance:

1. Check the alert log files of Primary and Standby databases frequently to monitor the database operations in a Data Guard environment.

2. Cleanup the archive logs on Primary and Standby servers.

I scheduled weekly Hot Whole database backup against my primary database that also backs up and delete the archived logs on Primary.

For the standby database, I run RMAN to backup and delete the archive logs once per week.

$rman target /@STAN;
RMAN>backup archivelog all delete input;

To delete the archivelog backup files on the standby server, I run the following once a month:

RMAN>delete backupset;

3. Password management
The password for the SYS user must be identical on every system for the redo data transmission to succeed. If you change the password for SYS on Primary database, you will have to update the password file for Standby database accordingly, otherwise the logs won’t be shipped to the standby server.

Refer to section b) 2, step 2 to update/recreate password file for the Standby database.

Source / Reference Links:

Database Schema Documentation

April 26, 2012 Leave a comment

Tools that provide options to generate database schema documentation are found all over web. But you may end up with a paid version of the tool in the end.

Oracle SQL Developer itself has an option to generate a HTML document on  database schema.

  • Open Oracle SQL Developer.
  • Create a connection to the desired database.
  • Right click on the connection and find ‘Generate DB Doc’ option.
  • Select the option and specify the target directory.

Other open source tools for database schema documentation:

Network Exports/Imports using Network_Link

March 27, 2012 Leave a comment

The NETWORK_LINK parameter identifies a database link to be used as the source for a network export/import. The following database link will be used to demonstrate its use.

CONN test/test
                    CONNECT TO scott IDENTIFIED BY tiger USING 'DEV';

In the case of exports, the NETWORK_LINK parameter identifies the database link pointing to the source server. The objects are exported from the source server in the normal manner, but written to a directory object on the local server, rather than one on the source server. Both the local and remote users require the EXP_FULL_DATABASE role granted to them.

expdp test/test@db10g tables=SCOTT.EMP network_link=REMOTE_SCOTT
        directory=TEST_DIR dumpfile=EMP.dmp logfile=expdpEMP.log

For imports, the NETWORK_LINK parameter also identifies the database link pointing to the source server. The difference here is the objects are imported directly from the source into the local server without being written to a dump file. Although there is no need for a DUMPFILE parameter, a directory object is still required for the logs associated with the operation. Both the local and remote users require the IMP_FULL_DATABASE role granted to them.

impdp test/test@db10g tables=SCOTT.EMP network_link=REMOTE_SCOTT
     directory=TEST_DIR logfile=impdpSCOTT.log
  • Ensure that the exporting user at the source database has the EXP_FULL_DATABASE role.This user must be specified when you create the database link.
  • Ensure that the importing user at the destination database has the IMP_FULL_DATABASE role.
  • Create and test a database link between the source and destination databases.
  • Run the following command, where import_user is the username for the importing user, and db_link is the name of the database link owned by the exporting user:
    IMPDP import_user/password NETWORK_LINK=db_link FULL=Y;
  • A log file for the import operation is written to the DATA_PUMP_DIR directory. You can discover the location of this directory by running the following command:
    SQL> select * from dba_directories where DIRECTORY_NAME like 'DATA_PUMP_DIR';

Source/Reference Links:

Oracle Database Running Out of Space ?

February 22, 2012 Leave a comment

Are you running out of space in the file system where database files (data files) are installed?

Try out this tip in order to move some of your data files from one drive to another and update the datafile location in your database.

The operating system: Oracle Enterprise Linux
The Database: Oracle Database 10g Enterprise Edition Release

In our environment we have installed our database’s files (data files) in /oracle/oradata/mydb location. The drive is getting up to 99% of utilization. Now we will move the system01.dbf from the above mentioned location to a new location in /oracle/hdb1/oradata/mydb location. /oracle/hdb1 is mounted from /dev/hdb1. The drive is nearly empty, that’s why I chose it.
Now for the real moving part, we will perform the following steps:
Login to SQL* Plus and shutdown the database
Logout from SQL* Plus and move the files from the source directory to destination directory.
Login to SQL* Plus as /nolog
Connect as SYSDBA to the database to an idle instance.
Issue startup mount command to bring up the database.
Issue ALTER DATABASE RENAME command to rename the data files from the source directory to destination directory.
Now finally open the database.
The above mentioned are the brief steps by which we can achieve our goal. Now to demonstrate see the following commands as well so that it serves as a live example:

Step 1
oracle@astrn10: sqlplus /nolog
SQL> conn /as sysdba
SQL> shutdown immediate;
Troubleshooting: If you are not able to get the SQL prompt, check your ORACLE_HOME, and ORACLE_SID.
Step 2
SQL> exit;
oracle@astrn10: mv /oracle/oradata/mydb/system01.dbf /oracle/hdb1/oradata/mydb/
Now check whether the file have been moved or not by issuing the following command:
oracle@astrn10: ls /oracle/hdb1/oradata/mydb/
total 429924352
-rw-r----- 1 oracle oinstall 429924352 Feb 12 11:00 system01.dbf
Now we are ready for the next step.
Step 3
oracle@astrn10: sqlplus /nolog
Step 4
SQL> conn /as sysdba
Connected to idle instance.
Step 5
SQL> startup mount;
Database mounted.
Step 6
SQL> alter database rename file '/oracle/oradata/mydb/system01.dbf' to '/oracle/hdb1/oradata/mydb/system01.dbf';
Database altered.
Step 7
SQL> alter database open;
Database opened.

That’s all. We are done with our agenda for moving data files from one drive to another. If this where Windows/any other operating system, then copy files as per your operating system commands/interface in Step 2.
In order to copy more files (in this demonstration we have moved only one file), then repeat Step #2 and Step # 6 for each file.
#End of tip


Running out of space? Want to move Oracle Datafiles? – Ask Anantha


Oracle Locks

February 21, 2012 Leave a comment


Query to identify the owner, object, object type, sid, serial number, status, OS user and machine to track locks:

select c.owner, c.object_name, c.object_type,b.sid, b.serial#, b.status, b.osuser, b.machine, a.session_id
 from v$locked_object a , v$session b, dba_objects c
where b.sid = a.session_id
and a.object_id = c.object_id;

More Queries:

s1.username || '@' || s1.machine 
|| ' ( SID,S#=' || s1.sid || ',' || s1.serial# || ' ) is blocking '
|| s2.username || '@' || s2.machine
|| ' ( SID,S#=' || s2.sid || ',' || s2.serial# || ' )'
AS blocking_status
v$lock l1,
v$session s1,
v$lock l2,
v$session s2
s1.sid = l1.sid
and s2.sid = l2.sid
and l1.BLOCK = 1
and l2.request > 0
and l1.id1 = l2.id1
and l2.id2 = l2.id2;
 v$locked_object a,dba_objects b
 a.object_id = b.object_id

Kill the session causing lock:

alter system kill session 'SID,SERIAL';
Eg: alter system kill session '323,1609';

Source / Reference Links:

Automatic Shared Memory Management – ASMM

February 19, 2012 Leave a comment

The Oracle Automatic Shared Memory Management is a feature that automatically readjusts the sizes of the main pools (db_cache_size, shared_pool_size, large_pool_size, java_pool_size) based on existing workloads for optimal performance.

About Automatic Shared Memory Management:

Automatic Shared Memory Management simplifies SGA memory management. You specify the total amount of SGA memory available to an instance using the SGA_TARGET initialization parameter and Oracle Database automatically distributes this memory among the various SGA components to ensure the most effective memory utilization. When automatic shared memory management is enabled, the sizes of the different SGA components are flexible and can adapt to the needs of a workload without requiring any additional configuration. The database automatically distributes the available memory among the various components as required, allowing the system to maximize the use of all available SGA memory. Oracle Database remembers the sizes of the automatically tuned components across instance shutdowns if you are using a server parameter file (SPFILE). As a result, the system does need to learn the characteristics of the workload again each time an instance is started. It can begin with information from the past instance and continue evaluating workload where it left off at the last shutdown.

The benefits of ASMM are:

  • Reduces the change of running out of shared pool memory
  • Uses available memory optimally
  • Improves database performance by constantly matching memory allocations and instance memory needs

Based on workload information, automatic shared memory tuning:

  • Captures statistics periodically in the background
  • Uses the different memory advisories
  • Performs ?what-if? analyses to determine best distribution of memory
  • Moves memory to where it is most needed
  • Has no need to configure parameters for the worst-case scenario
  • Resurrects component sizes from last shutdown if SPFILE is used

Oracle Automatic Shared Memory Management is enabled by setting:

  • a spfile used to specify init.ora values
  • sga_target parameter is set to a non-zero value
  • statistics_level parameter set to to TYPICAL (the default) or ALL (The other value of this parameter is BASIC, which will not allow changing the memory pools automatically.)
  • shared_pool_size must be set to a non-zero value

Oracle10g has introduced special double underscore hidden parameter to control ASMM:

  • __db_cache_size
  • __shared_pool_size
  • __large_pool_size

Once enabled, Oracle ASMM will morph the pool areas within the confines of sga_max_size.

When ASMM is enabled, then the following memory pools are automatically sized:

1 .Buffer cache (DB_CACHE_SIZE)
2. Shared pool (SHARED_POOL_SIZE)
3. Large pool (LARGE_POOL_SIZE)
4. Java pool (JAVA_POOL_SIZE)

The following pools are manually sized components and are not affected by ASMM.

3. DB_nK_CACHE_SIZE (non-default block size)

Reference Links: