I am in Berlin this week for an Administration Workshop II course. Unfortunately, it’s raining right now, so it would be no fun to stroll around the capital. Instead, there is time now for me to post something 🙂 Two major areas of that course are Performance Tuning and Backup & Recovery. One interesting feature that we discuss from the latter area is the ability to drastically speed up incremental backups (compared to versions before 10g) and to use these incremental backup for an actualization of level 0 image copies. Let’s look at an example:
SQL> alter database enable block change tracking using file '/home/oracle/bc.ora';
This command brings up the background process Change Tracking Writer (CTWR), that maintains a change tracking file. This file will contain the database block adresses of Oracle Blocks, modified after the next level 0 Full Backup. Because of this file, RMAN no longer has to scan all production datafiles and compare their blocks with the blocks of the level 0 backup in order to determine which blocks have changed. That was a very time consuming process before 10g. Now we will invoke RMAN with the following command:
RMAN> run {
recover copy of database with tag 'backrec';
backup incremental level 1 cumulative copies=1 for recover of copy with tag 'backrec' database;
}
Starting recover at 04-MAY-09
using target database control file instead of recovery catalog
allocated channel: ORA_DISK_1
channel ORA_DISK_1: sid=127 devtype=DISK
no copy of datafile 1 found to recover
no copy of datafile 2 found to recover
no copy of datafile 3 found to recover
no copy of datafile 4 found to recover
no copy of datafile 5 found to recover
no copy of datafile 6 found to recover
Finished recover at 04-MAY-09
Starting backup at 04-MAY-09
using channel ORA_DISK_1
no parent backup or copy of datafile 1 found
no parent backup or copy of datafile 3 found
no parent backup or copy of datafile 5 found
no parent backup or copy of datafile 2 found
no parent backup or copy of datafile 6 found
no parent backup or copy of datafile 4 found
channel ORA_DISK_1: starting datafile copy
input datafile fno=00001 name=/u01/app/oracle/oradata/orcl/system01.dbf
output filename=/u01/app/oracle/flash_recovery_area/ORCL/datafile/o1_mf_system_4zy6pnsg_.dbf tag=BACKREC recid=14 stamp=685996946
channel ORA_DISK_1: datafile copy complete, elapsed time: 00:00:35
channel ORA_DISK_1: starting datafile copy
input datafile fno=00003 name=/u01/app/oracle/oradata/orcl/sysaux01.dbf
output filename=/u01/app/oracle/flash_recovery_area/ORCL/datafile/o1_mf_sysaux_4zy6qqvp_.dbf tag=BACKREC recid=15 stamp=685996969
channel ORA_DISK_1: datafile copy complete, elapsed time: 00:00:25
channel ORA_DISK_1: starting datafile copy
input datafile fno=00005 name=/u01/app/oracle/oradata/orcl/example01.dbf
output filename=/u01/app/oracle/flash_recovery_area/ORCL/datafile/o1_mf_example_4zy6rk0w_.dbf tag=BACKREC recid=16 stamp=685996982
channel ORA_DISK_1: datafile copy complete, elapsed time: 00:00:07
channel ORA_DISK_1: starting datafile copy
input datafile fno=00002 name=/u01/app/oracle/oradata/orcl/undotbs01.dbf
output filename=/u01/app/oracle/flash_recovery_area/ORCL/datafile/o1_mf_undotbs1_4zy6rr46_.dbf tag=BACKREC recid=17 stamp=685996987
channel ORA_DISK_1: datafile copy complete, elapsed time: 00:00:03
channel ORA_DISK_1: starting datafile copy
input datafile fno=00006 name=/u01/app/oracle/oradata/orcl/users02.dbf
output filename=/u01/app/oracle/flash_recovery_area/ORCL/datafile/o1_mf_users_4zy6rv92_.dbf tag=BACKREC recid=18 stamp=685996987
channel ORA_DISK_1: datafile copy complete, elapsed time: 00:00:01
channel ORA_DISK_1: starting datafile copy
input datafile fno=00004 name=/u01/app/oracle/oradata/orcl/users01.dbf
output filename=/u01/app/oracle/flash_recovery_area/ORCL/datafile/o1_mf_users_4zy6rwc5_.dbf tag=BACKREC recid=19 stamp=685996988
channel ORA_DISK_1: datafile copy complete, elapsed time: 00:00:01
Finished backup at 04-MAY-09
Starting Control File and SPFILE Autobackup at 04-MAY-09
piece handle=/u01/app/oracle/flash_recovery_area/ORCL/autobackup/2009_05_04/o1_mf_s_685996989_4zy6rxvx_.bkp comment=NONE
Finished Control File and SPFILE Autobackup at 04-MAY-09If you look at the output above, you will bascially see that we did a level 0 Full Backup with Image Copies of all the datafiles. Now let’s modify some blocks in the database:
SQL> update sh.sales set amount_sold=amount_sold*1 where rownum<10000;
9999 rows updated.
SQL> commit;
Commit complete.
Now we run the same RMAN command again:
RMAN> run { recover copy of database with tag 'backrec';
backup incremental level 1 cumulative copies=1 for recover of copy with tag 'backrec' database; }
Starting recover at 04-MAY-09
using channel ORA_DISK_1 no copy of datafile 1 found to recover no copy of datafile 2 found to recover no copy of datafile 3 found to recover no copy of datafile 4 found to recover no copy of datafile 5 found to recover no copy of datafile 6 found to recover Finished recover at 04-MAY-09 Starting backup at 04-MAY-09 using channel ORA_DISK_1 channel ORA_DISK_1: starting incremental level 1 datafile backupset channel ORA_DISK_1: specifying datafile(s) in backupset input datafile fno=00001 name=/u01/app/oracle/oradata/orcl/system01.dbf input datafile fno=00003 name=/u01/app/oracle/oradata/orcl/sysaux01.dbf input datafile fno=00005 name=/u01/app/oracle/oradata/orcl/example01.dbf input datafile fno=00002 name=/u01/app/oracle/oradata/orcl/undotbs01.dbf input datafile fno=00006 name=/u01/app/oracle/oradata/orcl/users02.dbf input datafile fno=00004 name=/u01/app/oracle/oradata/orcl/users01.dbf channel ORA_DISK_1: starting piece 1 at 04-MAY-09 channel ORA_DISK_1: finished piece 1 at 04-MAY-09 piece handle=/u01/app/oracle/flash_recovery_area/ORCL/backupset/2009_05_04/o1_mf_nnnd1_TAG20090504T185902_4zy7ppmg_.bkp tag=TAG20090504T185902 comment=NONE channel ORA_DISK_1: backup set complete, elapsed time: 00:00:01 Finished backup at 04-MAY-09 Starting Control File and SPFILE Autobackup at 04-MAY-09 piece handle=/u01/app/oracle/flash_recovery_area/ORCL/autobackup/2009_05_04/o1_mf_s_685997943_4zy7pqr4_.bkp comment=NONE Finished Control File and SPFILE Autobackup at 04-MAY-09
Please notice the very short time it took RMAN to backup the (few) modified blocks during the incremental backup. Although this database is tiny of course, that would have taken a much longer time without block change tracking. The effect is way more dramatic in databases of more realistic sizes. Again we modify some blocks as before. After the third call of the command, we will see now always the following behavior:
RMAN> run {recover copy of database with tag 'backrec';
backup incremental level 1 cumulative copies=1 for recover of copy with tag 'backrec' database;}
Starting recover at 04-MAY-09
using channel ORA_DISK_1
channel ORA_DISK_1: starting incremental datafile backupset restore
channel ORA_DISK_1: specifying datafile copies to recover
recovering datafile copy fno=00001 name=/u01/app/oracle/flash_recovery_area/ORCL/datafile/o1_mf_system_4zy6pnsg_.dbf
recovering datafile copy fno=00002 name=/u01/app/oracle/flash_recovery_area/ORCL/datafile/o1_mf_undotbs1_4zy6rr46_.dbf
recovering datafile copy fno=00003 name=/u01/app/oracle/flash_recovery_area/ORCL/datafile/o1_mf_sysaux_4zy6qqvp_.dbf
recovering datafile copy fno=00004 name=/u01/app/oracle/flash_recovery_area/ORCL/datafile/o1_mf_users_4zy6rwc5_.dbf
recovering datafile copy fno=00005 name=/u01/app/oracle/flash_recovery_area/ORCL/datafile/o1_mf_example_4zy6rk0w_.dbf
recovering datafile copy fno=00006 name=/u01/app/oracle/flash_recovery_area/ORCL/datafile/o1_mf_users_4zy6rv92_.dbf
channel ORA_DISK_1: reading from backup piece /u01/app/oracle/flash_recovery_area/ORCL/backupset/2009_05_04/o1_mf_nnnd1_TAG20090504T185902_4zy7ppmg_.bkp
channel ORA_DISK_1: restored backup piece 1
piece handle=/u01/app/oracle/flash_recovery_area/ORCL/backupset/2009_05_04/o1_mf_nnnd1_TAG20090504T185902_4zy7ppmg_.bkp tag=TAG20090504T185902
channel ORA_DISK_1: restore complete, elapsed time: 00:00:02
Finished recover at 04-MAY-09
Starting backup at 04-MAY-09
using channel ORA_DISK_1
channel ORA_DISK_1: starting incremental level 1 datafile backupset
channel ORA_DISK_1: specifying datafile(s) in backupset
input datafile fno=00001 name=/u01/app/oracle/oradata/orcl/system01.dbf
input datafile fno=00003 name=/u01/app/oracle/oradata/orcl/sysaux01.dbf
input datafile fno=00005 name=/u01/app/oracle/oradata/orcl/example01.dbf
input datafile fno=00002 name=/u01/app/oracle/oradata/orcl/undotbs01.dbf
input datafile fno=00006 name=/u01/app/oracle/oradata/orcl/users02.dbf
input datafile fno=00004 name=/u01/app/oracle/oradata/orcl/users01.dbf
channel ORA_DISK_1: starting piece 1 at 04-MAY-09
channel ORA_DISK_1: finished piece 1 at 04-MAY-09
piece handle=/u01/app/oracle/flash_recovery_area/ORCL/backupset/2009_05_04/o1_mf_nnnd1_TAG20090504T190410_4zy80cd8_.bkp tag=TAG20090504T190410 comment=NONE
channel ORA_DISK_1: backup set complete, elapsed time: 00:00:03
Finished backup at 04-MAY-09
Starting Control File and SPFILE Autobackup at 04-MAY-09
piece handle=/u01/app/oracle/flash_recovery_area/ORCL/autobackup/2009_05_04/o1_mf_s_685998253_4zy80g2q_.bkp comment=NONEFirst RMAN actualizes the image copies of the datafiles in the recover copy section of the command. Then again a very fast incremental backup is done. In this fashion, we always get a quite fast backup and have our backup very close (in terms of time) to the production datafiles, so that a restore & recovery can be done fairly fast also. This picture illustrates the Recovery Area as well as the Recommended Backup Strategy that you have seen at work in this posting:
Conclusion: Incremental Backups can be taken very fast with the 10g New Feature Block Change Tracking. Together with Image Copies and the 11g New Feature Recover Copy this is the Recommended Backup Strategy for Oracle Databases. What is your reason why it is not implemented at your site?
Pickleball spielen 
#1 von Anand am Juli 24, 2009 - 01:10
Something new i learned.Thank you.
Regards,
Anand
#2 von prince oberoi am April 26, 2013 - 04:24
Well defined..
#3 von Tuba Altun am April 26, 2013 - 08:37
Thanks a lot for your sharing.
#4 von Uwe Hesse am April 26, 2013 - 18:07
Thank YOU guys for taking the time to leave a nice feedback 🙂
#5 von Krishna am Mai 5, 2013 - 16:17
Thank you for sharing use full information with us.
#6 von Bernhard am Januar 22, 2016 - 09:00
Hello Uwe Hesse,
Nice post!
Maybe you can also mention using block change tracking in an Data Guard environment you need to have Active Data Guard option. So when enabled on a Primary and you switch, your old Primary now (maybe unwanted) is using the ADG option. And the new Primary is not using the BCT. Unfortunate Oracle doesn’t take this into account and you need to be aware of this.
#7 von Uwe Hesse am Januar 28, 2016 - 11:12
Hi Bernhard, thank you for your comment! I am aware about what you said, but I think it’s futile for me trying to mention all the possible license traps you could fall into with Oracle Databases, so I focus almost exclusively on the technical part of it 🙂