B35 all you wanna know about rman by francisco alvarez

Advanced
Backup & Recovery

By:
Francisco Munoz Alvarez

About the Speaker

Francisco Munoz Alvarez
Oracle ACE Director
CLOUG (Chilean Oracle Users Group) President
LAOUC (Latin American Oracle Users Group Council) President
NZOUG (New Zealand Oracle Users Group) President
IOUC LA Spokesperson
8/9/10g/11g OCP, RAC OCE, AS OCA, E-Business OCP, SQL/PLSQL OCA, Oracle 7 OCM
Oracle 7, 11GR2 and OVM 3.1 Beta Tester
ITIL Certified
2010 ACE Director of the year by Oracle Magazine

Blog: http://oraclenz.wordpress.com - Email: mbatec@hotmail.com – Twitter : fcomunoz

Revera: Oracle Professional Services Manager

Grow up Mature
Born here Got Married Here Now Living here
here here

In this session…
• Point-in-time recoveries
• RMAN and Flashback database
• RMAN and encryption
• Performance tuning RMAN operations
• Database duplication
• Backup based database duplication
• Active database duplication (11g)
• Tablespace point-in-time recovery
• Reporting from the recovery catalog

RMAN Point-in-Time Recovery
• You can recover your database to a specific:
• Point-In-time
• System Change Number
• Log Sequence Number

• Useful for
• Restoring after a major error.
• Restoring development databases.
• Restoring from a lost online redo log.

RMAN Recovery
Incomplete Recovery – About Resetlogs

• It can be confusing sometimes, having to figure
out when to use resetlogs.
• In general, the database will always tell you
when you must use resetlogs.
• All incomplete recoveries will require the use of
the resetlogs command.
• If you restore a backup control file, you will need
to use the resetlogs command.
• In Oracle9i you should immediately backup the
database after a resetlogs operation.

RMAN Recovery
Incomplete Recovery

• RMAN supports incomplete recovery of your
Oracle database.

• RMAN supports
• Time based recovery
• SCN based recovery
• Log sequence based recovery

RMAN Recovery
Incomplete Recovery – Time based recovery

• Use the restore and recover commands using
the until time parameter to do a time based
restore.

• Example:
startup mount;
restore database UNTIL TIME
"TO_DATE('06/28/10 13:00:00','MM/DD/YY HH24:MI:SS')";
recover database UNTIL TIME
"TO_DATE('06/28/10 13:00:00','MM/DD/YY HH24:MI:SS')";
alter database open resetlogs;

RMAN Recovery
Incomplete Recovery – Change based recovery

the until time parameter to do a restore based on
a specific log sequence number.
• You can find log sequence numbers in the
v$log_history view.

• Example:
startup mount;
restore database UNTIL sequence 504033;
recover database UNTIL sequence 504033;

RMAN Recovery
Incomplete Recovery – SCN based recovery

the until time parameter to do a SCN based
restore.
• You can find the current SCN in the v$database
column current_scn.

• Example:
startup mount;
restore database UNTIL SCN 844039292;
recover database UNTIL SCN 844039292;

RMAN Recovery
Incomplete Recovery – How Far Back Can I Go?

• What is required to complete a database point-
in-time recovery?
• Backup of database datafiles
• All archived redo logs generated during
backup until the desired point of recovery.

RMAN Recovery

• You can use the list command to determine
how far back you can restore the database.
• First, set the nls_date_format to see the
date/time
Set nls_date_format=mm/dd/yyyy hh24:mi:ss

RMAN Recovery

• Start RMAN and use the list backup of
database summary command to see how far
back you can restore your database.
RMAN> list backup of database summary;

using target database control file instead of recovery catalog

List of Backups
===============
Key TY LV S Device Type Completion Time #Pieces #Copies
------- -- -- - ----------- ------------------- ------- -------
80 B F A DISK 11/05/2008 11:03:08 1 1
88 B F A DISK 11/05/2008 13:36:38 1 1
92 B F A DISK 11/05/2008 13:41:59 1 1

• Oldest backup

RMAN and Oracle Flashback
Technologies

RMAN and Flashback Technologies

• RMAN supports the Flashback Database
functionality with the flashback database
command.


• Must have configured the database for flashback
database operations

• Flash recovery area configured
• db_recovery_file_dest
• db_recovery_file_dest_size

• Set the flashback retention target (1440 default)
• db_flashback_retention_target


• Flashback logging configured

• To enable flashback you should perform a
clean shutdown
shutdown immediate;

• Mount the database
Startup mount

• Enable Flashback
alter database flashback on;


A inconsistent shutdown can lead to the following error:

SQL> alter database flashback on;
alter database flashback on
*
ERROR at line 1:
ORA-38706: Cannot turn on FLASHBACK
DATABASE logging.
ORA-38714: Instance recovery required.

RMAN and Flashback
Technologies
• Open the database

Alter database open;

• You can determine if the database is using flashback mode:

• V$DATABASE column FLASHBACK_ON

SQL> select flashback_on from v$database;

FLASHBACK_ON
------------------
YES

OEM Supports execution of these tasks too.

RMAN and Flashback
Technologies
• You can use the v$flashback_database_stat view to
determine how far you can flashback your database.

select begin_time, flashback_data,
db_data, redo_data, ESTIMATED_FLASHBACK_SIZE
from v$flashback_database_stat;

• You can use the view v$flashback_database_log to
determine how much space the recovery area requires
for flashback logs:

select estimated_flashback_size
from v$flashback_database_log;


• Using RMAN to flashback your database

RMAN> shutdown immediate

RMAN> startup mount

RMAN> FLASHBACK DATABASE
TO TIME=”TO_DATE('01/11/2009 12:40:00','MM/DD/YY HH:MI:SS')”;

RMAN SQL ’alter database open resetlogs’

• Don’t forget, you can open the database read only before
you complete the flashback operation.
• Make sure you have flashed back to the correct point in
time.

RMAN Performance Tuning
• First, we need to determine, is there a
bottleneck?
• Oracle provides database views to help us
discover if there are bottlenecks.

• V$RMAN_BACKUP_JOB_DETAILS
• V$SESSION
• V$PROCESS
• V$SESSION_LONGOPS
• V$BACKUP_ASYNC_IO
• V$BACKUP_SYNC_IO

• Where are the possible bottlenecks?
• Read operations
• Control file
• Recovery catalog
• Reading backup sets during recovery
• Reading memory buffers

• Write operations
• Control file
• Recovery catalog
• Writing to tape/disk
• Writing to memory buffers

• First… what is the history of our backups, are they getting slower,
backing up more?

• Sometimes we just think things are getting worse.

• Use V$RMAN_BACKUP_JOB_DETAILS view to find out for sure.

• Querying the view can be slow. (Bug 6868233) – Throw in a RULE
hint to work around this bug.

• View not reset when database is shutdown.

select /*+ RULE */ session_key, session_recid,
input_type, start_time, end_time, output_bytes,
elapsed_seconds
from v$rman_backup_job_details
where start_time >= sysdate-180
and status='COMPLETED'
And input_type='DB FULL';

• This view has one little problem you need to be aware of.
• Running multipule RMAN operations with the same
session can cause the view to accumulate totals.
• Oracle also offers a like Recovery catalog view
•RC_RMAN_BACKUP_JOB_DETAILS

select /*+ RULE */ session_key, session_recid,
start_time, end_time, output_bytes,
elapsed_seconds
from v$rman_backup_job_details
where start_time >= sysdate-180
and status='COMPLETED'
And input_type='DB FULL';

SESSION_KEY SESSION_RECID START_TIM END_TIME OUTPUT_BYTES ELAPSED_SECONDS
----------- ------------- --------- --------- ------------ ---------------
456 456 11-JAN-09 11-JAN-09 228353024 397
461 461 12-JAN-09 12-JAN-09 229755904 422

• In this case the operation on the 12th took 25 seconds
longer than that on the 11th. Probably not a big deal.


• I might be concerned if it looked like this:

----------- ------------- --------- --------- ------------ ---------------
456 456 1-JAN-09 11-JAN-09 228353024 1397
456 456 2-JAN-09 11-JAN-09 328353024 2355
456 456 3-JAN-09 11-JAN-09 428353024 3390
456 456 4-JAN-09 11-JAN-09 458353024 4505
456 456 5-JAN-09 11-JAN-09 498353024 10044
456 456 6-JAN-09 11-JAN-09 528353024 11597
456 456 7-JAN-09 11-JAN-09 628353024 12497
456 456 11-JAN-09 11-JAN-09 728353024 14497
461 461 12-JAN-09 12-JAN-09 829755904 19422


• Let’s look at our most current RMAN backup so we can find
session details.

select sid, serial#, program
from v$session
where lower(program) like '%rman%';

SID SERIAL# PROGRAM
------ ---------- ----------
125 149 rman.exe
128 130 rman.exe
134 164 rman.exe

• A summary of the active session history might help (make sure
you are licensed!)
Set lines 132
Column session_id format 999 heading ”SESS|ID”
Column session_serial# format 999 heading ”SESS|SER|#”
Column event format a40
Column total_waits format 9,999,999,999 heading ”TOTAL|TIME|WAITED|MICRO”
Select session_id, session_serial#, Event, sum(time_waited) total_waits
From v$active_session_history
Where session_id||session_serial# in (125149, 128130, 134164)
And sample_time > sysdate – 1 And program like ’%rman%’
And session_state=’WAITING’ And time_waited > 0
Group by session_id, session_serial#, Event
Order by session_id, session_serial#, total_waits desc;

• The resulting output
TOTAL
SESS TIME
SESS SER WAITED
ID # EVENT MICRO
---- ---- ---------------------------------------- --------------
125 149 control file single write 1,388,961
125 149 control file sequential read 45,964
125 149 control file parallel write 3,789
128 130 RMAN backup & recovery I/O 192,263,005
128 130 control file parallel write 529,012

• Total waits of about 195 seconds. Runtime job details. It took
635 seconds to backup 1.3GB of data. Can we make this faster?
----------- ------------- --------- --------- ------------ ---------------
468 468 11-JAN-09 11-JAN-09 1350017536 635


• Let’s see what happens if we use compression.
RMAN> CONFIGURE DEVICE TYPE DISK BACKUP TYPE TO COMPRESSED BACKUPSET;

• What do you think will be the result?


• Let’s look at the RMAN backup using compression. Here are
the session details.

select sid, serial#, program
from v$session
where lower(program) like '%rman%';

SID SERIAL# PROGRAM
---------- ---------- ----------
128 102 rman.exe
130 116 rman.exe
134 129 rman.exe

• The resulting output from v$active_session_history:
TOTAL
SESS TIME
SESS SER WAITED
ID # EVENT MICRO
---- ---- ---------------------------------------- --------------
128 102 control file sequential read 17,916
128 102 recovery area: computing backed up files 14,277
134 129 enq: TC - contention 9,653,878
134 129 RMAN backup & recovery I/O 5,414,782
134 129 control file parallel write 1,014,960

• Total waits of ~20 seconds for a backup that took a total of 397
seconds. Pretty good improvement there.
----------- ------------- --------- --------- ------------ ---------------
456 456 11-JAN-09 11-JAN-09 228353024 397

• How long will it take?
• Sometimes the question is not how long, but how much longer.
• You can use v$session_longops to get answers to such questions as
seen here:
• First, we have an rman backup running. Let’s see which sessions it is
connected to:

SQL> select sid, serial#, program
2 from v$session
3 where lower(program) like '%rman%';

SID SERIAL# PROGRAM
---------- ---------- ----------
128 138 rman.exe
134 178 rman.exe
138 267 rman.exe

• How long will it take?
• Now we query the v$session_longops view to see how long that
backup is expected to take:
Select sid, serial#, opname, time_remaining
From v$session_longops
Where sid||serial# in (128138,134178,138267)
And time_remaining > 0;

SID SERIAL# OPNAME TIME_REMAINING
---------- ---------- ------------------------------ --------------
128 138 RMAN: aggregate input 118
138 267 RMAN: full datafile backup 105

• The aggregate input line is the one we are interested in. It seems we
have 118 seconds remaining for the running operation.

• Ideas to improve performance
• Channels = available devices
• Parallelize as best you can.
• Don’t over do it.
• Datafile multiplexing
• Allows parallelized
datafile backups within a
backup set/piece.
• Impacts the size of the
datafile read buffers.

•The default settings for Multiplexing
are typically ok.
• Settings that impact Multiplexing:

Setting Default Meaning
filesperset 64 How many datafiles can be put in each backup set.
maxopenfiles 8 Defines how many datafiles RMAN can read at once.

•Multiplexing is dependent on filesperset
and maxopenfiles settings.

• Filesperset controls how many datafile
can be in each backup set.

• RMAN does not multiplex a datafile
when it is backed up using multi-section
backups in 11g.

• The Level of Multiplexing is dependent
on the setting of maxopenfiles.

• The level of multiplexing is the number
of files that will be put in one backup set
piece.

• As the level of multiplexing increases
the buffer size for input files will decrease.

•Memory allocated to channels based on
level of multiplexing:

• 4 LOM – 16 buffers@1mb per buffer

•>4 but < 8 LOM – Variable # of
channels of 512KB.
Total buffer size < 16MB.

• >8 LOB – 4 disk buffers of 128kb for each file.
Total buffer size per input file 512kb.

• Changing filesperset
• Backup command

Backup database filesperset=4;

• Changing maxopenfiles
• Allocate channel
• Configure channel

configure channel 1 device type disk maxopenfiles 4;

•Here are the results of one test where we adjusted maxopenfiles to settings
of 4, 8 and 16.

• Note in this test, the larger the setting, the slower the backup.
• Oracle suggests that higher settings may improve performance. Kind of
the opposite of this test.
• Note that your results may vary for many reasons.
Maxopenfiles setting Time to backup
4 6:35
8 6:46
16 7:45

• Your mileage may vary. Test before you invest.

• The Duration parameter (11g) can be used to minimize the
load of the database backup over a specific time.

• Use the duration command to indicate the maximum length
of the backup (can also use the option partial).
• Use the minimize load parameter to distribute the IO of the
backup over the duration of the backup.
• Careful with tape backups. This can impact the
performance of streaming.
• The minimize time parameter will cause the backup to
utilize as much IO bandwidth as possible.

• Are you incorrectly using RATE?
• Rate is designed to reduce performance, not
improve it!
• You can sometimes alter the size of the tape
buffers by using the PARMS and BLKSIZE options
of the configure or allocate commands..

• Large pool does not help unless you are facing
synchronous IO issues.
• Pretty rare these days not to have asynch IO.

• Incremental backups
• Improve backup performance by instituting an
incremental backup strategy.
• Level 1 backups only write changed blocks.
• Use a block change tracking file to further
speed up the backups.
• Higher levels of multiplexing with incremental
backups might be helpful.
•Smaller memory buffers get filled quicker
•Written out to tape quicker.


• Is it a disk bottleneck or a tape bottleneck?
• Backup validate may provide some insight.
• No tape IO
• Same disk IO.
• If the backup validate is faster, then your
problem is tape.
• If the backup validate is the same, then the
problem is disk IO.


•The V$BACKUP_ASYNC_IO view can provide
some insight into specific bottlenecks.
• Provides throughput information during backup.
• Total information maintained after backup.
• Cleared after instance restart.
• Maintains some historical record, but don’t
depend on it being there very long.

•The V$BACKUP_ASYNC_IO view can provide some insight
into specific bottlenecks.
• Provides throughput information during backup.
•Example:
select sid, serial, effective_bytes_per_second, filename
from V$BACKUP_ASYNC_IO;

EFF
BYTES
PER
SID SERIAL SEC FILENAME
---- ---------- ----------- ----------------------------------------------
134 16 86,878,436
134 16 C:ORACLEORADATAORCLSYSAUX01.DBF
134 16 C:ORACLEORADATAORCLSYSTEM01.DBF
134 16 1,697,564 C:ORACLEORADATAORCLUNDOTBS01.DBF
134 16 1,485,938 C:ORACLEORADATAORCLORCL_RETENTION_01.DBF
134 16 73,728 C:ORACLEORADATAORCLMY_DATA_01.DBF
134 16 73,728 C:ORACLEORADATAORCLTESTTBS.DBF
134 16 1,581,056 C:ORACLEORADATAORCLNEW_TBS_01.DBF
134 16 1,056,768 C:ORACLEORADATAORCLRMAN_DEMO_01.DBF
134 16 1,587,610 C:ORACLEORADATAORCLUSERS01.DBF

•The V$BACKUP_ASYNC_IO view can provide
some insight into specific bottlenecks.
• We can find the datafile that is causing us the
biggest problem:
• In this example, 2% of the IO’s were involved in a
long wait for this ASM tablespace :
SELECT LONG_WAITS/IO_COUNT, FILENAME
FROM V$BACKUP_ASYNC_IO
WHERE LONG_WAITS/IO_COUNT > 0
ORDER BY LONG_WAITS/IO_COUNT DESC;

LONG_WAITS/IO_COUNT FILENAME
------------------- ------------------------------------------------------------
.2 +DGROUP1/orcl/datafile/asm_tbs.257.674829161

• Large file backups…
• You have a 20GB file.
• What’s the fastest way to back that file up?
• Multiplexing
• Multi-threaded/process.
• Many files written in parallel to a single
backup set piece.
• One file limited to one channel/one backup set
(but many potential backup set pieces)

• Another option in 11g
• Multi-section backups (also called multisection
backups and interfile backup parallelism in the Oracle
docs).
• New in Oracle Database 11g.
• Spreads IO of datafiles across more than one
channel in parallel.
• Higher I/O
• Higher CPU
• Can reduce the time it takes to backup large
datafiles.

• If all else fails

• Create an RMAN trace file

• Use the trace parameter at the RMAN
command prompt.

• Example
rman target=/ trace=c:rmantrace.log debug

• Enable RMAN Debugging
• Enabled at the command line.
• Shows you in detail what calls RMAN is making.
• Enabling debug at command line

rman target=/ debug

•Enabling debug from RMAN prompt

Rman>debug on
Rman>debug on



•Debug options
•Io, sql, plsql, rcvman, rpc

•Enable RMAN Debugging

• Enable SQL tracing for RMAN

•At the end of the day, everything RMAN does happens
in the database.

• Wait events…etc…

• Use the SQL command to enable event 10046 tracing
for the RMAN session

• Review the trace files for wait events.

Example
sql "alter session set events ''10046 trace name
context forever, level 12''";

RMAN> backup as compressed backupset database plus
archivelog delete input;

• Carefully note the double single quotes.

• Unfortunately this does not trace everything…


Other solutions

• Trace recovery catalog database

• Logon trigger on backup database that checks
for RMAN logon and enables tracing.

• Example logon trigger to trace all RMAN sessions.
• Remember to grant alter session to the schema.
Create or replace trigger tr_rman_logon
After logon on database
Declare
v_program varchar2(1000);
Begin
SELECT distinct program into v_program
FROM v$session
WHERE audsid = userenv('sessionid')
and program like '%rman.exe%‘ and rownum < 2;
if v_program='rman.exe'
then
execute immediate 'alter session set events ''10046 trace
name context forever, level 12''';
end if;
Exception
when NO_DATA_FOUND then
NULL;
End;
/


• When we use a separate logon trigger we find that
there are actually many RMAN related processes
that we need to trace.

• Identify all the RMAN trace files being generated:
select a.sid, a.serial#, a.program, b.tracefile
from v$session a, v$process b
where a.paddr=b.addr
and a.program like '%rman%’;

SID SERIAL# PROGRAM
---------- ---------- --------------------
TRACEFILE
-------------------------------------------------------
136 453 rman.exe
c:oraclediagrdbmsorclorcltraceorcl_ora_12724.trc

137 500 rman.exe

170 304 rman.exe

• Use tkprof to process the trace files…
tkprof orcl_ora_12724.trc orcl_ora_12724.trc.tkprof
sort=prsela,exeela,fchela print=3 aggregate=no
explain=rman/rman

explain=rman/rman

explain=rman/rman

• Review the TKPROF output to determine the main
performance culprits.

• Sort the output by elapsed time.
• Some waits do not surface easily (RMAN backup
& recovery I/O for example).
• Typical problems
• Control file access
• Recovery catalog access
• Backup device I/O waits

• Problems we have seen in the past…
• Check Metalink for bugs
• Bugs in the recovery catalog.
• Indexes missing.
• May need to split into two catalogs.
• Control file issues
• Lots of RMAN records can cause slowing
response.
• Conflicts with different RMAN operations on
the same database running at the same time.
(Snapshot control file – Enqueues)

• Problems we have seen in the past…
• MML layer settings need tweaked
• Block size
• Network frame size
• Network throughput
• Do you have enough devices that you are backing
up too?
• Scheduling? Are to many backups happening at
the same time?
• Backup during the day?
• Are you validating to death?

•Summary of performance tuning tips
• Check your network settings.
• Check your MML vendor settings.
• How many devices do you have.
• Device contention – Scheduling.
• RMAN bugs
• Upgrade to 11g
• Duration
• Multi-section backups
• If backing up to tape, is the problem with your disks
or your tape backup?
• Use backup validate to determine.

• Review Metalink documents

• 247611.1 – Known RMAN Performance
Problems

• 463227.1 – Oracle 10g RMAN Recovery
catalog known performance issues.

• Commonly run into bugs..
• Bug 6476935 - RMAN Resyncs may take a
long time to complete
• Version affected : 10.2.0.3
• Version fixed : 11.2
• Affects : Catalog Schema
• Workaround : create index rsr_i_stamp on
rsr(rsr_sstamp, rsr_srecid);

• Bug 5219484 - CATALOG RESYNCS ARE
VERY SLOW - ROUT TABLE HAS 6 MILLION
ROWS +
• Version affected : 10.1
• Version fixed : 10.2.0.4
• Affects : Catalog Schema
• Diagnosis : Many rows in ROUT-table.
• Workaround : See Note 378234.1 Rman
Catalog Resync Operation is Very slow at
10G

RMAN Trouble Shooting
• So what happens if you get RMAN errors and you
can’t figure out what’s up?

•Metalink (bug lookup)

• Use slaxdebug

• Useful for syntax errors

• Useful for strings that are too long for RMAN

• Slaxdebug Example
Rman target=/ slaxdebug
RMAN> backup database;
debug: LOOKAHEAD token is "BACKUP_".
debug: SHIFT token "BACKUP_", enter state "5"
debug: (ACT STK) [5, SHFT]
debug: (PAR STK) [0] [5]
debug: LOOKAHEAD token is "DATABASE_".
debug: SHIFT token "DATABASE_", enter state
"154"
debug: (ACT STK) [154, SHFT]
debug: (PAR STK) [0] [5] [154]

• Commonly used views

• V$RMAN_BACKUP_JOB_DETAILS
• Information on RMAN backup jobs.

• V$BACKUP_ASYNC_IO
• Information on ASYNC performance of
RMAN operations

• V$BACKUP_SYNC_IO
• Information on SYNC performance of RMAN
operations


• V$PROCESS
• Information on processes in the database

•V$SESSION
• Information on sessions in the database.

• V$SESSION_LONGOPS
• Information on long running operations in the
database.


• V$RECOVERY_PROGRESS
• Show progress of RMAN user managed
recovery operations.

• V$SESSION_WAIT
• Displays wait events for sessions, including
RMAN sessions.


• V$RMAN_OUTPUT
• Provides the output from RMAN in the form of
a table.

• V$RMAN_STATUS
• Displays the status of finished and on-going
jobs.

RMAN Database Duplication
• RMAN provides an easy (sorta) way of duplicating
your database to another database.

• Backup based duplication
•Requires that a backup of the database being
duplication be available.

• Active database duplication (11g only)
• Backup the target database online duplicating it
to the auxiliary instance over the network. No
backup required.


•Duplicate from
•Source host
•Target database

•Duplicate to
• Destination host
• Auxiliary Instance

•Basic steps to prepare to duplicate a database

•Back up the target database (Backup based
database duplication only)

• Make backup images available to the destination
host. (Backup based database duplication only)


•Deciding where to put the duplicate database
related files.
• The destination host may have different file
systems.
• The duplicate command can make many of the
needed changes for you on the fly.
• May be duplicating to the same host too.

• Preparing the auxiliary instance for the
duplication.
• Configure OS Specific settings
• Configure the auxiliary instance parameters.
• Typically it’s a basic parameter file.
• Use the target database SPFILE as the final
copy.
• Create the password file.
• Configure networking
• Startup nomount the auxiliary instance


• Connecting to RMAN for duplication

Set oracle_sid=orcl

Rman target=/ auxiliary=sys/password@mydb

• The duplicate database command

•Used to execute the duplication

•Many different options

• RMAN (10g and later) will create the needed
tempfiles for you.


• Duplicate to a specific point in time.

• Duplicate to the point of the duplicate command.

• Duplicate to a restore point.

• Active or backup based database duplication.

• Examples – Backup based database duplication

Duplicate target database to neworcl
nofilenamecheck spfile;

nofilenamecheck spfile skip tablespace users;

nofilenamecheck spfile to restore point ’Test’;

nofilenamecheck spfile open restricted;


• Active database duplication

• Does not support until clause

• Does not support to restore point clause.


•Example - Active database duplication

duplicate target database to neworcl from active
database nofilenamecheck
spfile set control_files
'c:oracleoradataneworclcontrol01.ctl',
'c:oracleoradataneworclcontrol02.ctl'
set db_file_name_convert
’c:oracleoradataorcl’,
’c:oracleoradataneworcl’
set log_file_name_convert
’c:oracleoradataorcl’,
’c:oracleoradataneworcl’;

RMAN Tablespace Point Time
Recovery (TSPITR)
• TSPITR provides the ability to restore a specific
tablespace to a specific point in time in the database.

• RMAN provides this functionality.

• Minimal configuration
• Figure out where you want the temporary instance
files to reside.

•Figure out which tablespaces you want to recover.

Recovery (TSPITR)
• Use the RMAN recover command to start the restore:
recover tablespace users
until time '10/06/2008:22:42:00'
auxiliary destination 'c:oracleauxiliary';

• Tablespaces recovered are left offline

• You should back them up after the TSPITR.

• Bring them online with the alter tablespace
command using the online keyword.

Recovery (TSPITR)
• You should make sure the tablespace is fully self-contained.
•Tablespaces in the transport set can not have external object
references.

• Tables in Tablespace A

• Indexes in Tablespace B

• Must move both tablespaces A and B.

• Use the TS_PITR_CHECK view to determine if you are
moving self-contained tablespaces.

Recovery (TSPITR)
• Example of checking the transport set
SELECT ts1_name, ts2_name, reason
FROM SYS.TS_PITR_CHECK
WHERE ( TS1_NAME IN ('USERS') AND TS2_NAME NOT IN ('USERS') )
OR ( TS1_NAME NOT IN ('USERS') AND TS2_NAME IN ('USERS') );

TS1_NAME TS2_NAME
---------- ----------
REASON
----------------------------------------------------------------
USERS INDEX_TBS
Tables and associated indexes not fully contained in the
recovery set

Recovery (TSPITR)
• It is possible that you will lose objects in the
tablespace that were created after the point in time to
which you restore the tablespace.

• Determine which objects will be lost by querying the
view TS_PITR_OBJECTS_TO_BE_DROPPED

• Export the objects before the TSPITR and then re-
import them afterwards.

Recovery (TSPITR)
• Example of looking for objects that will be dropped.
SQL> SELECT OWNER, NAME, TABLESPACE_NAME,
2 TO_CHAR(CREATION_TIME, 'YYYY-MM-DD:HH24:MI:SS')
3 FROM TS_PITR_OBJECTS_TO_BE_DROPPED
4 WHERE TABLESPACE_NAME IN ('USERS')
5 AND CREATION_TIME >
6 TO_DATE('07-OCT-08:22:35:30','YY-MON-DD:HH24:MI:SS')
7 ORDER BY TABLESPACE_NAME, CREATION_TIME;
OWNER NAME
------------------------------ ----------------------------
TABLESPACE_NAME TO_CHAR(CREATION_TI
------------------------------ -------------------
SCOTT TESTTABLE
USERS 2008-10-07:19:34:39

Recovery (TSPITR)
• There are a number of rules related to TSPITR. Some
highlights:

• The database must be in ARCHIVELOG mode.

• You must have at least one backup take before the
point in time to which you wish to restore the
tablespace. All archived redo logs required after that
backup must also be available.

• If you rename a tablespace, you cannot perform
TSPITR to any point in time before the rename
operation.

Recovery (TSPITR)
• The normal no no’s continued…
• No replicated master tables.
• Complete tables must be transported (e.g. all
partitions).
• Tablespaces with tables that contain VARRAY or
nested tables are not allowed.
• Tablespaces with external tables are not allowed.
• UNDO tablespaces are not allowed.
• Any tablespace with SYS owned objects are not
allowed.

Recovery (TSPITR)
• After the recovery

•Datafiles added to the tablespace on the target
database after the point in time for the recovery will
be restored but empty.

• All backups associated with tablespaces in the
transport set taken before the point in time that you
restored the tablespaces to are no longer valid. You
should run a backup after the TSPITR.

Recovery (TSPITR)
• After the recovery

•You will not be able to run another TSPITR on that
tablespace to any time before the point to which
you restored the tablespace.

•Once a TSPITR is complete, you will not be able
to use the control file to restore any part of the
database to any point in time before the time that
you restored the tablespaces to during the TSPITR.

Recovery Catalog Shenanigans

• About the recovery catalog views

•Reporting against the recovery catalog

• Recovery catalog performance issues

Recovery Catalog Views
• You may already be familiar with the RMAN related
V$ views.

• Exposes RMAN related information contained in
the control file.

•Helpful in determining certain RMAN
information.

• You may already be familiar with the RMAN related
V$ views.

•The list and report command often are sufficient,
so the use of these views it probably rare.

• Examples:
•V$backupset
•V$backup_piece

• The recovery catalog has a number of views that are
roughly equivalent to the V$ views.

• All recovery catalog views start with RC_.

• Contain records for all databases registered with the
catalog.


• You can correlate most RMAN control file views
with recovery catalog views. The Oracle Database
Backup and Recovery Reference provides a table that
associates the V$ views with the RC_ views.
• Example
• V$BACKUPSET – RC_BACKUPSET
• V$BACKUP_PIECE – RC_BACKUP_PIECE

• Hmmm…. This seems like this could be useful….

• Enterprise reporting on backup status

• Tracking length and size of backups

• Looking for possible contention issues

• Databases not being backed up


• Hmmm…. This seems like this could be useful….

• Combined with a configuration management tool,
the recovery catalog can be even more powerful!

• Here is a script that displays databases that are
registered in the catalog and the last date they were
backed up.

Select a.db_key, a.dbid, a.name db_name,
b.backup_type, b.incremental_level,
b.completion_time, max(b.completion_time)
over (partition by a.name, a.dbid)
max_completion_time
from rman.rc_database a, rman.rc_backup_set b
where b.status = 'A'
and b.incremental_level = 0
and b.backup_type = 'D'
and b.db_key = a.db_key;

RMAN Encryption (11g)
• RMAN Supports 3 modes of encrypting backups.
• Transparent mode
• Requires configuration of the Oracle
encryption wallet.
• Password mode
• Requires that you define the password before
you start the backup.
• Dual mode
• Uses both transparent mode and password
mode encryption.

RMAN Encryption (11g)

• To use encryption, you must first enable it.

• This will use transparent encryption by default.

Configure encryption for database on;

• Example of enabling password encryption

Set encryption on identified by backup only;

RMAN has GUTS!
The RMAN Internal Packages

• At the end of the day, pretty much everything RMAN
does boils down to the use of packages built into the
database kernel.
• These packages are largely undocumented.
• The most common packages used are:

•DBMS_BACKUP_RESTORE

•DBMS_RCVMAN

•DBMS_RCVCAT

RMAN has GUTS!

• Another example.
• <10g, how to unregister a database?
• Use dbms_rcvcat.unregisterdatabase

• Need the DBID, the DBKEY from the recovery catalog:

SELECT db_key, db_id
FROM db WHERE db_id = 1231209694;

DB_KEY DB_ID
---------- ----------
1 1231209694

RMAN has GUTS!

• Then simply unregister the database:

begin
dbms_rcvcat.unregisterdatabase(1 , 1237603294);
End;
/

RMAN has GUTS!

• We can see an example of the use of these packages in
rmanRestoreDatafiles.sql.
• This script is created when you use DBCA to create
a database using a template with datafiles.
• Typically located in
• ORACLE_BASEcfgtoolsdbca<dbname> (11g)
• ORACLE_HOMEadmin<dbname> (10g)
• Also found in the scripts directory when DBCA
creates a script to create the database.
• Let’s look at an example of this script for a
minute…

RMAN has GUTS!

• Recover.bsq is the heart of RMAN.
• All RMAN operations are scripted in recover.bsq in
some way.
• Located in $ORACLE_HOME/rdbms/admin
• Let’s look in recover.bsq real quick….

RMAN has GUTS!

• If you are having frustrating problems with RMAN
the debug option may be of help.

•You can enable debug from
• When starting RMAN from the OS prompt
• RMAN Command line
• When allocating a channel
• When configuring a channel

RMAN has GUTS!

• You can enable debug in different modes
• IO
• SQL
• PL/SQL
• RCVMAN
• RPC
• ALL

• Example
Rman target=/ debug=io
Rman target=/ debug=all

Migration using RMAN
How to migrate a 32 bit DB to 64 bit
This procedure can be used for single instance or RAC (9i to 11g):

In the 32 bit source server :

Start making a full backup of the database including all archivelogs:

RMAN> run {
allocate channel c1 type disk;
allocate channel c2 type disk;
backup database plus archivelog;
backup current controlfile; }

In the target 64 bit server:

1) Copy backup files from source server
2) Create the pfile or copy the pfile fromsourceserver
3) Startup nomount the new instance.
4) Connect to rman and set the dbid:
$ rman target / nocatalog RMAN> set dbid=<dbid from the 32 server>;


5) Restore the controlfile:

RMAN> restore controlfile from ‘/disk1/backup_32/backup_piece_name’;
RMAN> alter database mount;
RMAN> restore database;

6) Identify the archivelog till you want to do the recovery and check if you have
all the backups you need to do it and run:

RMAN> run {
set until sequence xxxx;
recover database;
}
RMAN> exit


7) Now Migrating to 64 bit:

SQL> recover database until cancel using backup controlfile;
cancel
SQL> alter database open resetlogs migrate;
SQL> @$ORACLE_HOME/rdbms/admin/utlirp.sql
SQL> @$ORACLE_HOME/rdbms/admin/utlrp.sql
SQL> shutdown immediate;
SQL> startup

Note: Also please take a look in other methods in the following Metalink Note:
Migration of Oracle Instances Across OS Platforms - Doc ID: Note:733205.1

What else is new on 11G
RMAN Enhancements

The RMAN in 11g includes several interesting new features, including the following:

• Active (network-based) database duplication
• Parallel backup and restore of large files (multisection backups)
• Archival backups (long term)
• Improved lock media recovery performance
• New persistent configuration Parameters
• Backup failover to non-flash recovery areas

RMAN and RAC
Normal Architecture

Shared Disk Array:
– Datafiles
– Control File
– Snapshot Control File
Each node have their own:
– Redo logs
– Archive Logs

Since 10gR2 RMAN can utilize information gleaned from the CRS to provide better
integration with RAC, you not longer have to configure a channel to specifically connect
to a particular node. If you have 2 nodes and you set the parallelism to 2, RMAN will
query the CRS for the nodes information and it will automatically spread the two
channels across the nodes. CRS keeps track of node utilization and will spread the
backup jobs out to those nodes that currently being least utilized to avoid I/O traffic
jams.

RMAN and RAC
Tip:

Setup each node with a LOG_ARCHIVE_DEST_2 parameter that writes to the
another node, mount a shared directory in node 2 that was included in the
destination2 in the node 1 and vice versa.

When setup like this, Oracle writes archive logs from each node to the archive
destination of the other node. This gives us an elegant solution for backing up the
archive logs from a single node and provide us with fault tolerance in case a node is
lost.

RMAN and RAC
Better Tip yet:

You can think in deploy ASM as your volume manager for all your RAC files, then
you can create a second disk group and deploy your flash recovery area (FRA) to the
disk group. With FRA on an ASM disk group, managing archive logs in RAC will
become the same as managing them in a single-thread environment. If you
configure the database to archive to the FRA, both nodes will deposit the logs in a
single location where they are visible to both nodes at all times.

More things we can do:

• Duplicating a RAC Database to a Single-Node Database
• Create a Single-Node Standby Database from a RAC Database

Reading recommendations:

Oracle RMAN 11g – Backup and Recovery
By Robert G. Freeman and Matthew Hart

OCP Oracle Database 11g – New Features for Administrators
By Sam Alapati

Soon:

Pro Backup & Recovery – Apress
By Francisco Munoz Alvarez

B35 all you wanna know about rman by francisco alvarez

More Related Content

What's hot (20)

Similar to B35 all you wanna know about rman by francisco alvarez (20)

More from Insight Technology, Inc. (20)

Recently uploaded (20)

B35 all you wanna know about rman by francisco alvarez