miércoles, 10 de junio de 2015

Creating a Standby Database using RMAN and Tivoli Storage Manager



Creating a Standby Database using RMAN and Tivoli Storage Manager


 

Confirm that TSM is installed on the host/hosts and tdpo.opt file has been configured.

 

Contact the TSM administrator and confirm that all pre-requisite tasks related to the TSM set-up have been completed on BOTH primary as well as standby machines.

 

For example on Linux machines the tdpo.opt file should exist in the following location :

 

/opt/tivoli/tsm/client/oracle/bin/

 

or

 

/opt/tivoli/tsm/client/oracle/bin64/

 

If the ORACLE_SID is dgtest9i, then the tdpo.opt file is set up with the name tdpo.dgtest9id.opt

 

We can test the TDP set up using the sbttest command

 

export TDPO_OPTFILE=/opt/tivoli/tsm/client/oracle/bin64/tdpo.dgtest9id.opt

 

[oracle@itlinuxdevblade08 bin64]$ sbttest test

 

The sbt function pointers are loaded from libobk.so library.

-- sbtinit succeeded

Return code -1 from sbtinit, bsercoer = 0, bsercerrno = 0

Message 0 not found;  product=RDBMS; facility=SBT

 

 

Pre-requisite tasks to be completed on the Primary Database

 

Ensure that the database has been registered with the RMAN catalog and that the database is also running in ARCHIVELOG mode.

 

If already not created, create a password file. Please make a note of the password used as it should be the same one used when creating the password file on the standby site.

 

Also ensure that the parameter remote_login_password_file is set to EXCLUSIVE

 

Ensure that the Primary database is functioning in force logging mode.

 

It is a best practice to place the primary database in force logging mode so that all operation are captured in the redo stream. To place the primary database in force logging mode issue the following SQL

SQL > alter database force logging;

 

 

Take a backup of the Primary Database

 

Once we have tested the TSM setup on the machine, we can take an online backup of the Primary database using RMAN which will be restored on the standby site to create the standby database.

 

Ensure that the database has been registered with the RMAN catalog and that the database is also running in ARCHIVELOG mode.

 

Make a note of the current log sequence number. This is will be used as a starting point when we specify the range of archive log files that need to be backed up to tape.

 

$ rman target / catalog rman/rman@rmanp cmdfile=backup_primarydb.rcv

 

backup_primarydb.rcv

 

run {

allocate channel ch1 type 'sbt_tape' parms 'ENV=(TDPO_OPTFILE=/opt/tivoli/tsm/client/oracle/bin64/tdpo.mrhbld.opt)';

backup database  ;

release channel ch1 ;

}

 

Create the Standby Control File

 

Using RMAN we can take a backup of the current control file which will be restored on the standby machine as a standby control file.

 

$ rman target / catalog rman/rman@rmanp cmdfile=backup_primarycf.rcv

 

backup_primarycf.rcv

 

run {

allocate channel ch1 type 'sbt_tape' parms 'ENV=(TDPO_OPTFILE=/opt/tivoli/tsm/client/oracle/bin64/tdpo.mrhbld.opt)';

set limit channel ch1 kbytes=3145728 readrate 200;

backup current controlfile for standby;

release channel ch1 ;

}

 

 

READ RATE specifies the number of buffers to be read per second by RMAN backup or copy operations

 

 

 

 

Backup the Archive Log Files

 

While performing the recovery on the standby database after restoring the primary database backup, we need to go to a point in time after which we created the standby control file. We then need to record the last sequence number from V$ARCHIVED_LOG and ensure that the archivelogs until that particular sequence are backed up via RMAN.

 

SQL> alter system switch logfile;

 

SQL > alter system switch logfile;

 

SQL > select max(sequence#) from v$archived_log;

 

Make a note of this log sequence number. This will be used as the end point when we specify the range of archive log files that need to be backed up via RMAN to tape.

 

This can be verified by issuing the ‘archive log list’ command as well as physically checking the log_archive_dest location to crosscheck the last online redo log that has been archived.

 

We will then backup the range of archive log files that have been generated since the time the online backup of database was performed. These will be applied to the standby database as part of the recovery process.

 

$ rman target / catalog rman/rman@rmanp cmdfile=backup_primarydb_arch.rcv

 

backup_primarydb_arch.rcv

 

run {

allocate channel ch1 type 'SBT_TAPE'  parms 'ENV=(TDPO_OPTFILE=/opt/tivoli/tsm/client/oracle/bin64/tdpo.mrhbld.opt)';

backup   archivelog from logseq 16 until logseq  35 thread 1 ;

release channel ch1;

}

 

 

Preparing the Standby site for Data Guard

 

Create the directory structure on the standby site keeping in mind the mount points and directory structure on the primary node.

 

Remember while creating the $BDUMP, $UDUMP, $CDUMP and $PFILE directories that the ORACLE_SID will be the same as the primary database.

 

Also try and ensure that the size of the mount points exactly mirror the primary node otherwise we will face issues both while creating the standby database as well as when we perform housekeeping tasks on the primary like adding or resizing a datafile.

 

We then need to create the password file using the same password for the user SYS as what we have used while creating the password file on the primary node.

 

Copy the init.ora file from the primary node to $ORACLE_HOME/dbs on the standby node.

 

Create a pfile from this spfile

 

$ export ORACLE_SID=

 

SQL> create pfile from spfile;

 

 

In this example, we are making the following assumptions:

 

  • ORACLE_SID=dba01

 

  • Primary Node = itlinux01

 

  • Standby Node = itlinux02

 

  • Mount Points on the primary node are /itlinux01db01 and /itlinux01db02

 

  • Mount points on the standby node are /itlinux02db01 and /itlinux02db02

 

Important:

 

Please ensure that the DB_FILE_NAME_CONVERT and LOG_FILE_NAME_CONVERT  parameters are carefully entered and every mount point which has a datafile or logfile of the primary database is mentioned in the string list.

 

We will now edit the pfile to include parameters specific to the standby site ….

 

FAL_SERVER=dba01_itlinux01

FAL_CLIENT=dba01_itlinux02

DB_FILE_NAME_CONVERT=(‘/itlinux01db01’,’/itlinux02db01’,’/itlinux01db02’,’/itlinux02db02’) 

LOG_FILE_NAME_CONVERT=(‘/itlinux01db01’,’/itlinux02db01’,’/itlinux01db02’,’/itlinux02db02’) 

STANDBY_ARCHIVE_DEST=/itlinux02db01/ORACLE/dba01/arch

STANDBY_FILE_MANAGEMENT=AUTO

 

The following parameters will only apply when the standby database assumes the role of a primary database after a switchover is performed. They can be excluded at this point from the init.ora file, but a best practice is to include it now itself so that when a switchover is made in the future, no further changes to the init.ora need to be made.

 

LOG_ARCHIVE_DEST_2='SERVICE=dba01_itlinux01 ARCH’

LOG_ARCHIVE_DEST_STATE_2=DEFER

LOG_ARCHIVE_DEST_1='LOCATION=/itlinux02db01/ORACLE/dba01/arch’

LOG_ARCHIVE_DEST_STATE_1=ENABLE

LOG_ARCHIVE_FORMAT=%d_%t_%s.arc

 

 

Note: if we are using Oracle 10g then we can use the following parameters:

 

*.log_archive_dest_1='location=/itlinux02db01/ORACLE/dba01/arch','valid_for=(ALL_LOGFILES,ALL_ROLES)'

 

*.log_archive_dest_2='SERVICE=dba01_itlinux01 VALID_for=(ONLINE_LOGFILES, PRIMARY_ROLE) ARCH’

 

 

After making  the required changes to the init.ora create a spfile from this.

 

SQL > create spfile from pfile;

 

 

Network Configuration Files On Standby Machine

 

We will be configuring TNS aliases dba01_itlinux01 and dba01_itlinux02 in the tnsnames.ora file – these are nothing but the FAL_SERVER and FAL_CLIENT parameters that we are defining in the init.ora file.

j

The tnsnames.ora and listener.ora are defined keeping the following in mind ….

 

  • Client connections use the alias ‘dba01’ accessing via port 1526

 

  • Log shipping is using port 1522

 

  • On the primary machine there are two listeners running – one for accepting client connections to the database and the other for the log shipping. ( ports 1526 and 1522 respectively)

 

  • On the standby machine we have only one listener running which is being used for the log shipping between the two sites ( port 1522). The listener listening on port 1526 will ONLY be started when the standby site assumes the role of a primary site after a failover or switchover is performed.

 

This is a sample copy of the tnsnames.ora and listener.ora files – we need to ensure that tnsnames.ora file on BOTH sites have the same entries for FAL_SERVER and FAL_CLIENT as well.

 

listener.ora

 

# LISTENER.ORA Network Configuration File: /opt/oracle/product9206/network/admin/listener.ora

# Generated by Oracle configuration tools.

 

ITLINUX02 =

  (DESCRIPTION_LIST =

    (DESCRIPTION =

      (ADDRESS_LIST =

        (ADDRESS = (PROTOCOL = TCP)(HOST = itlinux02.hq.emirates.com)(PORT = 1522))

      )

      (ADDRESS_LIST =

        (ADDRESS = (PROTOCOL = IPC)(KEY = extproc))

      )

    )

  )

 

#LISTENER.ORA Network Configuration File

#Created by Oracle Enterprise Manager Clone Database tool

 

SID_LIST_ITLINUX02 =

  (SID_LIST =

    (SID_DESC =

      (SID_NAME = PLSExtProc)

      (ORACLE_HOME = /opt/oracle/product9206)

      (PROGRAM = extproc)

      (ENVS = "EXTPROC_DLLS=ANY")

    )

  

    (SID_DESC =

      (GLOBAL_DBNAME = dba01)

      (ORACLE_HOME = /opt/oracle/product9206)

      (SID_NAME = dba01)

    )

  )

 

TRACE_LEVEL_ITLINUX02 = OFF

LOGGING_ITLINUX02 = OFF

TRACE_LEVEL_PRIMARY = OFF

LOGGING_PRIMARY = OFF

 

PRIMARY =

    (DESCRIPTION =

    (ADDRESS = (PROTOCOL = TCP)(HOST = itlinux02.hq.emirates.com)(PORT = 1526))

  )

 

SID_LIST_PRIMARY =

  (SID_LIST =

    (SID_DESC =

      (GLOBAL_DBNAME=dba01)

      (ORACLE_HOME = /opt/oracle/product9206)

      (SID_NAME = dba01)

    )

  )

 

 

tnsnames.ora

 

 

dba01_itlinux01 =

  (DESCRIPTION =

    (ADDRESS_LIST =

        (ADDRESS =

          (COMMUNITY = TCP)

          (PROTOCOL = TCP)

          (HOST = itlinux01.hq.emirates.com)

          (PORT = 1522)

        )

    )

    (CONNECT_DATA =

      (SID = dba01)

     )

  )

 

dba01_itlinux02 =

  (DESCRIPTION =

    (ADDRESS_LIST =

        (ADDRESS =

          (COMMUNITY = TCP)

          (PROTOCOL = TCP)

          (HOST = itlinux02.hq.emirates.com)

          (PORT = 1522)

        )

    )

    (CONNECT_DATA =

      (SID = dba01)

     )

  )

 

 

dba01 =

  (DESCRIPTION =

    (ADDRESS_LIST =

        (ADDRESS =

          (COMMUNITY = TCP)

          (PROTOCOL = TCP)

          (HOST = itlinux01.hq.emirates.com)

          (PORT = 1526)

        )

    )

    (CONNECT_DATA =

      (SID = dba01)

     )

  )

 

 

 

Start the Standby Database in NOMOUNT mode

 

RMAN will restore the controlfile backup that we had taken on the primary site to the locations specified by the CONTROL_FILES parameter in the init.ora file.

 

We need to start the standby database in NOMOUNT mode.

 

SQL> startup nomount;

 

 

Restore and recover the Standby Database

 

 

On the PRIMARY Site we will start the standby database restore and recovery via RMAN using the TSM tape backup that we had earlier taken.

 

Any RMAN ‘duplicate’ command requires us to initiate an auxiliary channel as well.

 

The auxiliary channel will establish a connection to the remote standby database using the TNS alias that we earlier created i.e dba01_itlinux02

 

The ‘set until logseq’ should include the log sequence number which is one higher than the last log sequence pertaining to the archive log files that we had backed up earlier. So if we have backed up archive log files until sequence 1215 then we should include the command ‘set until logseq=1216’.

 

 

$ rman target / catalog rman/rman@rmanp \

auxiliary sys/syspassword @dba01_itlinux02 cmdfile=create_standby.rcv

 

create_standby.rcv

 

run {

allocate channel ch1 type 'sbt_tape' parms 'ENV=(TDPO_OPTFILE=/opt/tivoli/tsm/client/oracle/bin64/tdpo.mrhbld.opt)';

allocate  auxiliary channel aux1 type 'sbt_tape' parms 'ENV=(TDPO_OPTFILE=/opt/tivoli/tsm/client/oracle/bin64/tdpo.mrhbld.opt)';

set until logseq=35 thread 1; 

duplicate  target database for standby

dorecover;

release channel ch1 ;

release channel aux1 ;

}

 

 

 

These are the last few lines from the output of the above command. Note the last archive log sequence number applied was 1215 and NOT log sequence number 1216.

 

 

……………………..

………………………………………..

 

channel clone_default: deleting archive log(s)

archive log filename=/itlinux21ar01/ORACLE/cisl/arch/arch1213.log recid=3 stamp=583459445

archive log filename=/itlinux21ar01/ORACLE/cisl/arch/arch1214.log thread=1 sequence=1214

channel clone_default: deleting archive log(s)

archive log filename=/itlinux21ar01/ORACLE/cisl/arch/arch1214.log recid=2 stamp=583459433

archive log filename=/itlinux21ar01/ORACLE/cisl/arch/arch1215.log thread=1 sequence=1215

channel clone_default: deleting archive log(s)

archive log filename=/itlinux21ar01/ORACLE/cisl/arch/arch1215.log recid=1 stamp=583459425

media recovery complete

Finished recover at 27-FEB-06

Finished Duplicate Db at 27-FEB-06

 

released channel: ch1

 

released channel: aux1

 

Recovery Manager complete.

 

 

Start Managed Recovery on the Standby Database

 

We will then put the standby database in managed recovery mode by issuing the following command :

 

SQL> recover managed standby database disconnect;

 

Confirm that managed recovery is running by checking if the MRP process is running in the background.

 

SQL> !ps -ef |grep mrp

oracle   17740     1  0 07:26 ?        00:00:00 ora_mrp0_dba01

 

 

If for any reason some further recovery is required, we can manually copy all the archive log files from the production machine to the standby machine standby_archive_dest and do the recovery MANUALLY via the following command:

 

SQL> RECOVER STANDBY DATABASE  ( apply the required archive log file when prompted)

 

This example will put the standby database in MAXIMUM PERFORMANCE mode. In case we wish to configure the standby database for MAXIMUM AVAILABILITY or MAXIMUM PROTECTION we need to create additional standby redo log files ( at least one more than the number of the current online redo log files and make sure they are the same size as the online redo log files).

 

We also will need to use the LGWR keyword in the parameter log_archive_dest_2 and issue the command on the primary database while it is mount state.