HP Data Protector User manual

HP Data Protector Software
HP Data Protector and Oracle 10g RAC configuration best practices
Table of contents
Executive summary............................................................................................................................... 3
Overview............................................................................................................................................ 3
Target audience .................................................................................................................................. 3
Review of test configuration .................................................................................................................. 3
10gRAC configurations on hp-ux ....................................................................................................... 6
Oracle Virtual IP Address.................................................................................................................. 7
Oracle Services to manage workloads................................................................................................ 7
Oracle Services definition ................................................................................................................. 7
Oracle Online Backup Configuration ................................................................................................... 10
Configuration Guidelines ................................................................................................................ 10
Use the hostname and the Instance name to configure the backup ....................................................... 11
Use a VIP and the Instance names to configure the backup ................................................................. 12
Use a VIP and a service name to configure the backup....................................................................... 12
Use a VIP and allocate the channels to the instances manually ............................................................ 13
Storing the connection string in the Recovery Catalog......................................................................... 14
Backup Configuration Recommendations .......................................................................................... 16
SAN Backup versus LAN Backup......................................................................................................... 18
Disk Agent and Media Agent are NOT allocated on the same host...................................................... 18
Disk Agent and Media Agent are allocated on the same host.............................................................. 18
File System Backup: Oracle Base and Cluster configuration .................................................................... 19
Oracle Cluster Registry (OCR) ......................................................................................................... 20
Oracle Voting Disk......................................................................................................................... 20
Oracle SPFILE ................................................................................................................................ 21
ORACLE Directories ....................................................................................................................... 21
File system Backup Recommendations............................................................................................... 21
Oracle RMAN Restore and Recovery ................................................................................................... 22
Oracle Restore Enhancements.......................................................................................................... 22
Oracle RMAN Restore script............................................................................................................ 23
Oracle Restore Configuration .......................................................................................................... 24

Restore Recommendations Summary................................................................................................. 25
Recovery Considerations................................................................................................................. 25
Summary .......................................................................................................................................... 27
Appendix A. Reference documentation................................................................................................. 28
HP................................................................................................................................................ 28
Oracle.......................................................................................................................................... 28
Tools ............................................................................................................................................ 28
Appendix B. RMAN helpful Scripts ...................................................................................................... 29
B.1 How to monitor RMAN job progress .......................................................................................... 29
B.2 How to monitor RMAN waits..................................................................................................... 29
B.3 How to monitor RMAN wait states ............................................................................................. 30
B.4 How to monitor RMAN channels current speed............................................................................ 30
Appendix C. Session Reports .............................................................................................................. 31
Session 1: Backup using ita018 and RAC1....................................................................................... 31
Session 2: Backup using ita018-vip and RAC1, RAC2, RAC3, RAC4................................................... 35
Session 3: Backup using ita018-vip and service BACKUP ................................................................... 40
Session 4: Backup using ita018-vip and service MSL5000.................................................................. 45
Session 5: Backup allocating channels manually on RAC1 and RAC2.................................................. 49
Session 6: Restore allocating channels manually on RAC1 and RAC2 .................................................. 53
Session 7: Restore and Recovery using an incremental backup ............................................................ 56
For more information.......................................................................................................................... 60

3
Executive summary
This white paper provides guidance and best practices for performing backup and restore of Oracle
10g Real Application Clusters using HP Data Protector Software.
Since the release of Oracle 9i Real Application Clusters, there has always been uncertainty about
what needs to be backed up and how to backup the RAC database. We will try to address these
questions throughout this paper.
This paper is not:
•Intended to replace the HP Software Data Protector installation and integration guides with Oracle
•A performance whitepaper on backup and recovery of Oracle Real Application Clusters
•A step-by-step installation and configuration guide
Overview
HP Data Protector Software Oracle integration enables customers to perform online and offline
backups of an Oracle database by using the Oracle Recovery Manager and acting as a Media
Management Layer. In the past, the integration with an Oracle Real Application Cluster environment
was performed using just one of the instances of the cluster.
On the other side, Oracle’s Workload Management in a Cluster environment managed the workload
distribution, providing optimal performance. With Oracle 10g, the concept of Services was
introduced. It enables to group database workloads and distribute them over specific instances.
Integrating HP Software Data Protector with Oracle 10g Real Application Clusters, helps you benefit
from using the workload management functionality by distributing the backup and restore load over
more than one instance in the cluster.
Target audience
The target audience for this whitepaper are customers, system integrators and solution architects.
Basic knowledge of Data Protector and Oracle 10g RAC is advisable to understand this white paper
fully.
Review of test configuration
The following hardware configuration was used for the testing:
•5 HP Integrity Server rx2620 running HP-UX 11.23
•1 HP StorageWorks EVA5000
•1 SAN Tape Library MSL5000 with 2 SDLT320 drives
•2x 1000 Base-T per server
The following software configuration was used for the testing:
•HP Data Protector 6.0 Cell Manager running on HP-UX
•Oracle 10g RAC version 10.2.0.3 running on 4 nodes using Oracle clusterware
•Oracle 10g version 10.2.0.3 for the Oracle Recovery Catalog (RCVCAT)
The Oracle10g RAC database (RAC) was configured using 2 ASM disk groups:
•Disk Group DG_DB (database): 4 EVA LUN (Vdisk), each with 100 GB
•Disk Group DG_FRA (Flash Recovery Area): 4 EVA LUN (Vdisk), each with 25 GB

4
The following figure shows the environment described above:
Figure 1. Hardware architecture
The database had a size of 300GB and was created using the AXWORK suite. This suite created a
new tablespace called POOL, with 66 datafiles, each 4095MB in size. The following representation
shows the distribution of the datafiles. Note that they are all of equal size:
FILE_NAME MB
-------------------------------------------------- ----------
+DG_DB/rac/datafile/tbs_pool_001.dbf 4095
+DG_DB/rac/datafile/tbs_pool_002.dbf 4095
+DG_DB/rac/datafile/tbs_pool_003.dbf 4095
+DG_DB/rac/datafile/tbs_pool_004.dbf 4095
+DG_DB/rac/datafile/tbs_pool_005.dbf 4095
+DG_DB/rac/datafile/tbs_pool_006.dbf 4095
+DG_DB/rac/datafile/tbs_pool_007.dbf 4095
+DG_DB/rac/datafile/tbs_pool_008.dbf 4095
+DG_DB/rac/datafile/tbs_pool_064.dbf 4095
+DG_DB/rac/datafile/tbs_pool_065.dbf 4095
+DG_DB/rac/datafile/tbs_pool_066.dbf 4095

5
The following screenshot shows the graphical display shown by the Oracle Enterprise Manager
(Database Console):
Figure 2. Oracle Enterprise Manager: POOL datafiles view
Two extra LUN were allocated on the shared storage for the Oracle Cluster Registry (OCR) and
Oracle voting disks.
Below, a list of CRS resources defined in our cluster and their status:
oracle@ita018[RAC1]:/opt/oracle/product/CRS/bin$ ./crsstat.sh
HA Resource Target State
----------- ------ -----
ora.RAC.BACKUP.RAC1.srv ONLINE ONLINE on ita018
ora.RAC.BACKUP.RAC2.srv ONLINE ONLINE on ita021
ora.RAC.BACKUP.RAC3.srv ONLINE ONLINE on ita022
ora.RAC.BACKUP.RAC4.srv ONLINE ONLINE on ita023
ora.RAC.BACKUP.cs ONLINE ONLINE on ita018
ora.RAC.MSL5000.RAC1.srv ONLINE ONLINE on ita018
ora.RAC.MSL5000.RAC2.srv ONLINE ONLINE on ita021
ora.RAC.MSL5000.cs ONLINE ONLINE on ita018
ora.RAC.RAC1.inst ONLINE ONLINE on ita018
ora.RAC.RAC2.inst ONLINE ONLINE on ita021
ora.RAC.RAC3.inst ONLINE ONLINE on ita022
ora.RAC.RAC4.inst ONLINE ONLINE on ita023
ora.RAC.db ONLINE ONLINE on ita021
ora.ita018.ASM1.asm ONLINE ONLINE on ita018
ora.ita018.LISTENER_ITA018.lsnr ONLINE ONLINE on ita018
ora.ita018.gsd ONLINE ONLINE on ita018
ora.ita018.ons ONLINE ONLINE on ita018
ora.ita018.vip ONLINE ONLINE on ita018
ora.ita021.ASM2.asm ONLINE ONLINE on ita021
ora.ita021.LISTENER_ITA021.lsnr ONLINE ONLINE on ita021
ora.ita021.gsd ONLINE ONLINE on ita021
ora.ita021.ons ONLINE ONLINE on ita021
ora.ita021.vip ONLINE ONLINE on ita021
ora.ita022.ASM3.asm ONLINE ONLINE on ita022
ora.ita022.LISTENER_ITA022.lsnr ONLINE ONLINE on ita022
ora.ita022.gsd ONLINE ONLINE on ita022
ora.ita022.ons ONLINE ONLINE on ita022
ora.ita022.vip ONLINE ONLINE on ita022
ora.ita023.ASM4.asm ONLINE ONLINE on ita023
ora.ita023.LISTENER_ITA023.lsnr ONLINE ONLINE on ita023
ora.ita023.gsd ONLINE ONLINE on ita023
ora.ita023.ons ONLINE ONLINE on ita023
ora.ita023.vip ONLINE ONLINE on ita023

6
10gRAC configurations on hp-ux
Starting with Oracle10g RAC, there are different configurations possible for HP-UX; essentially, the
customer needs to decide about which clustering solution will be deployed:
•Service Guard Extension for RAC + Oracle Clusterware
•Oracle Clusterware only
Based on the storage volume manager choice, there are three main alternatives:
•Oracle Automatic Storage Management (ASM)
•HP Cluster File system (CFS)
•RAW devices
Important Note: RAW devices do NOT provide a shared location for storing archive logs (archived
logs cannot be located on RAW). In this case, an extra approach has to be addressed to have all
archived logs accessible by all RAC instances. This is needed for backup and recovery purposes. Two
possible approaches are to:
•Define an archivelog destination local on each node (i.e. /arch1 on RAC1, /arch2 on RAC2) and
export it and cross mount it on every system. In case one of the nodes is down, the archivelog
directory is also not available.
•Define a shared archivelog destination for all RAC nodes on an NFS share and use a clustering
solution as HP Service Guard to guarantee its availability.
Figure 3. Oracle 10gRAC on hp-ux supported configurations
SGeRAC &
Oracle
Clusterware
10g RAC
HP-UX
Oracle
Clusterware
SLVM CVM ASM
RAW ASM RAW CFS
SGeRAC &
Oracle
Clusterware
10g RAC
HP-UX
Oracle
Clusterware
SLVM CVM ASM
RAW ASM RAW CFS

7
Oracle Virtual IP Address
Since Oracle 10g, clients use a Virtual IP Address (VIP) to connect to the database. The VIP is a static
IP which is defined in the Oracle Clusterware installation.
When a RAC node fails, its associated VIP will failover automatically to one of the remaining nodes.
The new node hosting the failed VIP will associate a new MAC address for this VIP and will make it
known outside.
When clients are not using a VIP for their connections, in case of node failure, they run in a TCP
timeout. For availability and failover purposes, we have defined a VIP on each RAC node (ita018-vip,
ita021-vip, ita022-vip and ita023-vip). We will use the VIP to configure the integration, thus
providing Failover capabilities.
Oracle Services to manage workloads
Applications using an RAC database need to have their workload managed across the entire cluster.
For this purpose, Oracle introduced the Workload Management framework. Based on the use of
Services, it presents the application using the cluster an interface for managing their load in a very
simple way.
Oracle Services allow breaking down workloads in a manageable way, based on Service Level
Agreements, High Availability needs, or business needs and route them to the optimal instances
assigned. It is recommended to group load with the same service level. Load balancing among the
available instances is provided by Oracle Net Services.
When defining a new service, we can specify on which RAC instances the service should run on:
•Preferred: service should run on this instance
•Available: service should run on this instance if a preferred fails
•Not Used: the service should not use this instance
Transparent Application Failover (TAF) enables Oracle Net to failover to a different listener. Based on
the TAF method, the services can be defined as:
•None:
•Basic: connection to failover server will be established at failover time
•Preconnect: pre-established connections are allocated to provide faster failover
Oracle Services definition
To define a new service, you can use the command line interface, srvctl, or the database
configuration assistant, dbca:
$ srvctl add service –d RAC –s BACKUP –r RAC1,RAC2,RAC3,RAC4 –P PRECONNECT
In the above example, we add a new service called BACKUP in the database RAC. With the option –
r we specified the list of preferred instances and the option –P specifies the TAF policy.
The same can be achieved using the Database Configuration Assistant (dbca) GUI:

8
Figure 4. Database Configuration Assistant: Database Service BACKUP
The following figure is a graphical representation of the BACKUP service, which expands over all
available instances:
Figure 5. Oracle Service BACKUP
We define also a service called MSL5000, allowed to run on two of the four instances; we set two
instances to Preferred and two instances to Not Used.

9
Figure 6. Database Configuration Assistant: Database Service MSL5000
For the Oracle Service MSL5000, we can see below how only instances RAC1 and RAC2 are
available for this service:
Figure 7. Oracle Service MSL5000

10
The configuration can be verified using the following command:
oracle@ita018[RAC1]:/home/oracle$ srvctl status service -d RAC -s BACKUP,MSL5000
Service BACKUP is running on instance(s) RAC1, RAC4, RAC3, RAC2
Service MSL5000 is running on instance(s) RAC1, RAC2
Oracle Online Backup Configuration
This section discusses topics such as Backup Session Failover and Backup Load Balancing.
In a large customer environment, hundreds of Oracle Databases are backed up daily. Backup
sessions have to run independently from the status of some of the RAC instances. While it is critical to
have regular backups for the availability of the data, it is also crucial to backup the archive logs on a
daily basis to avoid the archive log destination to fill up
The idea of Session Failover is to be able to have the backup session running although the configured
instance is not available. Session failover applies only to the start of the backup session and not to an
already running session.
Backup Load Balancing distributes the Backup Load over more than one RAC node. As a result, more
than one node will read the datafiles and write the data to tape. This is also valid for a Restore
Session.
Configuration Guidelines
The following is a list of the main tasks needed for the Oracle integration. It is only for reference and
does not replace the Data Protector Oracle online integration manual.
1. Install the Data Protector Oracle agent on each and every RAC Node
2. Add a root account has to the Data Protector Admin User group in the Users configuration
menu for every RAC Node and every virtual hostname.1(1)
3. Add the oracle account to the Data Protector Admin User group in the Users configuration
menu for every RAC Node and every virtual hostname.
4. Link the Oracle SBT library (located under $ORACLE_HOME/lib) with the Data Protector SBT
library on every node as oracle user. It can be done online (the link process is described in
the Data Protector Oracle integration documentation). This is step is not required in Data
Protector 6.1.
5. Verity that the Oracle SQLNET connectivity works between each node and the RMAN
recovery catalog
6. Optional: register database RAC in the recovery catalog (this step is done through the
Integration). The integration needs a database account with SYSDBA privileges. You can use
the Oracle Database built-in SYS account or you may prefer to create a new account (i.e.
dp_integration) and grant it SYSDBA (NOTE: the grant has to be done on each node of the
cluster).
7. Import the Virtual IP Address (VIP) of every node in the RAC cluster; this will enable us to
configure the virtual IP as the client when doing the Integration.
1A new enhancement has been developed, which doesn’t require the root account but only the oracle account is required. Contact your support
organization to receive the needed patch.

11
The following screenshot shows a list of the available clients in our cell server:
Figure 8. Data Protector Cell Clients
At this point, there are 4 main strategies to configure a backup specification of the Oracle RAC
database
•Configure the backup using the node name and instance, i.e. ita018 and RAC1
•Configure the backup using a VIP and all instances, i.e. ita018-vip and RAC1,RAC2,RAC3,RAC4
•Configure the backup using a VIP and a service name, i.e. ita018-vip and BACKUP
•Configure the backup using a VIP and allocate the channels manually
We perform an Online Backup of the tablespace POOL for all following backups. As it has a large
number of datafiles, we can observe how the distribution of datafiles over channels is achieved. The
session report for each of these sessions can be found in the Appendix C.
Use the hostname and the Instance name to configure the backup
In this case we use the hostname ita018 as client and the instance name RAC1 to configure the
integration. The backup will run exclusively on one node, i.e. ita018. If the instance RAC1 is not
available, the backup will fail.
Figure 9. Oracle Integration configuration windows

12
This configuration has the drawback of not providing any failover in case the configured instance is
not available. If the session is scheduled and the instance is down, the backup session will fail.
In addition, only one node in the cluster is performing the backup.
Use a VIP and the Instance names to configure the backup
In this case, we need to use a VIP as a client to be able to provide failover. We provide the
integration a list with the Instances (RAC1, RAC2, RAC3, and RAC4).
Figure 10. Oracle Integration configuration windows
Note that we configured all available RAC instances names as a service name; this provides no load
balancing, but failover in case one instance is available. The Backup will run on just ONE instance,
but the session will not fail as long as there is at least one instance of the list running.
First of all, Data Protector verifies the connection to all specified instances and performs the Backup
on just one of them; in our case, all channels are allocated on the instance which is local to the VIP
configured (in our case RAC1, running on ita018). Refer the Backup Session 2 (Appendix C) for
more details.
As shown in the following output, node ita018 is currently listening to VIP ita018-vip:
root@ita018:/.root# netstat -i
Name Mtu Network Address Ipkts Ierrs Opkts Oerrs Coll
lan1 1500 10.0.0.0 ita018-priv 21210969 0 17790264 0 0
lan0 1500 16.57.72.0 ita018 20450505 0 67265569 0 0
lo0 4136 loopback localhost 34924594 0 34924590 0 0
lan0:804 1500 16.57.72.0 ita018-vip 118277099 0 325092 0 0
Use a VIP and a service name to configure the backup
We have defined two services for backup purposes previously. One of them, Service BACKUP,
allocates the channels amongst all four RAC instances. The second one, Service MSL5000, will use
only RAC Instances RAC1 and RAC2.

13
The Oracle integration uses a VIP as a client to guarantee failover.
Figure 11. Oracle Integration configuration windows
Node Affinity awareness: In a RAC environment, some nodes of the cluster have faster access to
specific datafiles. RMAN is aware of the node affinity, and will try to assign specific datafiles to the
channels of the nodes which have faster access to these datafiles.
In the Appendix C, you can find two session reports using this configuration:
•Backup using service BACKUP (Session 3)
•Backup using service MSL5000 (Session 4)
Use a VIP and allocate the channels to the instances manually
In the above examples, when we define a Service with Load Balancing (for instance BACKUP and
MSL5000), we let Oracle take the decision about the number of channels which will be started per
configured instance. As it can be seen in some session reports, it could happen that channel
distribution is not optimal for the backup (that is, only two of the four instances have channels or one
instance gets three out of four channels).
To overcome this situation, channels can also be allocated to specific instances manually. In this case,
it is recommended that one channel is allocated per drive device. The optimal situation is to have the
same number of drives as instances, thus allocating one channel per instance (each node has a
dedicated drive). This approach has a downside: if one of the instances is not available at backup
time, the backup session will fail (the channel cannot be allocated on that node).
In our environment, we allocate two channels, one on RAC1 and one on RAC2, writing to two tape
drives. To allocate channels manually, we need to edit the RMAN script in the GUI. For this purpose,
create a new backup specification; choose the drives for the backup with the right number of
channels (tape concurrency), save it and then edit the RMAN script by adding the connection string at
the end of the channel specification.

14
Although the backup specification could also be edited on the cell server, located in
$OMNICONFIG/barlists/oracle8, using the GUI is the only supported method to edit a backup
script.
Figure 12. Data Protector RMAN script edit window
The complete RMAN script is shown below:
run {
allocate channel 'dev_0' type 'sbt_tape'
parms 'ENV=(OB2BARTYPE=Oracle8,OB2APPNAME=RAC,OB2BARLIST=RAC_POOL_man_ch_1_2)'
connect 'sys/xxx@rac1';
allocate channel 'dev_1' type 'sbt_tape'
parms 'ENV=(OB2BARTYPE=Oracle8,OB2APPNAME=RAC,OB2BARLIST=RAC_POOL_man_ch_1_2)'
connect 'sys/xxx@rac2';
backup incremental level <incr_level>
format 'RAC_POOL_man_ch_1_2<RAC_%s:%t:%p>.dbf'
tablespace 'POOL';
}
As it can be seen, two channels will be allocated, one on instance RAC1 and one on instance RAC2.
For each of these channels, we have to specify the connection string using the instance name (not a
service).
In this scenario, if one of the instances is not available the channel allocation fails and the session will
abort.
The backup session report is in the Appendix C under Session 5.
Storing the connection string in the Recovery Catalog
Security concerns may arise when storing the connection string in the backup specification (located in
the Cell Manager File system); the channel allocation with the connection string can be defined and
stored in the recovery catalog as a configuration parameter.
•Create a backup specification for your database. Once saved, edit it and remove the channel
allocation parameters; they will be allocated by RMAN using the Recovery Catalog settings.
run {
backup incremental level <incr_level>
format 'RAC_POOL_sbt_rcvcat<RAC_%s:%t:%p>.dbf'
tablespace ''POOL'';
}

15
Figure 13. Data Protector RMAN script edit window
•Connect with RMAN to the instance and recovery catalog and specify the channel parallelism we
want to use; in our case parallelism 2 (two tapes devices)
RMAN> CONFIGURE DEVICE TYPE 'SBT_TAPE' PARALLELISM 2 BACKUP TYPE TO BACKUPSET;
•In the recovery catalog, configure the channels with the connection string. We define two SBT
channels, 1 connecting to instance RAC1 and 2 connecting to RAC2. For the ENV parameters:
–OB2BARTYPE = refers to the Oracle integration: Oracle8
–OB2APPNAME = Database Name (not instance): in our case RAC
–OB2BARLIST = Name of the backup specification defined in the first step:
RAC_POOL_sbt_rcvcat
–CONNECT connection information for the instance
RMAN> CONFIGURE CHANNEL 1 DEVICE TYPE 'SBT_TAPE' PARMS
'ENV=(OB2BARTYPE=Oracle8,OB2APPNAME=RAC,OB2BARLIST=RAC_POOL_sbt_rcvcat)' CONNECT
'sys/xxx@RAC1';
RMAN> CONFIGURE CHANNEL 2 DEVICE TYPE 'SBT_TAPE' PARMS
'ENV=(OB2BARTYPE=Oracle8,OB2APPNAME=RAC,OB2BARLIST=RAC
_
POOL
_sbt_rcvcat)' CONNECT 'sys/xxx@RAC2';
Below, the summary of Recovery Catalog saved parameters for the database RAC:
RMAN> show all;
RMAN configuration parameters are:
CONFIGURE RETENTION POLICY TO REDUNDANCY 1; # default
CONFIGURE BACKUP OPTIMIZATION OFF; # default
CONFIGURE DEFAULT DEVICE TYPE TO 'SBT_TAPE';
CONFIGURE CONTROLFILE AUTOBACKUP ON;
CONFIGURE CONTROLFILE AUTOBACKUP FORMAT FOR DEVICE TYPE DISK TO '%F'; # default

16
CONFIGURE CONTROLFILE AUTOBACKUP FORMAT FOR DEVICE TYPE SBT_TAPE TO '%F'; #
default
CONFIGURE DEVICE TYPE DISK PARALLELISM 4 BACKUP TYPE TO COMPRESSED BACKUPSET;
CONFIGURE DEVICE TYPE 'SBT_TAPE' PARALLELISM 2 BACKUP TYPE TO BACKUPSET;
CONFIGURE DATAFILE BACKUP COPIES FOR DEVICE TYPE DISK TO 1; # default
CONFIGURE DATAFILE BACKUP COPIES FOR DEVICE TYPE SBT_TAPE TO 1; # default
CONFIGURE ARCHIVELOG BACKUP COPIES FOR DEVICE TYPE DISK TO 1; # default
CONFIGURE ARCHIVELOG BACKUP COPIES FOR DEVICE TYPE SBT_TAPE TO 1; # default
CONFIGURE CHANNEL DEVICE TYPE DISK FORMAT '+dg_fra/rac/backup/%U';
CONFIGURE CHANNEL 1 DEVICE TYPE 'SBT_TAPE' PARMS
'ENV=(OB2BARTYPE=Oracle8,OB2APPNAME=RAC,OB2BARLIST=RAC_POOL_sbt_rcvcat)' CONNECT
'*';
CONFIGURE CHANNEL 2 DEVICE TYPE 'SBT_TAPE' PARMS
'ENV=(OB2BARTYPE=Oracle8,OB2APPNAME=RAC,OB2BARLIST=RAC_POOL_sbt_rcvcat)' CONNECT
'*';
CONFIGURE MAXSETSIZE TO UNLIMITED; # default
CONFIGURE ENCRYPTION FOR DATABASE OFF; # default
CONFIGURE ENCRYPTION ALGORITHM 'AES128'; # default
CONFIGURE ARCHIVELOG DELETION POLICY TO NONE; # default
CONFIGURE SNAPSHOT CONTROLFILE NAME TO
'/opt/oracle/product/RAC10g/dbs/snapcf_RAC1.f'; # default
As it can be seen, the connection string is hidden.
Backup Configuration Recommendations
Based on the observations during the testing and the results presented through this paper, this is a list
of recommendations and best practices to follow when configuring Data Protector with Oracle 10g
RAC:
Service versus channel allocated manually
The following table summarizes all four alternatives discussed previously:
Configuration Failover Load Balancing Performance
Hostname + Instance NO NO One Instance is used for the backup
VIP + all instances YES NO Any available instance is used for the
backup
VIP + Service YES YES Limited control over where channels are
allocated. In general, unbalanced
VIP + manual channel NO YES One channel on each host. Best results, Best
load distribution
The above table suggests using an Oracle Service to configure the integration for better availability.
Although the performance will not necessarily be the best, if you need to consider availability and
performance, it is the better option.
On the other side, if performance is the key factor for your backups, due to device or time constraints,
then allocating the channels manually will provide you the best throughput.
Use Concurrency of 1
In Data Protector the term Concurrency defines the number of channels allocated per tape device. For
instance, if we set a concurrency of 2 on one tape drive, we will have 2 RMAN channels writing
concurrently onto the tape; thus backup sets are written multiplexed to tape. This is not necessarily the
best approach when it comes to restores.
Normally, it shouldn’t be necessary to use more than one RMAN channel to stream the Tape Device.

17
When performing an Oracle restore, RMAN will only read with concurrency ONE from tape (just one
channel at a time), independent from the concurrency used during the backup.
The recommendation is to use always concurrency of 1 (one RMAN channel per tape drive)
Set Maxopenfiles to 1
RMAN channel parameter MAXOPENFILES should be set to 1 in case you are using a Disk Array
such as XP or EVA where the data is stripped over different physical disks. As a result, we guarantee
that every RMAN channel reads just one file at a time (no file multiplexing). Please note this has to be
set on each channel manually.
Tape block size should match RMAN output buffer size
The block size of your tape device will determine the I/O size of every write operation. In general,
the larger the block size, the faster backup you achieve. On the other hand, the tape block size
should match the Oracle write buffer size.
When an RMAN channel reads a datafile, each block is placed into an input buffer. After some block
validation checks are performed, it is copied into an output buffer and then sent to the Backup Device.
As documented by Oracle, when setting the parameter MAXOPENFILES < 4, each buffer will be 1MB
in size, with a total buffer size of 16MB (per channel), thus 16 Buffers per file. The buffer size used
currently, can be verified in the v$BACKUP_ASYNC_IO view.
When writing to the SBT device (TAPE), Oracle allocates also 4 output buffers per channel, each
buffer 256Kb, 1MB in total. The size of the write buffers can be adjusted with the BLKSIZE channel
parameter.
While the default output buffer size of 256K may be adequate for most environments, it may be worth
increasing the RMAN output buffer size and Tape block size to 1MB.
An example of an RMAN script with output buffer size set to 1MB and MAXOPENFILES to 1 is as
follows; as channel parameters, they have to be set for every channel.
run {
allocate channel 'dev_0' type 'sbt_tape'
parms 'ENV=(OB2BARTYPE=Oracle8,OB2APPNAME=RAC,OB2BARLIST=RAC_POOL_man_ch_1_2)
,BLKSIZE=1048576' connect 'sys/xxx@rac2' maxopenfiles 1;
allocate channel 'dev_1' type 'sbt_tape'
parms 'ENV=(OB2BARTYPE=Oracle8,OB2APPNAME=RAC,OB2BARLIST=RAC_POOL_man_ch_1_2)
,BLKSIZE=1048576' connect 'sys/xxx@rac3' maxopenfiles 1;
backup incremental level <incr_level>
format 'RAC_POOL_man_ch_1_2<RAC_%s:%t:%p>.dbf'
tablespace 'POOL';
}
Use Asynchronous I/O if available
Asynchronous I/O on HP-UX is only available when using RAW devices. In other case, consider using
the BACKUP_TAPE_IO_SLAVES (default FALSE) Oracle parameter to simulate asynchronous I/O to
tape. Synchronous I/O has a performance disadvantage as every server process has perform a task
at a time.
Use a Recovery Catalog
Use a recovery catalog to configure the Oracle integration as a centralized store for metadata and
scripts. The recovery catalog can also be exported (RMAN user schema) by Data Protector.
For further Backup-tuning recommendations, review the Appendix.

18
SAN Backup versus LAN Backup
Avoid having data being sent over LAN unnecessarily when working in a SAN environment. In the
following two figures, we illustrate two possible scenarios describing SAN backup versus LAN backup
in a RAC environment.
Note
Disk Agent: process which reads data from disk in a backup and writes data to disk in a restore. In
an Oracle environment, instead of Disk Agents, Oracle processes are started.
Media Agent: in a backup it writes the backup to disk or tape, in a restore reads the data from tape
or disk
Disk Agent and Media Agent are NOT allocated on the same host
1. A Disk Agent (DA) is started on host ita018 which reads data from the shared storage disks.
2. Once the data is on the host, it is sent to host ita023 (via LAN)
3. Ita023 starts a Media Agent (MA) to write the data to the SAN tape.
This is obviously extremely inefficient, as host ita018 should write directly to the SAN device.
Figure 14. Backup data flow: DA sends data to MA over LAN
Disk Agent and Media Agent are allocated on the same host
1. As in the above example, a Disk Agent (DA) is started on host ita018 to read data from the
shared devices.
2. A Media Agent is started on the same host to write the data to the SAN Library. In this setup,
the data transfer between DA and MA is done using memory buffers (and not LAN)

19
Figure 15. Backup data flow: DA and MA exchange data using memory buffers
File System Backup: Oracle Base and Cluster configuration
After reviewing the online backup configuration of the RAC database, we need to focus on the
backup of the Oracle software directories and the cluster configuration files. As these components are
not addressed in the Online backup, we need to define an additional File System backup
specification.
Important Note:
These files need to be backed up on a regular basis or when there are configuration changes.

20
Oracle Cluster Registry (OCR)
The Oracle Cluster Registry is a binary file which can be mirrored for redundancy. It holds the entire
cluster related information, such as instances names, resources, etc.
The location of the OCR is /var/opt/oracle/ocr.loc:
oracle@ita018[RAC1]:/home/oracle$ cat /var/opt/oracle/ocr.loc
ocrconfig_loc=/hpap/rdsk/hpap0
local_only=FALSE
Oracle provides an automatic mechanism which performs regular backups of the OCR. The OCR
daemons are in charge of these periodic snapshots. It has to be mentioned that these backups are
only taken on the OCRMASTER host (as long as it is available).
It can be verified with the following utility:
oracle@ita018[RAC1]:/home/oracle$ ocrconfig -showbackup
ita018 2007/09/10 10:43:57 /opt/oracle/product/CRS/cdata/crs
ita018 2007/09/10 06:43:57 /opt/oracle/product/CRS/cdata/crs
ita018 2007/09/10 02:43:56 /opt/oracle/product/CRS/cdata/crs
ita018 2007/09/09 02:43:54 /opt/oracle/product/CRS/cdata/crs
ita018 2007/09/04 14:43:11 /opt/oracle/product/CRS/cdata/crs
If we look in the default location of the OCR backups, $ORA_CRS_HOME/cdata/”cluster_name”, we
see the different backups which are kept:
oracle@ita018[RAC1]:/opt/oracle/product/CRS/cdata/crs$ date
Mon Sep 10 14:14:12 METDST 2007
oracle@ita018[RAC1]:/opt/oracle/product/CRS/cdata/crs$ ll -rta
total 109026
drwxrwxr-x 4 oracle oinstall 96 Jul 20 17:02 ..
-rw-r--r-- 1 root root 8654848 Sep 4 14:43 week.ocr
-rw-r--r-- 1 root root 9433088 Sep 9 02:43 day.ocr
-rw-r--r-- 1 root root 9433088 Sep 10 02:43 day_.ocr
-rw-r--r-- 1 root root 9433088 Sep 10 02:43 backup02.ocr
-rw-r--r-- 1 root root 9433088 Sep 10 06:43 backup01.ocr
drwxrwxr-x 2 oracle oinstall 1024 Sep 10 10:43 .
-rw-r--r-- 1 root root 9433088 Sep 10 10:43 backup00.ocr
These files will be backed when we perform a file system backup of the ORA_CRS_HOME directory.
Oracle Voting Disk
The Oracle Voting Disk is used mainly as a Disk Heartbeat by the Node Monitor to detect “split
brain” situations and decide on the node eviction.
There is no automatic mechanism provided by Oracle for backing up the Oracle Voting Disk. As we
need to guarantee the recoverability of the Voting Disk in case of loss, at least one backup needs to
be performed after the cluster installation and every time the Voting configuration is changed
(add/remove of nodes).
The location of the Voting Disk can be queried with this command:
oracle@ita018[RAC1]:/home/oracle$ crsctl query css votedisk
0. 0 /hpap/rdsk/hpap1
located 1 votedisk(s).