Slow Oracle Validate Restores...

[B14speedfreak]

Hi All,

Thanks in advance for reading and any posts.

I guess this post could be put into the Oracle 9i forum as well, however the problem seems to be more at the Netbackup side.

********Setup*********
We currently run an Oracle 9i database on a Solaris Server. We run the backups for this system on another Solaris server which has a sun bot, Solaris as an OS, and Netbackup 5.1 with the Oracle Plugin.

In Oracle we can run a Validate Restore on the full backups of the database we are running (its in Archivelog mode). We use the following script which is sheduled via the crontab on the Oracle Solaris Box:

*****script********

#
# Validate Script
#


#Variables

NLS_DATE_FORMAT=DD-MON-YYYY:HH24:MI:SS; export NLS_DATE_FORMAT
ORACLE_HOME=/u01/oracle/prod59db/9.2.0 export ORACLE_HOME
ORACLE_SID=PROD59 export ORACLE_SID
ORAID=oraprod export ORAID
today=`date +"%a"`

#main function

#main part of the script
<<EOF su - oraprod
<<EOF_1 rman

connect target /
connect catalog rman/rman@rman
#list incarnation;
#list backup summary;
run {
allocate channel ch1 type 'SBT_TAPE';
send "NB_ORA_CLASS=ORAhot_PROD59";
send "NB_ORA_CLIENT=eurfinunx001";
#######send "NB_ORA_SERV=eurcomunx001";
send "NB_ORA_SCHED=full";

restore validate database ;

release channel ch1;
}
exit
exit
EOF_1

EOF

exit 0

As you can see its a pretty straight forward script.

******Question************

We currently run this script for all of our production Systems with Oracle on board for the full backups. It runs at a decent rate for 2 of the systems. The other system is running a really slow rate (I.E. 7 hours to read a tape that only took 1 hour to complete the backup). The read rate for the Restore Validate for one of them was 57kb per second. Does anyone run restore Validates? And has anyone had any problems with them taking a LONG time to complete?

Thanks again in advance for any posting and for reading,

B14... aka marky mark...

 

[B14speedfreak]

Just some added info, if its any interest, we are using a 100mb line to connect up the media server to the production box. However as I say the backups seem to run fine, its just the restore validate that seems to run slowly.

Thanks again,

B14... aka... Marky Mark...

 

[comtec17]

Are the backups multiplexed?

 

[B14speedfreak]

The backups go straight to tape, there is no multi-plexing to another location (eg disk).

Thanks for the post,

B14... aka... Marky Mark...

 

[oracagdba]

We are experiencing very slow cross node restores from Veritas 6.0.mp2 from Oracle RMAN. The backup of a 200Gig database takes 30 minutes, the restore was not completed after 36 hours and frequently died. What was the resolution for your slow cross node restores? A point in time restore did complete in a few hours, but a cross node won't.

 

[B14speedfreak]

Firstly what platform are you on? I am currently using Solaris. The restore validates from tape do take a lot longer than the backups. Seconly what kind of line are you using. One thing I found with Solaris is that the line was not full duplex, simple setting but makes a big difference.

Also are you using RAC? If you are you can allocate channels from other nodes, helps speed up backups as well as validate restores, especially if you have mutiable tape drives.

I have some veritas documentunation for 5.1 that might be of help. I will have to post it when I am at work though...

hope it improves,

Mark..