disaster recovery using tdp with rman
DESCRIPTION
Disaster Recovery Using TDP with RMAN. Wylie Merritt ORACLE DBA Oklahoma Dept. of Human Services. Goals. Determine viability of TDP for z/Linux Develop a ‘hot’ backup strategy Test various recovery scenarios, including (especially) disaster recovery Make a recommendation by YE 2003. - PowerPoint PPT PresentationTRANSCRIPT
03/31/04
Disaster Recovery Using TDP Disaster Recovery Using TDP with RMANwith RMAN
Wylie Merritt
ORACLE DBA
Oklahoma Dept. of Human Services
GoalsGoals
Determine viability of TDP for z/LinuxDevelop a ‘hot’ backup strategyTest various recovery scenarios,
including (especially) disaster recovery
Make a recommendation by YE 2003
BackgroundBackground
OKDHS had previously always done file-level backups of a cold database
Migration to a new environment meant new tools
Institutional memory: Warm backups = trouble
BackgroundBackground
RMAN had never been implemented, much less used
IBM was already under contract and Tivoli was already in widespread use on the mainframe side
The BIG Question….The BIG Question….
Can we trust this mission-critical, high-visibility database to RMAN/TDP?
Plan of AttackPlan of Attack
Tivoli Storage Manager was already installed and in use (cold backups)
Tivoli Data Protector (beta) installed in late July ‘03
RMAN hooked up to TDP – it works!Now – How do we use RMAN?
Plan of Attack, Cont.Plan of Attack, Cont.
Preliminary testing promising – Backing up and restoring tablespaces,
data files, etc. works fine– Not excessively complex (though not
graphical)– Ability to do warm backups = a good
thing
BUT...
What about disaster recovery?
Replication as Disaster Recovery POPReplication as Disaster Recovery POP
‘Need to replicate’ already existed Identical Linux guest created for
replication – or so we thought! Lesson learned #1– Tivoli has to be ‘tricked’ into seeing the new
host as the original one– Some critical parameters:
• SErvername – in dsm.sys and dsm.opt• TDPO_NODE – in tdpo.opt
Replication as Disaster Recovery POPReplication as Disaster Recovery POP (Cont.) (Cont.)
What if your RMAN Catalog is gone?Lesson learned #2– Always use ‘controlfile autobackup’• Allows use of backup file (if accessible) to
determine which is the latest full backup, which log files will be required (and need restoring), etc.• As well as insuring you always have a
current control file!
Replication as Disaster Recovery POPReplication as Disaster Recovery POP (Cont.) (Cont.)
How do you connect to a non-existent database instance? (Hint: look at your listener.ora)
Lesson learned #3– Some ‘obscure’ listener.ora parameters come
in handy during ‘disaster recovery’• GLOBAL_DBNAME• SID_NAME• ORACLE_HOME
– In the SID_LIST_<lsnr> section
Replication as Disaster Recovery POPReplication as Disaster Recovery POP (Cont.) (Cont.)
Lesson learned #3 (cont.)– Also need to have defined (in init.ora)• DB_NAME• DB_DOMAIN• DB_NAME + DB_DOMAIN =
GLOBAL_DBNAME (from listener.ora)
Allows you to connect remotely to a non-existent (or down) database instance
Bottom Line – Must HavesBottom Line – Must Haves
Backup or some other way to replicate the environment, including:– Operating system– Oracle (including RMAN)– Tivoli (including Storage Manager and
Data Protector files)
Don’t forget...
All relevant (and relatively static) parameter files, such as tdpo.opt, dsm.opt, dsm.sys, init.ora, listener.ora & tnsnames.ora
Bottom Line – Must Haves Bottom Line – Must Haves (Cont.)(Cont.)
Connection to the new host– Local is good, but remote will work also
(if you’re prepared)A recent RMAN backup that includes– Controlfile autobackup– Log files
ScheduleSchedule
You are here
SO...
We have satisfied ourselves that this is a viable approach, and that TDP will perform as advertised
Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun
Proof of PrincipleProof of PrincipleProof of PrincipleProof of Principle
TestingTestingTestingTesting
ImplementationImplementationImplementationImplementation
Current PlanCurrent Plan
High-availability database will be backed up weekly using RMAN/TDP combination. Est. planned downtime at < 30 min./wk. (currently 1 hr.)
Monthly (?) Flash Copy All other z/Linux Oracle databases will be
backed up weekly using Storage Manager (cold backup, file level)
Go-live date: April 10th
Contact InformationContact Information
DBA – Wylie Merritt– [email protected]– 405.522.1364
Sys Admin (OS) – Chris Little– [email protected]– 405.522.1306
Sys Admin (Tivoli) – Frank Hubble– [email protected]– 405.522.1314