Title: Database Upgrade/Migration Options
1Database Upgrade/Migration Options Tips
- Sreekanth Chintala
- Database Technology Strategist
2Agenda
- Very high level overview of Migration/Upgrade
options - No technical details
3About the Speaker
- West Virginia University (M.S)
- Supporter of UT Longhorns
- 11 Yrs of Oracle, 17 years in IT
- RACSIG Web chair (2008-2010)
- Presenter at OIUG, OOW, Local user groups
- Technical Editor
- Sreekanthchintala_at_gmail.com
- Sreekanth_chintala_at_dell.com
3
4Database Upgrade
5Database Upgrade
10gR1
11gR2
6Database Migrations
- Migrations are typically done as part of
End-of-Life refresh, usually followed by Database
Upgrade
7Database Upgrade Methods
- Database Upgrade Assistant GUI
- Uses GUI to upgrade
- The DBUA performs pre-upgrade checks
- Manual Upgrade
- Performed via SQL-Plus Commands
- Use pre-upgrade script to validate
- Modify Cluster Registry entries for RAC Databases
- .profile changes, environment variables
- Tips
- Use GUI method, especially for RAC Databases
- Take a backup before the upgrade
- Check news groups to be aware of any known issues
8Factors
Factors that influence Migration Path
9Concerns
10Database Migration Options
- Complete Downtime
- Export and Import using Data Pump (or exp/imp)
- CTAS
- Backup Restore
- Transportable Tablespaces
- Minimal Downtime
- Logical/Physical Standby
- EMC SRDF Copy
- ASM Rebalance
- Near Zero Downtime
- Oracle Streams
- Golden Gate
11Export / Import
- Database needs to be down or open read only
- Take an export on the source system
- Copy the export dump to destination system
- Import into the database
- Advantages
- Simple to use, proven track record
- Works well for smaller DB sizes (say lt 50GB)
- Gotchas
- Size of the database dictates the amount of
downtime needed - Import time is 3-4 times longer than export
time - Older Unix system may have 2G file size
limitations.
12Export / Import
10gR1
Export Copy
11gR2
Export Copy
- Tips
- Use NFS mount that can be accessed from both
source and destination - Data Pump provides many features
- Review the nature of data. Bring over the static
tables, partitions ahead of time, reduce the
total downtime - In case of RAC instances, leverage all instances
to import different tables - Indexes can be re-built on the destination. No
need to import them - Collect stats after the import
- Beware of security and password changes
13SQL Plus-CTAS
- Database needs to be down or open read only
- Copy the Tables via DB Link
- Gotchas
- Size of the database and the network throughput
dictates the amount of downtime needed - Dont use it on Live tables
- Advantages
- No intermediate storage needed
- Operations can be performed in parallel for
non-related tables
14CTAS
10gR1
11gR2
- Tips
- Review the nature of data. Bring over the static
tables, partitions ahead of time, reduce the
total downtime - In case of RAC instances, leverage all instances
to import different tables - Indexes can be re-built on the destination. No
need to import them - Collect stats after the import
15Backup / Restore
- RMAN incremental backups can reduce the
downtime - Copy the backups to destination system
- Restore into the database
- Gotchas
- Heavy DML operations may prolong the
Restore-Sync time - Going through Tape can significantly add time
to backup and restore - Typically few restore (tape) resources are
allocated - Advantages
- Primary Database is up and running
- Can use RMAN compression, parallel operations
- RMAN only backups the changes and restores the
needed files - Tips
- Use a shared disk (NFS) between the target server
- If you have to go to tape, request dedicated
backup channels until the restore is complete
16Transportable Tablespaces
- Create the empty database
- Create the Metadata of the source database in
the new database - Database files are moved from existing database
to the other system - Gotchas
- Source and target systems should have the same
character set and national character set - Limitations on indexes ( Function based/domain
) - Limitations on snapshot/replication
- Tables and indexes to be part of the self
contained set - All the partitions of a partitioned table needs
to be moved as a set - Advantages
- Use this option to move the databases across
different platforms - Faster because it bypasses usual data extract
and load processes - Efficient than export/import
17Logical/Physical Standby Option
- Available for 10.1.0.3 and above
- Very minimal downtime( time to switchover )
- Must setup a Data Guard environment
- Gotchas
- Need to modify Primary to setup Logical/Physical
Standby - OS must be the same ( family) between primary and
standby - Primary needs to be shut down for DB Upgrade on
the target - SQL Apply reapplies the transactions. Heavy OLTP
systems are not a good candidate for Logical
Standby - Doesnt work for cross platform
- Advantages
- Proven technology
- Changes to the production are copied over to the
standby - Newer versions, you can use it to test the
performance using snapshot standby - Protection from Physical Corruption
18SRDF Copy Option
- Only available for EMC Storage - Symmetrix
Remote Data Facility - Manages real time copies of data volumes ( at
the storage level) - Gotchas
- EMC solution for Symmetrix storage
- Physical corruption will get carried over to the
other side - The target (individual) Lun size should be equal
to or larger than source Lun - Advantages
- Most simple solution to copy multi-terra byte DBs
- Most of the work is on the storage team to setup
SRDF - SRDF can copy the incremental block changes after
initial setup
19ASM Rebalance Option
- Leverage ASM rebalance feature to off-load data
from one storage array to another - Gotchas
- Will require downtime to bring up the new system
- Will require downtime to perform DB upgrades
- Advantages
- No disruption to primary database during
rebalance - No monitoring or governance during the operation
20ASM Rebalance Option
Storage Arrays
Storage Arrays
- Add the New LUNs on the OLD system (OS)
- Add the LUNs to the Disk Group
- Perform Rebalance
- Shutdown Old System
- Create CRS,DB,ASM on new system
- Copy SPFILEs (DB,ASM)
- Add the New LUNs on the New system (OS)
- Bring up the database
21Oracle Streams
- Little or no downtime
- Source and Targets can be any version to any
platform - Gotchas
- Limitations on some data types ( CLOB/BLOB)
- Extremely complex to implement maintain
- Longer development cycles
- Need highly skilled DBAs
- Advantages
- Can be used to migrate between different
platforms and versions
22Golden Gate Option
- Little or no downtime
- Source and Targets can be any version to any
platform - Gotchas
- Requires Golden Gate License
- investment of intermediate platform when moving
from one OS Family to another - Advantages
- Moderately complex to implement
- Provides Fall back options
- You can leverage the infrastructure many times
- Best near Zero Downtime Option
238i 9i ? 10g/11g Migration Overview
Production
Create Clone DB via RMAN
Goldengate Capture to Trails
Server Storage Provisioning
Fail back!
Reverse Replication
Real-Time Replication
Goldengate Apply from Trails
XTTS Transport Tablespace
Fail back Strategy Post-Migration
Switchover to RAC Database
DBA Begins Migration
App Testing
24In place Migrations of OS
- Complete Downtime to the server being upgraded
- Data Guard can be up and running while the
migration is taking place - Operating System Upgraded wipes out internal
drives - Newer Cluster Version wipes out the Clusterware
drives - Data Luns inside ASM are intact.
25In place Migration Steps
- DBA
- Create and save (elsewhere) pfile from the spfile
- Linux Engineer
- Re-image all cluster nodes to new OS
- Post-build steps an 11g RAC cluster.
- Check and validate if the cluster is healthy and
operational - DBA
- Restore the saved pfile configuration files and
start ASM. - Mount the expected ASM disk groups and start the
DB. - By starting the DB under the 11g binaries, this
converted the 10gR2/64 DB to 11g. Run DB Upgrade
script