Oracle 11g Real Application Clusters

1
Oracle 11g Real Application Clusters

• Brock Frank
• CTO

2
Agenda

• Introduction
• Oracle 11g New Features
• ASM
• CRS
• QA
• Oracle 11g Best Practices
• QA

3
Introduction

• Background

4
Oracle 11g New Features

• Install and Upgrade
• Storage
• Oracle Clusterware

5
Install and Upgrade

• DBUA
• Supports RAC, ASM, and patchsets (10.2.0.3 or
  higher) upgrades
• Disables archiving during upgrade
• Performs database cold backup prior to upgrade
• Moves database files between ASM and file system
• Hot Patching using OPatch
• No instance restart or relinking of software
  needed
• Patch shipped as dynamic/shared library
• Memory requirement of one OS page per Oracle
  process
• Listed in OPatch inventory
• Persistent across instance restarts
• Replace with conventional patch at convenient
  downtime

6
Storage

• ASM Fast Mirror Resync
• ASM Preferred Mirror Read
• Direct NFS Client

7
Without ASM Fast Mirror Resync
Oracle Database 10g and 11g
8
With ASM Fast Mirror Resync
Oracle Database 11g
9
ASM Preferred Mirror Read
10
Direct NFS Client

• Performance
• Scalability
• HA

11
NFS Performance

• Direct NFS Client improves performance of NFS
  storage
• Direct I/O
• Bypasses OS Caches
• Eliminates OS write-ordering locks
• Reduces system CPU utilization
• Asynchronous I/O

12
NFS Scability and HA

• Direct NFS Client increases high availability and
  optimizes scalability of NFS storage
• Supports up to 4 parallel network paths to NAS
  system
• Automatically load balances across all available
  network paths (No extra configuration needed!)
• Resends packets over available network paths in
  the event of a network path failure
• No NIC Bonding required

13
NFS DSS Performance Study
14
Oracle Clusterware

• Added the ability to perform rolling upgrades
• No need to shutdown CSR completely to perform an
  upgrade
• Upgrade one node at a time

15
A
16
RAC Best Practices

• More Nodes
• Common Setup
• RAC Setup Recommendations
• Verifying Setup
• Keeping your System Up

17
More Nodes Greater HA
CRM
OE
Payroll
Email
18
More Nodes Greater HA
CRM
OE
Payroll
Email
Node failure has less impact
19
Oracle RAC Customers 9i
20
Oracle RAC Customers 10g
21
Oracle RAC Customers 11g
22
RAC Setup Recommendations

• Full Redundancy Networks, Switch, etc.
• Bonding/Teaming your Networks/SAN
• Use a Switch
• Mirroring OCR
• Setup 3 Voting Disks

23
Verifying Setup Destructive Testing
Crash of a client connection
Crash of the DB instance
Crash of 1 element of the interconnect
Inst DB
Inst DB
Crash of the ASM instance
Inst ASM
Inst ASM
Servers
Crash of each fiber card
Crash of the SAN connection
Storage
Crash of a storage system
24
Keeping your System Up

• Adhere to strong Systems Life Cycle Disciplines
• Comprehensive test plans (functional and stress)
• Rehearsed production migration plan
• Change Control
• Separate environments for Dev, Test, QA/UAT,
  Production
• Backup and recovery procedures
• Security controls
• Support Procedures

25
Test Clusters

• When
• Any application, parameter or component change
• Why
• Doesnt Oracle/OS Vendor/HW Vendor test?
• Your Unique Environment
• What happens if you dont?

26
Change Management

• Plan changes to minimize downtime and service
  disruption
• May mean overnight or weekend work
• Avoid critical periods in application cycles,
  such as month- or year-end processing
• Consider staged changes
• One function at a time
• One node at a time (if possible)
• Include time and resources to back out changes if
  necessary

27
Summary

• To build the a full HA architecture with Real
  Application Clusters
• Use Services FAF
• Use the full stack Oracle Clusterware, ASM or
  NFS
• Validate your environment w/ every change
• Follow MAA Best Practices
• Have a Test Cluster

28
A