Mark E. FullerSenior Principal InstructorOracle UniversityOracle Corporation

Oracle and LinuxA Winning Combination -RAC and Linux

Oracle9i Real Application Clusters Database with instances on separate

nodes Physical or logical access to each

database file Software controlled data access

Nodes

Database

Why Implement RAC?

Enables systems to scale up by increasing throughput

Increases performance by speeding up database operations

Provides higher availability Provides support for a greater number of

users

Oracle9i RAC Cache Fusion

100

0.01

120

Blockaccesstime(milli-seconds)

Without Cache Fusion

WithCacheFusion

Block inlocal cache

Block inremote cache

Blockon disk

Why Choose Linux?

Enterprise-class offerings for mission critical systems

Open and cost-effective solution Opportunity to reduce IT infrastructure

costs Oracle Corporation provides direct

technical support of the operating system

Linux RAC Architecture

Hardware– Intel-based hardware– External shared SCSI or Fiber Channel disks– Interconnect by using NIC

Operating system– Red Hat 7.1 and Red Hat 2.1 Advanced

Server– SuSE 7.2 and SuSE SLES7– UnitedLinux 1.0

Linux RAC Architecture

Oracle software– Oracle Cluster File System– Oracle Cluster Management System– Oracle9i Enterprise database

RAC on Linux Storage

Storage options for RAC on Linux:– Oracle Cluster File System (OCFS)– Raw devices named /dev/raw[1-255]

Up to 255 raw devices can be addressed. The tool that is used to set up and query raw

devices is raw.

Currently, Linux has no cluster file system.

– SuSE has a Logical Volume Manager (LVM).– Red Hat has no LVM.

Oracle Cluster File System Is a shared file system that is designed

specifically for Oracle RAC Eliminates the need for database files to

be linked to logical drives Volumes can span one shared disk or

multiple shared disks Guarantees consistency of metadata

across nodes in a cluster

OCFS Features

Node-specific files and directories Unique cluster name integrity

– Allows a hardware cluster to be segregated into logical software clusters

– Simplifies storage area network management

Automatic configuration of new nodes

Cluster Management on Linux oracm maintains both

node status view and Oracle instance status view.

The hangcheck thread driver monitors oracm and reconciles with the hangcheck-timer at defined intervals.

The timer resets the node if a new thread is not started within a specified time.

User mode

Kernel mode

Oracle Cluster Management System (OCMS)

Oracle instanceOracle instance

oracmoracm

Hangcheck thread Hangcheck thread driverdriver

Hangcheck-timerHangcheck-timer

Oracle Cluster Management System OCMS consist of:

– Cluster Monitor (CM)– Node Monitor (NM)– Hangcheck-timer Module

Binaries are located in:– $ORACLE_HOME/ocms/bin

Configuration files located in:– $ORACLE_HOME/oracm/admin/cmcfg.ora– $ORACLE_HOME/oracm/admin/ocmargs.ora

The Hangcheck Timer

Replaces the watchdog daemon in Oracle 9.2.0.2

Is loaded as a kernel module Is specified by the KernelModuleName

parameter in the CMCFG.ORA file Already included in the Linux 2.5 Kernel

The Node Monitor (NM)

Maintains a consistent view of the cluster, and reports the node status to the cluster manager

Uses a heartbeat mechanism Works with hangcheck-timer and acts

depending on the type of failure Is integrated into the cluster monitor

process, oracm, in Oracle 9.2.0.2

The Cluster Monitor (CM)

Maintains the process-level cluster status.

Accepts registration of Oracle instances to the cluster and provides a consistent view of Oracle instances.

When an Oracle process that writes to the shared disk quits abnormally, the CM on the node detects it and takes appropriate action.

RAC/Linux Installation Outline Build hardware cluster and interconnect Install Linux

– Remove IBMJava2 Package– Add pdksh and ssh packages

Prepare Operating System for OCFS,OCMS– Kernel Parameters– Create oracle account and dba group– Create $ORACLE_HOME and /var/opt/oracle

directories– Adjust login profile script for oracle account

RAC/Linux Installation Outline Download and install the following

– http://otn.oracle.com/tech/linux/content.html ocfs-support-1.0-1.i686.rpm ocfs-tools-1.0-1.i686.rpm ocfs-2.4.9-3typeversion.rpm

Start ocfstool gui and generate configuration Create mount point for shared disk Use fdisk to partition the shared disk Start ocfstool gui and format shared disk from

1 node only

RAC/Linux Installation Outline Load ocfs module and mount shared volume Use runInstaller to install OCMS 9.2.0.1

– Accept default watchdog parameter– It will be disabled later

Download and install the hangcheck module– http://otn.oracle.com/tech/linux/content.html

Hangcheck-timer-2.4.9-e.3-0.4.0-2.i686.rpm

Disable watchdog daemon from system startup

RAC/Linux Installation Outline Update OCMS 9.2.0.1 to 9.2.0.2 with

runInstaller Automate OCFS and OCMS commands to

load and execute on system startup Setup rlogin and rsh facilities so the

Universal installer can complete work on both nodes

RAC/Linux Installation Outline Use runInstaller to install the Oracle 9.2.0.1

RDBMS– Select Custom install– Select Oracle9i Real Application Clusters option– Select ‘The management server will use an

existing repository’ Prevents installer from creating OMS repository

now– Choose not to create a database at this time– Cancel Enterprise Manager configuration

RAC/Linux Installation Outline Update RDBMS 9.2.0.1 to 9.2.0.2 with

runInstaller Start gsd daemon with gsdctl utility Run DBCA to create cluster database

– Gsd must be running to succeed

Oracle University Related Offerings Managing Oracle on Linux ( 3 days ILT ) Oracle9i Real Application Clusters ( 3

days ILT ) Oracle9i Database: Real Application

Clusters on Linux ( 2 days ILT )

Oracle Database 10G RAC -Complete, Integrated Clusterware

Complete Oracle cluster software solution

Single-vendor support Low Cost

– No need to purchase additional software

– Easy to install, manage High quality and

functionality across all platforms

Common event and management API’s

– Multi-tier HA & minimal MTTR

Support for third-party clusterware

ConnectivityMessaging and Locking

Cluster Control/RecoveryServices Framework

Oracle Database 10G RAC -Automatic Server Provisioning Push-button add/drop server to cluster

Hands-free allocation and re-allocation of servers to services

– Automatic routing of service requests to appropriate server with lowest load

– On server failure, automatic re-allocation of surviving servers to services

Works across all platforms

AQ&Q U E S T I O N SQ U E S T I O N SA N S W E R SA N S W E R S