Rent-A-RAC

1
Rent-A-RAC

• Cnetics Oracle 10g Real Application Cluster
  Prototype

2
TPC Benchmark Leader

• Fastest and most cost effective platform is
• Oracle 10g RAC HP Itanium2 RedHat Linux3
  (April 14, 2004)
• TPC-C benchmark of 1.1 M transactions at
  5.52/tpmC.
• Faster than HP Integrity Superdome on HPUX with
  Oracle 10g (1.0 M transactions at 8.33/tpmC.

3
(No Transcript)
4
Oracle RAC Portable RAC
Public network
Jamaica2
Jamaica3
Jamaica1
Cluster interconnect
Oracle 10g RAC instance jamc1
Oracle 10g RAC instance jamc2
Oracle 10g RAC instance jamc3
S3
S4
Sn
DB cache
DB cache
DB cache
Cache to cache
Oracle clusterware
Oracle clusterware
Oracle clusterware
Operating system
Operating system
Operating system
Shared storage
Concurrent access from every node scale out
More nodes higher availability
Redo log instance 1
Redo log instance 3
Control files
Database files
5
Rent-A-RAC Software Components

• The Oracle 10g Real Application Cluster has
  several software components that compose the
  system
• Clustered file system
• Clustered servers
• The clustered file system for the Cnetics RAC is
  Oracle Cluster File System.
• The clustered servers are managed by Oracle
  Cluster Ready Services.
• Management of a clustered database is managed by
  Oracle Enterprise Manager Database Control.

6
Oracle Cluster File System

• Oracle Cluster File System (OCFS) presents a
  consistent file system image across the servers
  in a cluster. OCFS allows administrators to take
  advantage of a file system for the Oracle
  database files (data files, control files, and
  redo logs) and configuration files. This eases
  administration of Oracle Real Application
  Clusters.
• http//oss.oracle.com/projects/ocfs/documentation

7
OCFS Features

• The Oracle Clustered file system has several
  features and advantages
• Shared Disk Cluster File system specifically
  suited to Oracle RAC
• Improve management of data for Oracle Real
  Application Clusters (RAC)
• Open Source initiative for Linux.
• Solved the 255 raw device limit on Linux and
  difficulty in management
• Single view of data files across all nodes
• Use of standard OS tools to handle file operations

8
OCFS Limitations

• The Oracle Clustered file system has several
  current limitations
• No Shared Oracle home installation
• Binaries and generic file support on CFS
• Support for other applications (ISV applications)
• Volume management
• Dynamic resizing of partitions
• Defragmentation of filesystem
• Online FSCK (file system check)

9
Cluster Ready Services (CRS)

• CRS daemons are crsd, evmd and ocssd
• Corresponded executables are located in
  CRS_HOME/bin/
• Cluster Ready Services are started at boot by
  /etc/rc.d/init.d/init.crs script.
• To stop and start crsd, cssd, and evmd manually,
  execute
• /etc/rc.d/init.d/init.crs startstopenabledisabl
  e
• Also use service command
• service init.crs startstopenabledisable

10
Clusterware Log Files

• Cluster Ready Services Log Files Cluster Ready
  Services (CRS) has daemon processes that generate
  log information.
• Log files for the CRS daemon (crsd)
• ORA_CRS_HOME/crs/log/ltnode namegt.log
• Log Files The Oracle Cluster Registry (OCR)
• ORA_CRS_HOME/srvm/log/
• Cluster Synchronization Services (CSS) Log Files
  by the OCSSD daemon
• ORA_CRS_HOME/css/log/ocssdltnumbergt.log
• ORA_CRS_HOME/css/init/ltnode_namegt.log
• Event Manager Log Files Event Manager (EVM)
  information generated by evmd
• ORA_CRS_HOME/evm/log/evmdaemon.log
• ORA_CRS_HOME/evm/init/ltnode_namegt.log
• Oracle High Availability Log Files for RAC
• ORACLE_BASE/database_name/admin/hdump

11
RAC Processes

• LMON global enqueue monitor manages global
  enqueues, process expirations global cache
  recovery
• LMD global enqueue daemon lock agent for
  enqueues.
• LMS Lock manager server process global cache
  service requests, cross instance call operations
• LCK cross instance broadcasts
• DIAG diagnosability daemon monitors instance
  health and data for instance process failures

12
RAC Parameter File

• Parameter file is shared by all instances
• Some parameters are global, some are specific to
  each instance
• Set parameters with the alter system command
• Tip Backup the server parameter file
• Initialization parameters types in Real
  Application Clusters include
• Multi-Valued Parameters
• Parameters That Must Be Identical Across All
  Instances
• Parameters That Must Be Unique Across All
  Instances

13
RAC Manage Instances

• Start up / shut down instances with Enterprise
  Manager, SQLPlus or SRVCTL
• Shutting down one instance does not interfere
  with the operation of other running instances
• Shutdown normal immediate do not require
  recovery. Abort does require recovery. Recovery
  is performed by other instances
• Use shutdown transaction local so shutdown does
  not have to wait for other instances

14
Instance Management with Enterprise Manager

• Enterprise Manager http//jamaica15502/em
• Navigate to the instance then shutdown or startup
• OEM asks for host and database credentials
• Choose advanced option for shutdown/startup
  options

15
Enterprise Manager

• Enterprise Manager is a webserver process which
  connects to database
• emctl status dbconsole
• emctl stop dbconsole
• emctl start dbconsole

16
Instance Management with SQLPlus

• Connect with SQLPlus to instance as SYSDBA or
  SYSOPER
• Issue shutdown or startup
• Confirm that connection is made to the correct
  instance
• select instance_name from vinstance
• Only one instance can be managed at a time

17
Instance Managementwith SVRCTL

• SVRCTL can check status of instances databases.
• srvctl status database -d jamc
• SVRCTL can manage instances cluster database
• Start all instances
• srvctl start database -d jamc
• Start stop one instance
• srvctl stop instance -d jamc i jamc3 o
  transactional
• svrctl start instance d jamc i jamc3

18
RAC Load Balancing

• See Metalink Note220970.1
• All load balancing available currently (9i-10gR1)
  occurs at connect time. The following options
  exist
• Random
• Either client side load balancing or hardware
  based methods will give a randomize the
  connections to the instances. On the negative
  side this method is unaware of load on the
  connections or even if they are up meaning they
  might cause waits on TCP/IP timeouts.
• Load Based
• Listener side load balancing redirects
  connections by default depending on the RunQ
  length of each of the instances.
• Session based
• Session based load balancing takes into account
  the number of sessions connected to each node and
  then distributes connections to balance the
  number of sessions across the different nodes.
• Session count balancing is method used when you
  set a listener parameter, prefer_least_loaded_node
  _off.

19
Demo-Transaction Failover

• RAC will carry one connection activity even when
  an instance fails
• Connect to node - Query from a table
  rac_demo.persons
• Use preconnect connection for faster failover
• Shutdown connected node
• Query from the gvsession view for session status

20
Demo - Scalability

• Oracle RAC will distribute load across clustered
  instances
• Generate workload on one node
• Watch the machine load
• Turn on second instance
• Watch the machine load
• Look at connection spread
• Turn on third instance
• Watch the machine load
• Look at connection spread

21
RAC Parallel Execution

• A Parallel action can run across multiple
  instance group
• Parallelism can be set
• For a session
• On a table
• With a hint
• Alter session forceenable parallel query
• Use INSTANCE_GROUP PARALLEL_INSTANCE_GROUP

22
Rent-A-RAC

• Cnetics Oracle 10g Real Application Cluster
  Prototype

23
Questions?

• Cnetics Oracle 10g Real Application Cluster
  Prototype