Oracle RAC –

adding and removing

cluster nodes

WLCG Service Reliability Workshop

CERN, November 30th, 2007
Jacek Wojcieszuk, CERN IT

CERN - IT Department
CH-1211 Genève 23
Switzerland LCG
www.cern.ch/it
 Part I

Node addition

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 2
 Methods
• Silent cloning procedures
• Enterprise Manager Grid Control cloning and
adding instance
• addNode.sh and DBCA
• Manual

DEMO
In the presented demo as well as in the command examples appearing in the
presentation, the following names are being used:
demorac – cluster name
itrac18 (demorac1, +ASM1), itrac19 (demorac2, +ASM2) -
existing cluster nodes/instances
itrac20 (demorac3, +ASM3) – node/instances that are being added to
the cluster

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 3
 Steps
• Setup hardware for a new node
• Install and configure OS
• Configure account equivalency
• Add Oracle Clusterware to the new node
• Configure ONS for the new node
• Add ASM and DB home to the new node
• Configure a listener on the new node
• Add an ASM and a database instances

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 4
 Hardware setup
• New node of the cluster should be:
– of similar type as other cluster nodes (the same
architecture e.g. x86_64)
– of similar size to other cluster nodes
– physically connected to the shared storage used
by the cluster
– physically connected to the cluster private
network(s)

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 5
 OS installation and configuration
• Install OS on the new node
– The same flavor and version as installed on other nodes (e.g.
RHEL 4.0)
– The same kernel version recommended
• Make sure that all required packages are installed
– Ideally the list of software packages should be identical on all
cluster nodes
• Configure kernel parameters and shell limits
• Create Oracle-related OS groups and users
• Configure network interfaces
– On one of the old cluster nodes edit /etc/hosts file adding
entries related to the new node; distribute this file to all cluster
nodes
• Configure hangcheck timer module

# as root on the new node:

echo "/sbin/modprobe hangcheck-timer" >> /etc/rc.d/rc.local
echo "session required /lib/security/pam_limits.so" >>
/etc/pam.d/login
/sbin/modprobe hangcheck-timer
CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 6
 Account equivalency
• Passwordless access from one Oracle software
owner account to another has to be configured
– Needed only during software installation
– The easiest is to re-configure the equivalency for the
whole cluster
– Use sshUserSetup.sh script available on metalink
# as oracle user from one of the cluster nodes
./sshUserSetup.sh -hosts „itrac18 itrac18.cern.ch itrac19
itrac19.cern.ch itrac20 itrac20.cern.ch" -advanced

• One can use clufvy tool to verify if the new

machine is ready for clusterware installation
# as oracle user from one of the existing cluster nodes
$ORA_CRS_HOME/bin/cluvfy stage -pre crsinst -n
itrac18,itrac19,itrac20 –r 10gR2 –osdba ci –orainv ci

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 7
 Adding Oracle Clusterware and
configuring ONS
• Run addNode.sh script from an existing cluster node
cd $ORA_CRS_HOME/oui/bin; ./addNode.sh

• After Oracle Universal Installer is started, follow the wizard:

– Provide information related to the node being added
– Launch the installation
– Run the scripts as root on new and old nodes as requested by the
pop-up window
• Close OUI and verify if nodeapps associated with the new
node have been properly added to the Cluster registry
– Use crs_stat command

• Configure the new ONS:

# From one of the old cluster nodes run:
cat $ORA_CRS_HOME/opmn/conf/ons.config |grep remoteport
remoteport=6200
cd $ORA_CRS_HOME/bin
./racgons add_config itrac20:6200
CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 8
 Adding ASM and DB home
(addNode script)
• Depending on the chosen configuration can be
done in either 1 or 2 steps
• Run addNode script on an existing cluster node
cd $ORACLE_HOME/oui/bin; ./addNode.sh

• After Oracle Universal Installer is started, follow the

wizard
– Tick node that you want to add (should be ticked by
default)
– Install the software
– Run root.sh script as requested by the pop-up window

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 9
 Adding ASM and DB home
(cloning)
• Go to an existing cluster node and tar ORACLE_HOME
directory
# on the first node of the cluster
cd $ORACLE_HOME/..
tar cvfp rdbms.tar rdbms

• Copy created tar file to the new node and unpack it

# on the new node
cd $ORACLE_HOME/..
scp oracle@itrac18:$PWD/rdbms.tar .
tar xvfp rdbms.tar

• Remove from the $ORACLE_HOME/dbs directory instance specific

files (pfiles and password files)
• Run cloning procedure
# on the new node
cd $ORACLE_HOME/clone/bin
perl clone.pl ORACLE_HOME="/ORA/dbs01/oracle/product/10.2.0/rdbms"
ORACLE_HOME_NAME="OraDb10g_rdbms"
'-O"CLUSTER_NODES={itrac18,itrac19,itrac20}"' '-O"LOCAL_NODE=itrac20"'

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 10
 Listener configuration
• From the new node run netca to configure a listener:
– Choose ‘Cluster configuration’
– Select the new node only
– Choose ‘Listener configuration’
– Choose ‘Add’
– Leave the default listener name
– Leave intact the default protocol choice
– Specify proper port
– Complete the configuration
• Check with the crs_stat command if the new listener has
been registered to clusterware
• Edit listener.ora and tnsnames.ora files on the new
node:
– You may want to remove extproc related entries
– In tnsnames.ora add/edit entries related to the new listener
• Distribute updated tnsnames.ora file to all nodes of the
cluster

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 11
 Adding ASM and database
instances (with DBCA)
• Although this procedure is recommended by Oracle
we have seen that it is quite unreliable and
therefore we strongly recommend using manual
procedure described later
• Run DBCA from an old cluster node:
– Choose ‘Oracle Real Application Clusters database’
– Choose ‘Instance Management’
– Choose ‘Add an instance’
– Specify SYS username and password
– Select name of the new node
– Modify custom service definitions
– Agree to extend ASM to the new node
– Complete the operation
CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 12
 Adding ASM and database
instances (manually)
• Edit /etc/oratab file on the new node
– add entries describing new ASM and DB instances
• Create dump directories for ASM and DB instances as defined
by initialization parameters
• Create ASM pfile in the default location
($ORACLE_HOME/dbs)
• Create a password file for ASM
orapwd file=$ORACLE_HOME/dbs/orapw+ASM3 password=xxxx

• Connect to an existing ASM instance and define values for init

parameters specific for the new instance
export ORACLE_SID=+ASM1; sqlplus / as sysdba
ASM> alter system set instance_number=3 scope=spfile sid=‘+ASM3’;
ASM> alter system set local_listener=‘LISTENER_+ASM3’
scope=spfile sid=‘+ASM3’;

• Define ASM instance in the cluster registry and start it up

srvctl add asm -n itrac20 -i +ASM3 -o $ORACLE_HOME
srvctl start asm -n itrac20
CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 13
 Adding ASM and database
instances (manually) (2)
• Connect to an existing database instance
– Create an undo tablespace and a redo log thread for the new
RAC node
– Set properly init parameters specific to the new instance
export ORACLE_SID=demorac1; sqlplus / as sysdba
SQL> create undo tablespace undotbs3 datafile size 1G autoextend
on next 1G maxsize 30G;
SQL> alter database add logfile thread 3 group 31 size 512m;
SQL> alter database add logfile thread 3 group 32 size 512m;
SQL> alter database enable public thread 3;
SQL> alter system set instance_number=3 scope=spfile sid=‘demorac3’;
SQL> alter system set thread=3 scope=spfile sid=‘demorac3’;
SQL> alter system set undo_tablespace=‘UNDOTBS3’ scope=spfile
sid=‘demorac3’
SQL> alter system set local_listener=‘LISTENER_DEMORAC3’
scope=spfile sid=‘demorac3’;

• Create DB pfile in the default location ($ORACLE_HOME/dbs)

• Create a password file for the DB instance
orapwd file=$ORACLE_HOME/dbs/orapwdemorac3 password=xxxx
CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 14
 Adding ASM and database
instances (manually) (3)
• Define clusterware targets related to the new
database instance
srvctl add instance -d demorac -i demorac3 –n itrac20
srvctl modify instance -d demorac –i demorac3 -s +ASM3

• Start the new instance using srvctl tool

srvctl start instance -d demorac -i demorac3

• Verify with crs_stat tool if all the clusterware

targets are up

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 15
 Final steps
• Install OEM agent
• Redefine services if needed
srvctl stop service -d demorac -s demorac_lb
srvctl modify service -d demorac -s demorac_lb
–i demorac1,demorac2,demorac3 -n
srvctl start service -d demorac -s demorac_lb

• Update your public tnsnames.ora or LDAP

• During next maintenance you may want to increase
value of the cluster_database_instances
initialization parameter

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 16
 Part II

Node removal

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 17
 Steps
• Stop and delete the DB instance on the
node to be deleted
• Clean up the ASM instance
• Remove the listener from the node to be
deleted
• Remove RAC and ASM software from
oraInventory
• Remove ONS configuration from the node
• Remove the node from the clusterware

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 18
 Deleting DB instance on the node
to be deleted from the cluster
• Redefine services in a way that they do not use the
node being deleted (using srvctl tool)
• From the node that is not being deleted run DBCA:
– Choose ‘Oracle Real Application Clusters database’
– Choose ‘Instance Management’
– Choose ‘Delete an instance’
– Provide SYS credentials
– Select the instance to delete
– Proceed with the deletion
• Verify with the crs_stat command that the
instance has been removed from the cluster registry

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 19
 Cleaning up the ASM instance

• Stop the ASM instance

# on any cluster node
srvctl stop asm -n itrac20

• Remove the ASM instance

# on any cluster node
srvctl remove asm -n itrac20

• Check with crs_stat command if the ASM

instance has been removed from OCR
• On the node being deleted edit /etc/oratab file
– remove entries related to removed DB and ASM instances

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 20
 Removing the listener
• Run netca on the node being deleted:
– Choose ‘Cluster configuration’
– Select the node being deleted
– Choose ‘Listener configuration’
– Choose ‘Delete’
– Complete the deletion
• Check with crs_stat tool if the listener has
disappeared from OCR

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 21
 Removing database software
• Update oraInventory on the node being removed
# on the node being deleted:
$ORACLE_HOME/oui/bin/runInstaller –updateNodeList
ORACLE_HOME=$ORACLE_HOME CLUSTER_NODES=itrac20 -local

• Uninstall the DB software using OUI

# on the node being deleted:
$ORACLE_HOME/oui/bin/runInstaller

• Update oraInventory on other cluster nodes

# on the first node of the cluster
$ORACLE_HOME/oui/bin/runInstaller –updateNodeList
ORACLE_HOME=$ORACLE_HOME CLUSTER_NODES=itrac18,itrac19

• If you use separate software installation for ASM

repeat the steps above for this ASM_HOME
CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 22
 Removing ONS configuration and
cluster software
• Remove ONS configuration
# From the first node of the cluster run:
$ORA_CRS_HOME/bin/racgons remove_config itrac20:6200

• Delete clusterware software from the node

# On the node to be removed as root run:
$ORA_CRS_HOME/install/rootdelete.sh

• Run rootdeletenode.sh script as root from the

first node of the cluster to complete clusterware
removal
# On first node of the cluster as root:
$ORA_CRS_HOME/install/rootdeletenode.sh itrac20

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 23
 Removing cluster software (2)

• Update oraInventory on the node being removed

# on the node being deleted:
$ORA_CRS_HOME/oui/bin/runInstaller –updateNodeList
ORACLE_HOME=$ORA_CRS_HOME CLUSTER_NODES=itrac20 CRS=true -local

• Uninstall the Clusterware software using OUI

# on the node being deleted:
$ORA_CRS_HOME/oui/bin/runInstaller

• Update oraInventory on other cluster nodes

# on the first node of the cluster:
$ORA_CRS_HOME/oui/bin/runInstaller -updateNodeList
ORACLE_HOME=$ORA_CRS_HOME CLUSTER_NODES=itrac18,itrac19 CRS=true

• Verify with the crs_stat command that the node has

been properly removed from the cluster
CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 24
 Final steps (optional)
• Remove binaries left on the removed node
• Deconfigure OS oracle account equivalency
• Stop and remove OEM agent from the node
• Deconfigure private network interfaces

CERN - IT Department
CH-1211 Genève 23
Switzerland
www.cern.ch/it
RAC – adding and removing nodes –
WLCG Service Reliability Workshop, Nov 2007 - 25