Test Plan for oracle RAC database server
Database failover
Node1
Node
1
Database
Client
Node2
Node
2
Objective Action Expectation Observation Pass/fail
To test the Create a user named HR and a table The sql connection The select still Pass/fail
failover owns by HR. should be able to proceed on
session of a Sql> create user HR identified by HR swing over to another while instance
connected default tablespace USERS temporary instance and still #1 has shut
oracle user in tablespace TEMP; continue the select down.
the event of query.
an instance Sql> create table test (Col1 number);
failure #1 – sqlplus
** note that hr user must be granted connection from
with the rights to select v$intsance client to db is still
views. available
insert over 100,000 lines of data into #2 – the select
the table statement will still
sql> declare proceed on
sql> x number;
sql> begin
sql> for x in 1..10000 loop
sql> insert into test values (x);
sql> end loop;
sql> end;/
sqlplus from the observation client
with Transparent Application Failover
configuration setup.
Ensure that the HR user is connecting
to one of the instances.
C:\ sqlplus hr/hr@mansion
Sql> select instance_name from
v$instance;
Execute a select statement. While the
statement is running, shutdown
immediate on the instance #1.
SQL> select * from test;
telnet into the unix node that the HR
user is connecting to and do a
shutdown.
C:> telnet <oracle_server>
$ ORACLE_SID=mansion1; export
ORACLE_SID
$ sqlplus / as sysdba
Sql> shutdown immediate;
To test the Connect from the observation client The sql session Pass/fail
complete with transparent application failover should not be
failover of configuration to the database via affected and failover
the session in sqlplus to the next node
event of a C:\ sqlplus hr/hr@mansion without any
node (e.g. disruption.
sudden check for the instance ID
shutdown) Sql> select instance from v$instance; The select output
Sql> select * from test; process should not be
affected.
power off the unix box in which the
sqlplus connection is at.
To test the create a win batch job to simulate a Read the instance Across the 20 Pass/fail
load query on the database and print out the name as indicated in connection
balancing of instance in which the connection was the [Link]’s output. made, 50% of
the RAC made [Link]; them are to
under sqlplus hr/hr@mansion @[Link] Out of the 20s the instance
numerous connection, at least #1 and the
connection [Link]; half of them should others to
select instance_name from v$instance; be balanced across instance #2
the two nodes
double-click on the [Link] 20 times
all at one go.
check for the number of
connections .e.g 20 concurrent, and
which instance they have connected
to.
To test the Connect from the observation client The sql session Pass/fail
stability of with transparent application failover should not be
the db by configuration to the database via affected and failover
killing oracle sqlplus and perform select statement. to the next node
server C:\ sqlplus hr/hr@mansion without any
process Sql> select * from tab; disruption.
Kill off a process related to the oracle The select output
daemons from a terminal. This should process should not be
crash the database processes. affected.
$ ps –ef |grep smon |cut –c 10-15|
xargs kill -9
Hardware failover
Objective Action Expectation Observation Pass/fail
To test the pull out one of the interconnect wires The ping should
failover of the and do a ping to the other node via the be successful with
private private address; return messages
interconnect # ping [Link]
To test the Plug out one of the cable connecting The ping from the
failover of the the node to the public network. client should be
network card able to go thru’
of the public Ping from the observation client to the
IP address node
$ ping [Link]
To test the
failover of the Not available…
storage in
event of a
storage device
failure
To test the At one node, pull out the mirror disk The console
failure of a and observe the state of the console should continue
root mirror session. its function
disk without any
Plug it back and do meta replace… disruption.
After the syncing is complete, pull out
the other disk and observe the output
at the console screen.
To test the Plug out 1 hard disk from the storage
failure of a array.
single hard Plug it back in.
disk in the
storage array
<Correct this jdbc string depend on BSP environment setup>
jdbc:oracle:thin:@(DESCRIPTION=
(ADDRESS_LIST=
(ADDRESS=(PROTOCOL=TCP)HOST=sg-int1)(PORT=1521))
(ADDRESS=(PROTOCOL=TCP)HOST=sg-int2)(PORT=1521))
(FAILOVER=on)LOAD_BALANCE=off))
(CONNECT_DATA=(SERVER=DEDICATED)(SERVICE_NAME=mansion)))
