Documente Academic
Documente Profesional
Documente Cultură
Introduction
Partition the Disks
ASMLib Installation
ASM Creation
Database Creation
Performance Comparison
Related articles.
Introduction
Automatic Storage Management (ASM) simplifies administration of Oracle related files by
allowing the administrator to reference disk groups rather than individual disks and files,
which ASM manages internally. On Linux, ASM is capable of referencing disks as raw
devices or by using the ASMLib software. This article presents the setup details for using
either raw devices or ASMLib, as well as the procedures for converting between both
methods.
The article assumes the operating system installation is complete, along with an Oracle
software installation. The ASM instance shares the Oracle home with the database instance. If
you plan on running multiple database instances on the server the ASM instance should be
installed in a separate Oracle home.
Note: When running Oracle 10g Release 2 on RHEL 4 you should consider reading this
article: Using Block Devices for Oracle 10g Release 2 in RHEL 4
# ls sd*
sda sda1 sda2 sdb sdc sdd
# fdisk /dev/sdb
Device contains neither a valid DOS partition table, nor Sun, SGI or OSF
disklabel
Building a new DOS disklabel. Changes will remain in memory only,
until you decide to write them. After that, of course, the previous
content won't be recoverable.
The number of cylinders for this disk is set to 1305.
There is nothing wrong with that, but this is larger than 1024,
and could in certain setups cause problems with:
1) software that runs at boot time (e.g., old versions of LILO)
2) booting and partitioning software from other OSs
(e.g., DOS FDISK, OS/2 FDISK)
Warning: invalid flag 0x0000 of partition table 4 will be corrected by
w(rite)
Command (m for help): n
Command action
e
extended
p
primary partition (1-4)
p
Partition number (1-4): 1
First cylinder (1-1305, default 1):
Using default value 1
Last cylinder or +size or +sizeM or +sizeK (1-1305, default 1305):
Using default value 1305
Command (m for help): w
The partition table has been altered!
Calling ioctl() to re-read partition table.
Syncing disks.
#
The remaining disks ("/dev/sdc" and "/dev/sdd") must be partitioned in the same way.
ASMLib Installation
This step is only necessary if you want to use ASMLib to access the ASM disks.
Determine your kernel version using the following command as the root user.
# uname -r
2.6.9-34.ELsmp
#
Download the ASMLib software from the OTN website, making sure you pick the version
that matches your distribution, kernel and architecture. For this example I used CentOS 4.3,
so the following packages were required.
oracleasm-support-2.0.1-1.i386.rpm
oracleasmlib-2.0.1-1.i386.rpm
oracleasm-2.6.9-34.ELsmp-2.0.1-1.i686.rpm
[100%]
[ 33%]
[ 67%]
[100%]
[
[
[
[
[
OK
OK
OK
OK
OK
]
]
]
]
]
Once the kernel module is loaded, stamp (or label) the partitions created earlier as ASM
disks.
# /etc/init.d/oracleasm createdisk
Marking disk "/dev/sdb1" as an ASM
# /etc/init.d/oracleasm createdisk
Marking disk "/dev/sdc1" as an ASM
# /etc/init.d/oracleasm createdisk
Marking disk "/dev/sdd1" as an ASM
#
VOL1 /dev/sdb1
disk:
VOL2 /dev/sdc1
disk:
VOL3 /dev/sdd1
disk:
OK
OK
OK
If this were a RAC installation, the disks would only be stamped by one node. The other
nodes would just scan for the disks.
# /etc/init.d/oracleasm scandisks
Scanning system for ASM disks:
#
OK
Run the following commands and add them the "/etc/rc.local" file.
chown
chown
chown
chmod
chmod
chmod
oracle:oinstall /dev/raw/raw1
oracle:oinstall /dev/raw/raw2
oracle:oinstall /dev/raw/raw3
600 /dev/raw/raw1
600 /dev/raw/raw2
600 /dev/raw/raw3
ASM Creation
Creation of the ASM instance is the same, regardless of the use of ASMLib or raw devices.
When using ASMLib, the candidate disks are listed using the stamp associated with them,
while the raw devices are listed using their device name.
To configure an ASM instance, start the Database Configuration Assistant by issuing the
"dbca" command as the oracle user. On the "Welcome" screen, click the "Next" button.
Select the "Configure Automatic Storage Management" option, then click the "Next" Button.
If the Oracle Cluster Syncronization Service (CSS) is not currently running, a warning screen
will be displayed. Follow the instructions and click the "OK" button. Once you've returned to
the previous screen, click the "Next" button again.
Enter a password for the ASM instance, then click the "Next" button.
Once the ASM instance is created, you are presented with the "ASM Disk Groups" screen.
Click the "Create New" button.
On the "Create Disk Group" screen, enter Disk Group Name of "DATA" and select the
required level of redundancy:
External - ASM does not mirror the files. This option should only be used if
your disks are already protected by some form of redundancy, like RAID.
Normal - ASM performs two-way mirroring of all files.
In this example, the "High" redundancy is used. Select all three candidate disks and click the
"OK" button. The following image shows how the candidate disks are displayed when using
ASMLib.
When using raw devices, the candidate discs are listed using the devide names.
Database Creation
Before continuing with the database creation, check the listener is up and the ASM instance
has registered with it. Start the listener using the following command.
$ lsnrctl start
LSNRCTL for Linux: Version 10.2.0.1.0 - Production on 29-APR-2006 14:35:46
Copyright (c) 1991, 2005, Oracle.
The ASM instance is not registered, so we can force the registration by doing the following.
$ export ORACLE_SID=+ASM
$ sqlplus / as sysdba
SQL*Plus: Release 10.2.0.1.0 - Production on Sat Apr 29 14:37:06 2006
Copyright (c) 1982, 2005, Oracle.
Connected to:
Oracle Database 10g Enterprise Edition Release 10.2.0.1.0 - Production
With the Partitioning, OLAP and Data Mining options
SQL> alter system register;
System altered.
SQL> exit
Disconnected from Oracle Database 10g Enterprise Edition Release 10.2.0.1.0
- Pr oduction
With the Partitioning, OLAP and Data Mining options
$
Checking the status of the listener shows that the ASM instance is now registered.
$ lsnrctl status
LSNRCTL for Linux: Version 10.2.0.1.0 - Production on 29-APR-2006 14:37:32
Copyright (c) 1991, 2005, Oracle.
Connecting to (ADDRESS=(PROTOCOL=tcp)(HOST=)(PORT=1521))
STATUS of the LISTENER
-----------------------Alias
LISTENER
Version
TNSLSNR for Linux: Version 10.2.0.1.0 Production
Start Date
29-APR-2006 14:35:47
Uptime
0 days 0 hr. 1 min. 46 sec
Trace Level
off
Security
ON: Local OS Authentication
SNMP
OFF
Listener Log File
/u01/app/oracle/product/10.2.0/db_1/network/log/listen er.log
Listening Endpoints Summary...
(DESCRIPTION=(ADDRESS=(PROTOCOL=tcp)(HOST=centos2.localdomain)
(PORT=1521)))
Services Summary...
Service "+ASM" has 1 instance(s).
Instance "+ASM", status BLOCKED, has 1 handler(s) for this service...
Service "+ASM_XPT" has 1 instance(s).
Instance "+ASM", status BLOCKED, has 1 handler(s) for this service...
The command completed successfully
$
Go back to the DBCA and create a custom database in the normal way, selecting the
"Automatic Storage Management (ASM)" storage option.
Enter the ASM password if prompted, then click the "OK" button.
Select the "DATA" disk group, then clicking the "Next" button.
Accept the default "Oracle-Managed Files" database location by clicking the "Next" button.
Enable the "Flash Recovery Area" and Archiving, using the "+DATA" disk group for both.
Continue with the rest of the DBCA, selecting the required options along the way.
Perform the ASMLib Installation, but stop prior to stamping the ASM disk. If you attempt to
stamp the disks using the createdisk command it will fail.
Alter the ASM disk discovery string to exclude the raw devices used previously, then
shutdown the ASM instance.
SQL> ALTER SYSTEM SET asm_diskstring = 'ORCL:VOL*' SCOPE=SPFILE;
System altered.
SQL> SHUTDOWN IMMEDIATE;
ASM diskgroups dismounted
ASM instance shutdown
SQL>
If you are planning to remove the raw device mappings (Raw Device Setup), you could
simply reset the ASM_DISKGROUP parameter.
SQL> ALTER SYSTEM RESET asm_diskstring SCOPE=SPFILE SID='*';
System altered.
SQL>
At this point the disks will not be used by ASM because they are not stamped. As mentioned
previously, the createdisk command used to stamp new disks would fail, so we must issue
the renamedisk command as the root user for each disk.
# /etc/init.d/oracleasm renamedisk /dev/sdb1 VOL1
Renaming disk "/dev/sdb1" to "VOL1":
# /etc/init.d/oracleasm renamedisk /dev/sdc1 VOL2
Renaming disk "/dev/sdc1" to "VOL2":
# /etc/init.d/oracleasm renamedisk /dev/sdd1 VOL3
Renaming disk "/dev/sdd1" to "VOL3":
#
OK
OK
OK
Notice, the stamp matches the discovery string set earlier. The ASM instance can now be
started.
SQL> STARTUP
ASM instance started
Total System Global Area
Fixed Size
Variable Size
ASM Cache
ASM diskgroups mounted
83886080
1217836
57502420
25165824
bytes
bytes
bytes
bytes
SQL>
The ASM instance is now using ASMLib, rather than raw devices. All dependent databases
can now be started.
Alter the ASM disk discovery string to match the raw devices you plan to set up, then
shutdown the ASM instance.
SQL> ALTER SYSTEM SET asm_diskstring = '/dev/raw/raw*' SCOPE=SPFILE;
System altered.
SQL> SHUTDOWN IMMEDIATE;
ASM diskgroups dismounted
ASM instance shutdown
SQL>
Perform all the steps listed in the Raw Device Setup, then start the ASM instance.
SQL> STARTUP
ASM instance started
Total System Global Area
Fixed Size
Variable Size
ASM Cache
ASM diskgroups mounted
SQL>
83886080
1217836
57502420
25165824
bytes
bytes
bytes
bytes
The ASM instance is now using the disks as raw devices, rather than as ASMLib disks. All
dependent databases can now be started.
Performance Comparison
Some documents suggests using ASMLib with Oracle 10g Release 2 gives superior disk
performance, while others say it only reduces the time searching for candidate disks, and
hence ASM startup time. I decided to compare the performance of the two methods myself to
see if I could tell the difference.
My first thought was to perform a simple insert/update/delete test, so I created the following
user and schema for the test in a database using and ASM instance using ASMLib.
export ORACLE_SID=DB10G
sqlplus / as sysdba
Then, as the test user, I ran the following code several times and calculated an average time
for each operation.
SET SERVEROUTPUT ON
DECLARE
l_loops NUMBER := 1000;
l_data VARCHAR2(32767) := RPAD('X', 4000, 'X');
l_start NUMBER;
BEGIN
l_start := DBMS_UTILITY.get_time;
FOR i IN 1 .. l_loops LOOP
INSERT INTO test_tab (id, data) VALUES (i, l_data);
COMMIT;
END LOOP;
DBMS_OUTPUT.put_line('Inserts (' || l_loops || '): ' ||
(DBMS_UTILITY.get_time - l_start) || ' hsecs');
l_start := DBMS_UTILITY.get_time;
FOR i IN 1 .. l_loops LOOP
UPDATE test_tab
SET
data = l_data
WHERE id
= i;
COMMIT;
END LOOP;
DBMS_OUTPUT.put_line('Updates (' || l_loops || '): ' ||
(DBMS_UTILITY.get_time - l_start) || ' hsecs');
l_start := DBMS_UTILITY.get_time;
FOR i IN 1 .. l_loops LOOP
DELETE FROM test_tab
WHERE id = i;
COMMIT;
END LOOP;
DBMS_OUTPUT.put_line('Deletes (' || l_loops || '): ' ||
(DBMS_UTILITY.get_time - l_start) || ' hsecs');
EXECUTE IMMEDIATE 'TRUNCATE TABLE test_tab';
END;
/
The code is purposely inefficient, using a single statement and a commit within a loop for
each operation. Remember, the ASM instance is using high redundancy, so each physical
write operation is effectively done 3 times.
Once the tests on ASMLib were complete, I switched to using raw devices and repeated the
tests. The average results for 1000 of each operation are listed below.
Operation
==============
Inserts (1000)
Updates (1000)
Deletes (1000)
ASMLib (hsecs)
==============
468
956
1281
You can instantly see that the ASMLib results are better than those of the raw devices, but the
testing is suspect for the following reasons:
For each single run of the script, only 1000 operations of each type were
performed. That equates to about 4M of data in the table when it is full.
When you consider the use of the buffer cache, this is a pitiful amount of
data. I originally intended to perform many more operations, but my disk
was grinding so badly I thought better of it.
The tests were performed using VMware virtual disks, so really all this
work was being done on a single SATA disk. I can't be sure if these results
aren't just an artifact of the setup.
Although the average results look convincing, the raw data was so eratic
I'm not convinced these results mean anything.
For these reasons, I decided not to continue to persue the performance tests, much to the
delight of my hard drive. If I get access to a more realistic setup I will attempt some large
scale tests and report the outcome.
For more information see:
Installing ASMLib
Migrating Raw Devices to ASMLib