Monday, October 29, 2007


Oracle's 11g Automatic Storage Management on CentOS 5 (x86_64)




Command "ipcs -m" appears to be almost useless for getting Oracle 11g shared memory
allocation maps.
The basic steps for shared memory monitoring and tuning /dev/shm to handle instances with MEMORY_TARGET more then 1024M had been explained in Tanel Poder’s blog ([1]).
Posting bellow follows advices from Tanel Poder’s blog to build Oracle 11g (64 bit) ASM database with MEMORY_TARGET 1.1 GB on CentOS 5 .
Changes have been made to SPFILE after database install. Restarting database I've got "ORA-00845: MEMORY_TARGET not supported on this system" due to insufficient size
of tmpfs. View [3] regarding definition of MEMORY_MAX_TARGET and MEMORY_TARGET.
I've tuned OS configuration files and oracle shell environment for 64 bit Oracle 11g as advised in [2] for 32 bit Oracle 11g .


Install ASMLIB:-



# rpm -Uvh oracleasm-2.6.18-8.el5-2.0.4-1.el5.x86_64 \
oracleasmlib-2.0.3-1.el5.x86_64 \
oracleasm-support-2.0.4-1.el5.x86_64




Then run:-



# /etc/init.d/oracleasm configure
# /etc/init.d/oracleasm enable




Create ASM disks as requiered (for example):-



# /etc/init.d/oracleasm createdisk VOL1 /dev/sda7
# /etc/init.d/oracleasm createdisk VOL2 /dev/sda8




Tune /dev/shm to handle Oracle Instance with MEMORY_MAX_TARGET 1.2 GB.
MEMORY_MAX_TARGET initialization parameter is a maximum amount of memory that you would want to allocate to the database. It determines the maximum value for the sum of the SGA and instance PGA sizes.Please, view [3] for detailes.








Start OUI in advanced mode and create ASM instance in ORACLE_HOME=/u01/app/oracle/product/11.6.0.1/asm




















Start OUI in advanced mode and create production instance in ORACLE_HOME=/u01/app/oracle/product/11.6.0.1/db_1
























Shared Memory report for production instance:-






Mapping production SGA on it's LGWR daemon address space:-








Mapping ASM SGA on it's LGWR daemon address space:-








Run "ls -l /dev/shm" :-











EM screens snapshots:-








OS detection screen:-








Initialization parameters screen:-








ASM disks used by 11g database:-








Database files layout:-








References
1.http://blog.tanelpoder.com/2007/08/21/oracle-11g-internals-part-1-automatic-memory-management/
2.http://oracle-base.com
3.http://download.oracle.com/docs/cd/B28359_01/server.111/b28310/memory003.htm#BGBCIBHF
4.http://download.oracle.com/docs/cd/B28359_01/server.111/b31107/asminst.htm

Saturday, October 20, 2007


Install Solaris DomU (64 bit) on OpenSUSE 10.3 (x86_64) Dom0



Installing OpenSUSE 10.3 on the box I selected Xen entry into Grub menu after
first phase was completed and system rebooted for final configuration.
It brought me into text mode setup very similar graphical one when you
select ordinary kernel. During this setup I was not given an option to configure any one of three network interfaces (Dual Marvell Yukon Gigabit Ehernet integrated
on the board plus Realtek 8139 Ethernet plugged into PCI slot ).
I have to notice, that SLES 10 SP1 behaves exactly the same way, but during
text mode setup it gave me an option to configure each one of ethernet
interfaces on the box.
When OpenSUSE 10.3 got loaded on top of it's Xen 3.1, ifconfig reported:-








As eth0 was selected one of interfaces provided by Dual Marvell Yukon
Gigabit Ethernet.Yast reported that interface eth0 was configured as DHCP.
Yast also properly recognised all three ethernet interfaces,but configured
only eth0.


Despite that I brought eth0 up with static IP address and
double checked "brctl show" report. It kept stay the same.











Then I made sure that box is available on local network and
replaced corrupted file "libvirt.py" as advised in [1]:-



rpm2cpio libvirt-python-0.3.0-30.i586.rpm | cpio -idv
cd usr/lib/python2.5/site-packages/
cp libvirt.py /usr/lib64/python2.5/site-packages/




Starting from this point I reproduced procedure described in details in [2]:-









Then installed Tight VNC on Solaris DomU ([3]) :-























I've attempted to assign eth0 static IP address before loading Xen 3.1 Kernel
coming with OpenSUSE 10.3. However, this action caused system to hang at boot up.
I've also tried to perform complete install OpenSUSE 10.3 (x86_64) along with Xen in graphical mode (manually configuring ethernet interfaces with static IP addresses)
and after all boot into Xen Kernel. System hanged at boot up again.
I am not sure is this issue hardware dependent or no. Just in case I've also
made a snapshot of "xm info" report:-








References
1.http://lxer.com/module/forums/t/26203/
2.http://lxer.com/module/newswire/view/93016/index.html
3.http://lxer.com/module/newswire/view/93273/index.html

Friday, October 12, 2007


Latest and Greatest Solaris Xen 'drop'



I just have decided that snapshots of the most recent posts at OpenSolaris Xen
Discuss forum would be the best form of advertising regarding upcoming Solaris
Xen Drop.











Jerry Kemp wrote:
> So it would be correct to state that from Solaris Express build 75 on,
> xVM/Xen will be included as a standard part of Solaris Express for x86/x64?

Yes.

MRJ

Tuesday, October 09, 2007


Install Xen 3.0.4 OpenSolaris DomU on SLES 10 SP1 Dom0 (64 bit)



In general installation follows [1]. However, when SLES 10 is placed on top of it's Xen 3.0.4, "ifconfig" shows network interface eth0 without any IP and brings up eth1 with original IP value been assigned to eth0. Command "brctl show" reports that virual network interfaces peth0,vif0.0 are connected to bridge xenbr0. To make OpenSolaris PVM be able to communicate with SLES 10 SP1 Dom0, i did:-



# ifconfig eth1 down
# ifconfig eth0 up "Original IP address" netmask 255.255.255.0




"brctl show" report kept it's output the same. Procedure described in [1] went smoothly
and properly built OpenSolaris DomU on SLES 10 SP1 Dom0
Start up screenshots follows bellow:-

















References

1.Install Xen 3.1 Solaris DomU on CentOS 5 or Debian Etch Dom0 (64 bit)

Tuesday, October 02, 2007


Install Oracle 10.2.0.1 on OpenSolaris DomU at Linux Dom0 (64 bit)



This installation has been done to evaluate OUI (disk I/O) performance on OpenSolaris DomU
versus OUI performance on Linux (CentOS 4.5,5.0) HVM VMs on the same box.
Difference appears to be very noticeable.


Regarding VNC install on OpenSolaris DomU at Linux Dom0 view "Install VNC on Solaris DomU at CentOS 5 or Debian Etch Dom0 (64 bit) at Lxer.com ([2]).
Login as "oracle" to OpenSolaris DomU graphical console.








As root tune oracle's project , create required directories ,then "su - oracle" and create .bash_profile as advised in [1].
Start OUI in advanced mode with option "Install software only".








When done, run "dbca" and create database:-








Then run "netca" to create LISTENER and perform Net Service Name configuration:-






Next :-



[oracle@ServerSun511]$ vi /var/opt/oracle/oratab
To replace "N" by "Y" at the end unique entry line
[oracle@ServerSun511]$ emctl stop dbconsole
[oracle@ServerSun511]$ dbshut
[oracle@ServerSun511]$ lsnrctl stop
[oracle@ServerSun511]$ lsnrctl start
[oracle@ServerSun511]$ dbstart
[oracle@ServerSun511]$ isqlplusctl start
[oracle@ServerSun511]$ emctl start dbconsole




Verify connections to Oracle Instance on OpenSolarisDomU from Linux Dom0:-














References.
1.http://www.oracle-base.com/articles/10g/OracleDB10gR2InstallationOnSolaris10.php
2.http://lxer.com/module/newswire/view/93273/index.html