Active/Active Cluster with Oracle RAC

[Mousa55]

Hi All,

I done the Active/Passive Cluster MCSG Cluster in 2-node environment (rp8420) with same version (11.11)
with this procedure

sysA: # vgexport –v –p –s –m /tmp/map/vg01.map /dev/vg01
sysA: # rcp vg01.map ruxerp01:/tmp/map
sysB # mkdir /dev/vg01
sysB # mknod /dev/vg01/group c 64 0x010000
# mkdir /mcsg # vgchange –a n vg01 (node2)
# vgchange -a n /dev/vg01 (node 2)
# vgchange -a r /dev/vg01 (read only) (node 2)
etc etc ......

But In the Active/Passive Cluster I can’t mount the VG at two server in the same time
Except mounted as read only option on two node.

My questions how I can configure Active/Active MCSG Cluster with Oracle RAC on (rp8420) server, And with different node version (node 1 (11.11) and node 2 (11.23) ).
How to mount the VG at two server With R/W option with Active/Active cluster.
Please provide the detailed procedure to configure Active/Active Cluster with Oracle RAC.

Thanks & kind Regards

[Hein van den Heuvel]

You have a lot of reading to do my friend!

Start with a google for "Active/Active MCSG Cluster with Oracle RAC" (verbatim from your topic!).

You'll find topics like:

http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1358005

There are good references in there.
For example: http://docs.hp.com/en/T1859-90017/T1859-90017.pdf

Is that vg01 your main (only?) volume group?
You'll want to establish a 'rac' volume group.

For better help you may want to indicate HP-UX and Oracle target versions.

Please consider involving an experienced resource for this transation, be it internal or external (consultant). For that you might look at HP, Oracle or indepdendent companies.

Good luck!
(Judging by your question you'll need some ;-)

Best regards,
Hein.

[Mousa55]

Hi,

thanks for your replay,
"Is that vg01 your main (only?) volume group?" I said that vg01 only for example.
Because I have more VG.
and I read that document before, but is not clear and is not describe how to configure the cluster with oracle RAC .
what is different between Active/Passive and Active/Active in configuration? and how to make 2 server with R/W option on 2 node in Active/Active Cluster with Oracle RAC.
And what is software needed for oracle rac cluster?


Thanks & kind Regards

[Aneesh Mohan]

Hi Nejad,

Using MCSG cluster with Oracle RAC extension (or OPS bundle) you can make active/active volume groups in multiple nodes.It is using shared LVM concept to accomplish that.For example if you are planning to configure VG01 in active/active RAC cluster ,then with the help of RAC extension you can activate VG01 in shared server mode in one node and remaining all other node configured to use VG01 will be in shared client mode .

I belive you can not mix different version of operating systems in a Oracle RAC environment for a same database..

Regards,
Aneesh

[Aneesh Mohan]

continue...

1)
you have to use VGCHANGE="vgchange -a s" default in the pkg cntl files for the RAC vgs.


2)If you are planning RAC with RAW chunks then oracle:dba should have permission to read and write corresponding LV`s.


Aneesh

[Mousa55]

Hi,

did I can make Active/Active (oracle RAC) Cluster by using Active/Passive procedure.
if I can do it, what is the different steps or commands between two type.
To be all node use R/W option.

Thanks

[Aneesh Mohan]

Hi Nejad,

HP MC Serviceguard with Oracle RAC extension is the pre-requistie for making ORACLE Database instance active/active setup.

The difference here compared to our traditional MC Serviceguard Active/Passive setup is you can share a volume group among multiple configured nodes (master/slave modes) using serviceguard packages with SLVM functionality.

You can you the same commands to configure the cluster except in the package configuration.

Example :- If you are using two node Oracle RAC with MC serviceguard with volumes groups vg01 and vg02.

a) you should have MC Serviceguard RAC extension (OPS bundle) installed in both nodes.

b) Configure MC Serviceguard cluster with two nodes (traditional way).
c) Configure pkg1(node 1) with vg01 and vg02 and change the default volume group activation mode similar to VGCHANGE="vgchange -a s" in package cntl files.
d)) Configure pkg2(node 2) with vg01 and vg02 and change the default volume group activation mode similar to VGCHANGE="vgchange -a s" in package cntl files.

During the package activation both the volume groups will be activated correspoding nodes. (master/slave modes).Offcouse Oracle user should have proper permission to use the raw chunks created on the volume groups.

regards,
Aneesh

[Mousa55]

Hi Aneesh,

and thanks for your replay And the details clear.
My last question is
you said:
"You can you the same commands to configure the cluster except in the package configuration"

What is the difference between package configuration in Active/Passive and Active/Active. Is it the(VGCHANGE="vgchange -a s " only, and What are the exceptions?


thanks a lot

[Aneesh Mohan · Accepted solution]

>>> What is the difference between package configuration in Active/Passive and Active/Active. Is it the(VGCHANGE="vgchange -a s " only, and What are the exceptions?


that is what the minimum requirement of Oracle RAC for active/active setup from MC Serviceguard.

As stated above you have to grant required permissions to the oracle user to use the RAC volumegroups.

Additional things:-

You can add Oracle clusterware startup and shutdown also in the package scripts if you are using (above l0g release).

Package start up, we can add oraclware startup script (check with DBA )

Package shutdown script, we can add oraclware stop script then deactivate Shared VG.


Aneesh

[Michael Steele_2]

Hi

It's a little more complicated than what is being listed above. There a several parameter values to change in the package in cluding the prority of failing over.

cmmakepkg
package_type

For package_type this can be the type of failover, multi_node, or
system_multi_node. Default is failover. You can configure only failover or multi-node packages; see â Types of Package: Failover,
Multi-Node, System Multi-Nodeâ on page 277.
node_name The node on which this package can run, or a list of nodes in order of priority, or an asterisk (*) to indicate all nodes.

Read chapter six

[Stephen Doud]

SGeRAC does not allow for mounted file systems on multiple servers unless you also install a version of Storage Management Suite that contains Cluster File System.

The reason why you should NEVER mount a file system to multiple nodes without CFS is that hpux does not have an inter-node communication linkage that transmit inode/file modifications between nodes. Hence, if a file is modified, added or deleted on one node, the other node's in-core inode structure for that file system will not be updated, and becomes non-current. If the other node eventually attempts to use that file, it will discover the in-core file inode structures are not current, causing at the least a file system I/O error, and more likely a kernel panic! You don't want -that- in an HA enviroment.

Documents on Storage Management Suite are located here:
http://docs.hp.com/en/ha.html#HP%20Serviceguard%20Storage%20Management%20Suite

See also http://h20338.www2.hp.com/hpux11i/us/en/solutions/serviceguard-storage-management-suite-overview.html?jumpid=reg_R1002_USEN


When you are done with that, check out the Oracle implementation strategies for Oracle 10gR1 and R2 and 11gR1 discussed in various Whitepapers at this location:
http://docs.hp.com/en/ha.html#Serviceguard%20Extension%20for%20Real%20Application%20Cluster%20%28ServiceGuard%20OPS%20Edition%29