Home » Server Options » RAC & Failsafe » Extended (Distance / Stretched) RAC Cluster with Existing Setup. (12.1.0.2)
 Extended (Distance / Stretched) RAC Cluster with Existing Setup. [message #647135] |
Wed, 20 January 2016 04:28  |
naiduva
Messages: 4
Registered: November 2008
Location: India - Chennai
|
Junior Member |
|
|
We currently have a existing RAC setup as below:
Datacenter 1 - Node 1 & HP 3PAR SAN 1
Datacenter 2 - Node 2 & HP 3PAR SAN 2
Oracle RAC 12c Release 1
Both Node 1 and Node 2, now access the ASM data on SAN 1. SAN 2 is currently ideal. Now the requirement is to replicate the SANs so that incase one DC goes down, the data is not lost and application is up and available.
I want to implement the Oracle Extended Stretched Cluster with this setup. As SAN 2 is ideal, is it possible to attach to the current network and start replicating data from the ASM on SAN1 to ASM on SAN2 and make it as a single ASM.
The current ASM structure on SAN1 is as follows:
ACFS 61428 MB EXTERNAL REDUNDANCY
ARCH 1048576 MB EXTERNAL REDUNDANCY
DATA01 1048576 MB EXTERNAL REDUNDANCY
OCRVOTE 1048576 MB HIGH REDUNDANCY
REDO 1048576 MB EXTERNAL REDUNDANCY
Now if I get SAN2 into the network, how do I start replicating the data from SAN1 to SAN2 using ASM. How do I create the diskgroups? Is this possible at all. Further how do I add the Voting/Quorom disk in 3rd Site. Can I use NFS mounted FS?
Request any help with this matter.
Regards,
Vasant
|
|
|
|
|
|
| Re: Extended (Distance / Stretched) RAC Cluster with Existing Setup. [message #647141 is a reply to message #647136] |
Wed, 20 January 2016 07:34  |
 |
Mahesh Rajendran
Messages: 10708
Registered: March 2002
Location: oracleDocoVille
|
Senior Member
Account Moderator |
|
|
First let us first breakdown to bare bone requirements and make the terminologies clear.
1) RAC.
Provides protection from a single point failure. Multiple nodes are involved.
If one node fails, others will still serve (availability).
If the central storage fails, RAC is of no use. Dataguard might be helpful.
RAC alone is not a DR Solution.
RAC on Extended clusters provide better availability. Still not a DR.
Dataguard is a DR solution.
2) ASM.
Just a volume/storage manager, specifically built for Oracle.
3) SAN replication is not an ideal DR solution.
If a block gets corrupted, it is replicated to DR site.
>>As SAN 2 is ideal
I have no idea on what you mean by ideal here.
Idle, may be?
RAC on extended cluster (Geographically separated) might look good on paper.
How can you ensure that this 50KM of network is always available, fast and secure?
If you can do that, We can talk further.
>>start replicating data from the ASM on SAN1 to ASM on SAN2 and make it as a single ASM.
You contradict yourself in the same statement.
Either you have two separate ASM on different SAN
or
A single ASM using 2 different SAN.
With this you ask ASM to consider SANS as disks.
Use normal redundancy, With SAN1 as failgroup1 and SAN2 as failgroup2.
But, you still have the issue of Geographically separated SANS.
I would just look into Dataguard, where I have much control over everything,
Even if the Datacenters are separated.
[Updated on: Wed, 20 January 2016 07:35] Report message to a moderator |
|
|
|
| Re: Extended (Distance / Stretched) RAC Cluster with Existing Setup. [message #647179 is a reply to message #647141] |
Thu, 21 January 2016 05:06 |
naiduva
Messages: 4
Registered: November 2008
Location: India - Chennai
|
Junior Member |
|
|
Hi Mahesh,
Thank you for your reply.
Yes its idle 
My customer already has the setup up and running in both the data centers and wants to attache the idle SAN to the existing setup and make it as a stretched cluster using ORACLE RAC. They already have the plan to build a DR site with stand alone database in a different site/DC. So I do have to go with this build without a choice as they wont even accept a SAN to SAN replication.
So I assume that I just attach the SAN2 to the existing setup, present its disk and create the new diskgroups on both SAN1 and SAN2. Both the DCs are actually within 50 Kms by road. So I think this setup should be quite feasible.
Now the question of actually configuring the ASM on both the SANs is I actually require. So suppose I create the following diskgroups on SAN1:
2 x 1TB LUN for DATA - 1 disk of 1TB is from SAN1 and another disk of 1TB is from SAN2)
4 x 10GB LUN for REDO - 2 disk of 2.5GB each is from SAN1 and 2 disk of each is from SAN2
2 x 500GB LUN for ARCH - 1 disk of 250GB from SAN1 and another 250GB from SAN2
3 x 30GB LUN for OCRVOTE - 1 disk of 10GB from SAN1 , 1 disk of 10GB from SAN2 and the 3rd 10GB disk from a NFS mounted filesystem on a 3rd site
Is my above understanding correct? Does this architecture take into account the failgroups. OR could you give me a better plan?
Regards,
Vasant
[Updated on: Thu, 21 January 2016 05:59] Report message to a moderator |
|
|
|
| Re: Extended (Distance / Stretched) RAC Cluster with Existing Setup. [message #647182 is a reply to message #647179] |
Thu, 21 January 2016 06:14 |
|
Mahesh Rajendran
Messages: 10708
Registered: March 2002
Location: oracleDocoVille
|
Senior Member
Account Moderator |
|
|
Sounds good.
NFS for OCR should work.
Essentially, you mirror everything ( since you are using normal redundancy).
Something like this
create diskgroup data normal redundancy
failgroup controller1 disk
'/SAN1/disk1' NAME SAN1disk1
failgroup controller2 disk
'/SAN2/disk1' NAME SAN2disk1
;
And yes.
The customer is always right. Hope they have direct/dedicated channels to start with.
Even with that, be prepared to get more calls (with latency issues) down the road.
Good for you (I mean, more business)
[Updated on: Thu, 21 January 2016 06:21] Report message to a moderator |
|
|
|
Goto Forum:
Current Time: Sat May 03 03:39:15 CDT 2025
|
] 
Blog
Search
Help
Members
Register
Login
Home
