Mirror view /sync setup for Oracle DB

When you mentioned Geo Cluster, I am assuming WAN link is in the picture and distance is large.

Mirrorview/S is not applicable for large distances, for large distances, you will need to use Mirrorview/A.

So if you decide to use Mirrorview/A then set up would be with Consistency Group.

Refer Diagram:

Refer attached Mirrorview knowledge book, Page 40

Something about, Geo Clustering (Googled):

WAN clustering, also called geoclustering, high-availability clustering or remote clustering, is the use of multiple redundant computing resources located in different geographical locations to form what appears to be a single highly-available system. The goal of WAN clustering is to support enterprise business continuity by providing location-independent load balancing and failover. WAN clustering can be used for just about any computing resource, including mainframes, file servers and software application stacks. The biggest challenge in WAN clustering is to make sure that system states and their associated data are concurrent at multiple locations. Two advancements have helped administrators meet this challenge -- faster wide area network (WAN connection speeds and the ability to create and manage a clustered resource through a single virtualized master identity.)

Hi Ankit,

Thanks for reply.. But the distance is not much large.. So it was set up using MV/S already for exisitng geo clusters.

This is new setup.. My point is whether it is MV/S or MV/A, for DB luns to be consistent during failover , we need to create a consistency group and place all the LUNs (DB+Archive) in that group.. Is this setup is correct?

Mirrorview/A is required for this site. If we are using M/V S during failover it may cause Data Loss or corruption, reason being it a database LUN IOPS will be constant.

Is this a banking sector site?

Not sure why there would be data loss/corruption using MV/S.  Since it is synchronous, all write IOs are sent to both arrays before acknowledgement to the host, so they are literally mirror images of each other.  At a very base level, when using Synchronous mirroring, you almost don't need a consistency group, but it's a good idea anyway.  If site A was lost unexpectedly, the failover to site B would look like a power outage as far as database consistency goes.  When Oracle restarts on the second array, it will perform transaction replay/rollback as necessary and come online.  If the failover is graceful, ie: administrator invoked, then Oracle will stop on one side, VCS will instruct MV/S to promote the target devices, then Oracle will start at the other side.  There would be no data loss in this scenario.

You are using VCS, so you are likely using VxVM/VxFS as well right? Be sure ALL LUNs that are part of the LVM group are in a single consistency group.  If a single database instance spans multiple LVM volumes, then all of the LUNs for all LVMs involved need to be in the same consistency group.  Other LUNs like for Archive Logs, Redo, etc should also be in the consistency group. 

This technically applies whether you are using MV/Sync or MV/Async.  The consistency group guarantees that every LUN in that group is write-order consistent with each other at the replication target.  This is required for ACID compliant databases (Oracle, MSSQL, etc) to recover from a failure.

Richard Anderson wrote: At a very base level, when using Synchronous mirroring, you almost don't need a consistency group, but it's a good idea anyway.

I would go as far as suggesting that consistency groups are mandatory even for MV/S.  However, Richard, I do get your points about "at a very base level" and "amost don't need".  While you provide an example where an entire site goes down, we need to consider individual SP A or B communication failures and to ensure a consistent fracture occurs if for instance one LUN is owned by SP A and another LUN is owned by SP B.

Christopher Imes wrote: Richard Anderson wrote: At a very base level, when using Synchronous mirroring, you almost don't need a consistency group, but it's a good idea anyway.  I would go as far as suggesting that consistency groups are mandatory even for MV/S.  However, Richard, I do get your points about "at a very base level" and "amost don't need".  While you provide an example where an entire site goes down, we need to consider individual SP A or B communication failures and to ensure a consistent fracture occurs if for instance one LUN is owned by SP A and another LUN is owned by SP B.

Ah.. yes.  Totally agree!

Thanks for the response.. But it is configured Keeping in mind during  SP failure also, consistency should be maintened.

It is well documented in EMC white paper... But the scenario now is:

It is VCS cluster across sites on Linux systems.. Accroding to VCS documentation, for MV/Sync consistency group is not supported...(For MV/aysnc it is supported).. So question raised to Symantec already ,will it take care of the consistency if individual mirrors are provided (instead of Consistency group) for VCS config during SP failures...

Please let me know  your ideas if you have come across this kind of setup already....

Hi Anderson,

Thanks for the clear explanation... I am surprised when i see the Symantec has not commented on it.. Symantec is implementing this new project setup... But it is setup by MV/Synchronous.

Anyway thanks for the knowledge share...