topic SAN Replication Configuration Help for Oracle Database (11gR2) in HPE 3PAR StoreServ Storage

SAN Replication Configuration Help for Oracle Database (11gR2)

roydba — Mon, 19 Jun 2017 05:35:40 GMT

Hi Experts,

We are using HP 3PAR storage for Oracle RAC database and HP remote copy for replicating data to DR. Our storage is capacity is full. We have another 3PAR box and presented LUNs from that box to database server.

Question: Can I add all LUNs from Box-1 and Box-2 under same Remote Copy Group.

Requirement is, If Box-2 link failed, Box-1 should stop replicating and vice versa, otherwise there will be difference between Box-1 and Box-2 LUNs. Oracle requires all LUNs should stop replication at same time.

Would anyone please advise me a solution.

Thanks in advance.

Re: SAN Replication Configuration Help for Oracle Database (11gR2)

aberus — Fri, 11 Aug 2017 12:31:35 GMT

You may be in a pickle my friend because your Oracle LUN's are pulling storage from different physical arrays. In order for Oracle recovery to work you need to have consistency between all of the LUN's which is what the RC groups provides. About the only thing that I could think of, and this would require testing, would be to place the DB into hot backup mode kick off both of the RC groups then take the DB out of hot backup mode. Taking this approach you will at least know that your Oracle headers are in a "frozen" state.

I've created a script that does all of this (3PAR Users Group), but my situation is a little difference because I am running Oracle on Windows so I am not able to use Recover manager.

Re: SAN Replication Configuration Help for Oracle Database (11gR2)

Sheldon Smith — Sun, 13 Aug 2017 19:31:37 GMT

A Remote Copy Group (RCG) is tied to a target pair. You have two target pairs: Box-1 <> DR, and Box-2 <> DR, so you would have two RCGs: Box-1_DR and Box-2_DR.

HPE 3PAR Recovery Manager Central for Oracle requires all volumes be on one system. So, what Aberus said.
Here's an idea, not sure if this will do what you need or not, it would require testing;
Using Asynchronous Periodic replication, you set both RCG periods to zero, so they require manual (scripted) resynchronization. Then after you put the database into Hot(?) Backup mode (so you are working with consistent database volumes):

Create consistent group snapshots of both groups of DR (secondary) volumes with "creategroupsv"
Manually synchronize both RCGs with "syncrcopy"
Put database back to Normal mode.
After both RCGs complete synchronization, delete the DR snapshots from step 1.

In case of failure, you stop both groups, fail them over to the DR system, and promote the pre-synchronization snapshots to the now primary volumes.