IBM FlashSystem

 View Only

 PBR for Oracle Standard Edtion

Lukasz Borek's profile image
Lukasz Borek posted Mon October 20, 2025 06:57 AM

Hi,

My team is working on Oracle Standard Edition replication POC using Policy Based Replication. Main issue here is lack of crash-consistency, so database need to be switched to backup mode on active site and we need to transfer this snapshot to secondary.  How to do it smart using PBR?

I see two options : 

(1) Take snapshot on both primary and replica volume group : 

  • Move oracle to backup mode
  • Take snapshot on primary site
  • Wait until replication catch up
  • Take snapshot in secondary site

Main challenge here is how to make sure replication is completed on secondary site and we can take a snap? location2_running_recovery_point = 0? 

(2) Use snapshot, clones and refreshfromsnapshot

  • Move oracle to backup mode and take a snapshot
  • Create a clone based on snap above
  • Set up policy based replication for clone above
  • Annual backup schedule : create snapshot while Oracle is in backup mode and do refreshfromsnapshot 
  • Point in time snapshot will be replicated to secondary site. 

Looks like a plan but is complicated and failovers wont be as simple as in option 1.

Tried IBM Storage Defender Copy Data Management tool, but it can do only crash consistent snaps on secondary. 

Does it make any sense? Or do we have another option on the table?

Thanks! 

Claiton Zanella's profile image
Claiton Zanella IBM Champion posted Mon October 20, 2025 10:52 AM

Lukasz,

Usually, when you use replication with consistency groups, the storage system ensures that writes are replicated in order to maintain data consistency.
In the event of a disaster with asynchronous replication, it’s expected that some of the most recent data will be lost. Databases are typically designed to recover from that and return to a consistent state.

Any alternative involving point-in-time copies would result in a higher RPO and RTO. In such cases, you could use backup replication to achieve better recovery objectives.

Good luck!

Claiton

Lukasz Borek's profile image
Lukasz Borek posted Mon October 20, 2025 12:51 PM

Hi Claiton,

Databases are typically designed to recover from that and return to a consistent state

Its true for Oracle Enterprise Edition with Storage Snapshot Optimization. For Standard Edition you still need to put database into backup mode to ensure consistency. 

Backup replication is possible, but RTO is not acceptable. 

We did some basic testing using above scenarios and both were successful, but wondering if its correct way to do so. 

I'd prefer doing snapshot on both sites, during databse in backup mode. Don't know how to be sure if change was already replicated. I can think of scenario where secondary would be always few minutes behind primary and location2_running_recovery_point will never fall to 0. Calculating ${current time} - ${location2_running_recovery_point} sounds reasonable? 

Claiton Zanella's profile image
Claiton Zanella IBM Champion posted Mon October 20, 2025 03:19 PM

Even without SSO, the Oracle Database should be able to recover from a failure — after all, reboots are part of life!

I understand your point and I think first Option is an elegant way to ensure a guaranteed crash-consistency copy in case Oracle cannot recover by itself. It could also be triggered by an event in destination site. 

I don't have a lab available to test it, but I think you are in the right direction.

Nezih Boyacioglu's profile image
Nezih Boyacioglu IBM Champion posted Wed October 22, 2025 05:58 AM

Hi Lukasz, 
Storage replication by its nature provides crash-consistent solution. If you need application aware replication you need to use application layer replication solutions like data guard etc. (and yes it's on enterprise edition). If you have Oracle Enterprise edition it would also crash consistent with storage layer replication. For the recovering from crash consistent replication. You need to promote target site to readwrite mode and mount it to your server and open and if it's fails recover database on oracle.

Related Content