Replica: Restore points on continuous schedule

[brichardson1991_obk]

Hi all,

We currently are replicating a live database (finance) server every 2/3 hours.

We are looking to move this to continuous or every half hour replication.

The question is though, how do we maintain enough restore points that we are covered should something bad happen to the live server it doesn't matter due to a restore point on the replica.. The idea is that we run off the replica while we restore from backup or fail the changes from the replica to the live.

[foggy]

Ben, the number of restore points for replication jobs is limited by 28. Isn't that enough to promptly react with a failover to replica VM in your case?

[brichardson1991_obk]

Thanks Alexander.

I will have to research more but that certainly helps.

We are just trying to guarantee we have enough of a window that if someone didn't notice an issue for a couple of hours with the database we could fail over to replica and get what we needed.

Based on current metrics we would get 8 hours worth of restore points. But if it is running continuously this will be less as the amount of change will be less so the job will take less time surely ?

[skrause]

You will need to take into account the amount of time for snapshot creation/removal and if you are using VM hot-add in VMware, the reconfiguration of the proxies. In most of my replication jobs those tasks are a good 75% of the time the job takes to run. You will also need to factor in the potential performance impact on your database of running off of a snapshot more often.

What type of database are you running (MSSQL, Oracle, etc).

[foggy]

Based on current metrics we would get 8 hours worth of restore points. But if it is running continuously this will be less as the amount of change will be less so the job will take less time surely ?

Theoretically, yes, however you should perform some tests to see how this works in your particular environment.

[brichardson1991_obk]

What would happen if we ran a continous job with only a few restore points.
But also a scheduled (say 3 hourly) job for the same server.

How would this work (if possible) this would give us the oversight that if something went wrong further back than the continuous allowed for we could fix it?

[foggy]

Each replication job can have its own 28 restore points. This will work, however sometimes one job will wait for another to release the VM.