search

MicrosoftDocs / azure-docs

Open source documentation of Microsoft Azure
https://docs.microsoft.com/azure
Creative Commons Attribution 4.0 International
10.24k stars 21.41k forks source link

Is there any latency to backup user-defined restore points in one azure datacenter and restore in remote Azure Datacenter #71596

Closed leizhang1984 closed 3 years ago

leizhang1984 commented 3 years ago

Hi,

I have one question about latency to backup user-defined restore points in one azure datacenter and restore in remote Azure datacenter.

Assuming I create a user-defined restore point in Azure Synapse SQL Pool in East Asia Region at 2021-03-04 5:00 p.m. (assuming restore operation will successfully immediately) and restore this point in remote Azure datacenter Synapse SQ Pool in Southeast Asia Region at 2021-03-05 5:20 p.m.

Is there any data replicate latency between East Asia region backup file and Southeast Asia Region restore file? Will Azure Synapse SQL Pool start to perform a full copy from East Asia Region to Southeast Asia Region at 2021-03-05 5:20 p.m.? If East Asia Region encounter entire Datacenter outage after 2021-03-05 5:20 p.m., will remote Azure Datacenter Southeast Asia Region restore successfully?

Thanks.

Document Details

⚠ Do not edit this section. It is required for docs.microsoft.com ➟ GitHub issue linking.

SamaraSoucy-MSFT commented 3 years ago

Thanks for the question! We will review and get back to you shortly.

SamaraSoucy-MSFT commented 3 years ago

The snapshot would begin copying at 5:20. It does take time to move data, how long depends on how much data you are transferring and current network conditions, so I'm unable to give you an exact number. We publish the average latency between regions, so that presents a minimum transfer time. Between East Asia and Southeast Asia the average latency in December 2020 was 34 milliseconds. An outage before the transfer completes would cause the restore to fail.

To allow for disaster recovery in the case of a full data center outage, Azure automatically makes a backup of your dedicated SQL pool every 24 hours. This is automatically turned on for your pool unless you expressly opt out. In your case, the region paired with East Asia is the Southeast Asia region, so the backup will be in the desired data center.

If you need to be able to do a restore with <=20 minutes of data loss across regions, you may want to look at Azure SQL as an alternative. Hyperscale tier in particular. There you can geo replicate the whole database to minimize possible data loss. Even without active georeplication of the full database, transaction backups are run every 5-10 minutes by default, preferably to georedundant storage, so the full backups could be located in both East Asia and Southeast Asia for much faster restore in the secondary region.

Please do let me know if that resolves your question.

leizhang1984 commented 3 years ago

@SamaraSoucy-MSFT, Thank you for your reply.

May I double check your reply?

When I restore user-defined restore point (I created in East Asia Region at 2021-03-04 5:00 p.m) in remote Azure Data Center Synapse SQL pool. After restored successful, the restored data is same version as 2021-03-04 5:00 p.m. , not 24 hour ago version, am I correct?

SamaraSoucy-MSFT commented 3 years ago

So there are two types of restores you can do- one that is the restore point that you were asking about. In that case it will work exactly as you've described- the copy starts when you begin the restore.

The second type is a full backup. If the data center goes down fully, the restore points may not be available and your restore point transfer would most likely fail, but the backup would be. The backup available for disaster recovery is only updated every 24 hours.

Does that clarify what I meant at all?

leizhang1984 commented 3 years ago

@SamaraSoucy-MSFT , Thank you for your clarify and please close this issue.

Appreciate your strong support.