This post is about a new SQL Azure Database feature called a “Business Continuity Feature”, called “Disaster recovery/geo-replication”. This feature was announced last week as a preview. For the Premium tiers, this is a lovely feature that include “Active geo-replication” (their term) and cmdlets (and portal) for controlling it. For Basic and Standard tiers, you can use “Database copy + Manual export” according to the chart here: By now, you know what “Database copy + Manual export” means (BACPACs, CREATE DATABASE AS COPY OF).

Terms concerning geo-replication are documented here: and there’s an entire docs section devoted to how to use it. A few points need to be called out:
1. Secondaries are readable.
2. You can create secondaries in the same region/different server or different region.
3. The target server must have an available Premium database quota to create the active secondary. You start with a quota of 2 and can “request” (from Azure support?) to have your quota increased.
4. The secondaries are always transactionally consistent but can run behind the primary.
5. Secondaries must be the same performance level (P1,P2,P3) as the primary. This has charge repercussions, as you could imply (nowhere is it directly stated but it seems obvious) that each secondary costs the same as the primary.
6. Each primary and secondary consists of a database that is itself replicated (passively and unseen by you) 3 times, because this is the design of ASD. And there is no info that this internal design has changed.

You start a continuous copy to the secondary using the portal or the (soon-to-be-released) PowerShell cmdlet Start-AzureSqlDatabaseCopy. This starts a seeding process, and eventually catches up to close to the primary. You can monitor this. Your can stop or cancel a continuous copy by using portal, the DROP DATABASE command on the copy, or Stop-AzureSqlDatabaseCopy. You need to use Stop-AzureSqlDatabaseCopy -ForcedTermination parameter to cancel the operation while it is still in progress or stop the replication immediately. See

There are two ways to failover, planned or forced. Planned is performed only on the primary. Forced can be performed on the primary or secondary. Their terms for the types of types of disaster recovery that can be designed are:
1. Active-passive compute with coupled failover
2. Active-active compute with decoupled failover
3. Active-passive compute with decoupled failover

I know some folks in the disaster recovery biz have a problem with the terms active-active and active-passive, so perhaps these terms will change in future. There are a few recovery scenarios documented, I’d like to try these with the PowerShell cmdlets, when they appear.

This is an excellent feature for Premium edition but, to reiterate, nothing new in this space is currently announced for Basic and Standard, past the original, built-in and internal, 3 database replicas as with the original ASD. A change to the internal single-database design was not announced.

In the next post (which will be the last post, if I can make it short enough), I’ll talk about performance levels in the tiers and comment on the current state of ASD functionality.

Cheers, @bobbeauch