General availability of enhanced Geo-Disaster recovery with read replicas

Microsoft

Apr 16, 2024

We're thrilled to announce the General availability (GA) of our enhanced disaster recovery (DR) features for Azure Database for PostgreSQL - Flexible Server. Following a successful public preview period, where we gathered valuable feedback and made significant improvements, these features are now fully available to help you bolster your data resilience across multiple regions.

Our enhanced disaster recovery capabilities center on two major innovations: Virtual endpoints and the Promote to primary server operation. These features are designed to simplify your DR processes and ensure robust data protection:

Virtual endpoints: Regardless of the server role swaps, the writer endpoint always connects to the current primary server, and the reader endpoint points to the current replica. Virtual endpoints ensure you don't need to change connection strings in your applications, keeping things straightforward and hassle-free.

Promote to primary server: This feature enables a quick and seamless transition between your primary database and its replica, minimizing downtime and maintaining continuity during unexpected regional disruptions.

To locate this feature, simply navigate to the replication blade in the Azure portal. Here, you can manage these settings and familiarize yourself with the seamless disaster recovery options available.

Screenshot of replication blade in the Azure portal.

Disaster recovery simplified

Our enhancements make disaster recovery straightforward. They support hassle-free, manual switchovers or failovers to a read replica, whether in the same region or a different one, and ensure a smooth transition back to the primary server. This ease of use is crucial for conducting regular disaster recovery drills, enabling quick and effective responses with minimal footprint on your applications.

An accompanying diagram illustrates the system configuration before a promotion, showing how virtual endpoints are linked to the current primary and replica server. After the promote to primary server operation, the server roles switch, and the virtual endpoints automatically update to reflect the new roles.

Diagram of promote to primary server flow.

What about the Previous Promote Feature?

We also updated the terminology for existing functionalities for better clarity:

Promote to independent server and remove from replication: This action, previously known as the 'promote' option, is renamed to better distinguish it from the 'Promote to primary server' capability, which allows switching between primary and replica server roles within a replication setup. This action in Flexible Server is similar to the 'stop replication' feature in Azure Database for PostgreSQL Single Server, which is set to retire.

You can use virtual endpoints with the 'Promote to independent server and remove from replication' action, but it's optional. The accompanying diagram illustrates how server roles change and how virtual endpoints adjust following the promotion.

Diagram of promote to independent server flow.

Data Integrity During Promotion

You can promote a read replica in two ways: planned or forced. Both methods are available whether you choose to 'promote to primary server' or 'promote to independent server and remove from replication.'

Planned: Ensures full data synchronization before accepting client connections, safeguarding against data loss.
Forced: Suitable for urgent recovery needs, such as during a regional outage, prioritizing quick restoration without data synchronization.

Changes from public preview to general availability

As we transitioned from public preview to general availability, we implemented significant enhancements to improve functionality and security:

Server symmetry checks for forced promotions: We introduced new checks for the 'forced' promote option when promoting a replica to a primary server. These checks ensure that the primary and replica servers meet server symmetry requirements. This improvement help maintain the integrity of the replication and avoids the need to recreate the old primary when failing back. If a replica doesn't meet these symmetry requirements, the promotion attempt results in an error. However, in scenarios where a region is down, the replica is promoted without any checks to ensure swift recovery.

Private link support: We expanded our networking options to include Private link support. This addition complements the existing networking capabilities, which include public access and private access via virtual network integration. Now, users can choose from three networking options to best suit their security and connectivity needs.