FIX: A split-brain scenario occurs after a failover when using AlwaysOn Availability Groups with external cluster technology in SQL Server 2017

Applies to: SQL Server 2017 DeveloperSQL Server 2017 Enterprise CoreSQL Server 2017 Enterprise More

Symptoms


Assume that you use AlwaysOn Availability Groups on an instance of Microsoft SQL Server 2017 that's managed by an external cluster technology for high availability and disaster recovery (for example, Pacemaker on Linux.) When the external cluster fails over the primary server to another server that's running SQL Server, the original primary server doesn't quit as expected. Therefore, a split-brain scenario occurs.

Resolution


This problem is fixed in the following cumulative update for SQL Server:

       Cumulative Update 9 for SQL Server 2017

Status


Microsoft has confirmed that this is a problem in the Microsoft products that are listed in the "Applies to" section.

References


Learn about the terminology that Microsoft uses to describe software updates.