You experience poor synchronization performance between a primary replica and a secondary replica in a Microsoft SQL Server 2012 AlwaysOn availability group.
This issue can occur if the availability mode for a replica is changed from asynchronous commit to synchronous commit.
Service pack information for SQL Server 2012
To resolve this problem, obtain the latest service pack for SQL Server 2012. For more information, click the following article number to view the article in the Microsoft Knowledge Base:
2755533 How to obtain the latest service pack for SQL Server 2012
To work around this issue, suspend the secondary replica while the synchronous replication mode is used. Then, resume the secondary replica.
Note This workaround reestablishes the data movement session and causes the thread to read the replication mode flag again.
Microsoft has confirmed that this is a problem in the Microsoft products that are listed in the "Applies to" section.
To verify that you are experiencing the issue that is described in the "Symptoms" section, follow these steps.
Note When the session queries the sys.dm_exec_requests dynamic management view, the session performs DML against the availability databases. In this state, the prevalent reported waittype is HADR_SYNC_COMMIT.
Start Performance Monitor.
Track the following performance counters:
Availability Replica:Receives from Replica/Sec
Availability Replica:Sends to Replica/Sec
Compare the statistics that are reported by the performance counters in step 2. If the statistics are not similar, you may be experiencing the issue that is described in the "Symptoms" section.