Failover Detection Utility – Availability Group Failover Analysis Made Easy

To analyze the root cause for the failover of an Availability group, users are required to perform coordinated analysis of various logs including the Cluster Logs, SQL Error Logs, and the Availability groups extended events logs. This coordinated analysis of the logs can be difficult and requires extensive knowledge of the internals and the various…

31

Behavior Change for VSS Backups on VMs with Basic Availability Groups

This week we released SQL Server 2016 SP2 CU2 and SQL Server 2017 CU9. In this blog, I’d like to go into a little more detail on one of the fixes that is part of these CUs which involves the SQL Writer Service. If you are running SQL Server 2016 (or later) Standard Edition with…

0

SQL Server 2016/2017: Availability group secondary replica redo model and performance

When availability group was initially released with SQL Server 2012, the transaction log redo was handled by a single redo thread for each database in an AG secondary replica. This redo model is also called as serial redo. In SQL Server 2016, the redo model was enhanced with multiple parallel redo worker threads per database…

0

New in SSMS – Always On Availability Group Latency Reports

With SQL Server 2012 we introduced Always On Availability Groups, and the Always On Availability Group Dashboard in SQL Server Management Studio (SSMS). This dashboard can be utilized by database administrators to view the current health of an Availability Group and its availability replicas and databases. While the dashboard can be configured to provide information…

14

Troubleshooting High HADR_SYNC_COMMIT wait type with Always On Availability Groups

HADR_SYNC_COMMIT indicates the time between when a transaction ready to commit in the primary replica, and all secondary synchronous-commit replicas have acknowledged the hardening of the transaction commit LSN in an AG. It means a transaction in the primary replica cannot be committed, until the primary replica received greater hardened LSNs from all secondary synchronous-commit…

0