Since Exchange 2010’s release, database availability groups (DAGs) are the Microsoft way to provide highly-available Exchange services to your organization and mitigate data loss. They can present a challenge to some backup software as we recently found out. When using Veeam Backup & Replication version 9 to backup an Exchange DAG we saw errors such as
VSS: Backup job failed. Cannot notify writers about the 'BACKUP FINISH' event. A VSS critical writer has failed.
A server that is a member of a DAG can host active or passive copies of an Exchange database and you can have multiple databases covering your organization. In order to spread load and be able to tolerate the loss of a node hosting a database that is a member of a DAG, we chose to host an active database copy and a passive database copy on each node of a three-node cluster. In that way, we could lose a node system and still be able to provide Exchange services to all of our users without interruption. This has saved our bacon on several occasions when Windows or Exchange has gone out to lunch on a particular node. In general, DAGs just work.
Backup software that leverages the Volume Shadow Copy Service (VSS) can have problems backing up an active copy of a database. Veeam uses the MS Exchange VSS writer to perform backups when using their application-aware processing. This is a fantastic feature that allows for post-backup triggering of log truncation for example. However, these features fail when the underlying VSS operations fail on active databases.
So why not just backup the passive copy? That’s just the answer, but if you use Veeam, there is not an exposed option in the user interface to enable that behavior. To do so, you need to create a registry key on your Veeam Backup & Replication server:
HKEY_LOCAL_MACHINE\SOFTWARE\Veeam\Veeam Backup and Replication\EnableDBExclusions.
Create it as type
REG_DWORD, and set the value to
1. No service restart or system reboot is required. You should then have no errors when backing up passive copies of Exchange DAG databases.