DAGs build on WSFC
A Database Availability Group is an Exchange-level abstraction layered on a Windows Server Failover Cluster. Each Mailbox server that joins a DAG becomes a cluster node, and Exchange manages database replication and failover on top. You do not manage the cluster directly for normal operations — but you must understand quorum and the witness server, because a lost quorum takes every database offline.
For odd-numbered DAGs the built-in quorum model is fine; for two-node DAGs configure a witness server (a third Exchange server or a standalone Hub/CAS) to maintain quorum when one node fails.
Replication and activation
Each database has copies on multiple servers with an activation preference. Under a server failure, Exchange activates the lowest-preference healthy copy automatically. Use activation preferences to control which server owns a database in normal operation, and block automatic activation on the DR server so a site failover is deliberate.
- Deploy at least two mailbox copies per database
- Set activation preference to control primary ownership
- Block automatic activation on the DR copy (Datacenter Activation Coordination)
- Monitor copy status and replay queue length daily
Disaster recovery with DAC mode
Enable Datacenter Activation Coordination (DAC) mode on every DAG. Without it, a network partition between data centres can cause split-brain — both sites activating databases independently and diverging. DAC mode requires a quorum of DAG members plus the witness to be up before databases can mount, preventing the split-brain scenario.
Lessons learned
- Always enable DAC mode on multi-site DAGs
- Configure a witness server for two-node DAGs
- Block automatic activation on the DR copy
- Test a datacentre switchover annually and document the runbook
Written by
Jamshed Alam
Global MSP & IT Infrastructure Consultant with 20 years across Exchange, Microsoft 365, Azure, security and automation. Available for project and contract work worldwide.