Plan and Configure HA/DR
Built-in versus you-built
Azure SQL Database and Managed Instance are always highly available inside a region: the platform maintains multiple replicas and fails over automatically without user action. Zone redundancy spreads those replicas across availability zones within the region.
Cross-region protection is a design decision: active geo-replication (up to four readable secondaries, manual failover), auto-failover groups (grouped databases with read-write and read-only listeners plus an automatic failover policy and grace period), and geo-restore from a geo-redundant backup as a last resort.
Backup and retention
Point-in-time restore retention is 1–35 days (default 7). Long-term retention adds weekly, monthly, and yearly copies for up to 10 years. Backup storage redundancy is chosen at creation (LRS, ZRS, GRS, RA-GRS); geo-restore requires GRS or RA-GRS.
On VMs and Managed Instance
Managed Instance failover groups behave like the Azure SQL Database version. On SQL Server on Azure VMs you build Always On availability groups or a failover cluster instance, or use Azure Site Recovery; back up to URL into blob storage for off-box copies.
Traps
- The failover group's read-only listener already exists; you do not create a second group for read traffic.
- Automatic failover honors a grace period (default one hour); it is not instantaneous.
- Async geo-replication has a small but non-zero RPO; only synchronous-commit availability groups on VMs give zero data loss.
- PITR cannot reach past your retention setting; long-term retention is a separate restore path, not an extension of PITR.
Rule of thumb
Same region → rely on built-in HA and add zone redundancy. Different region → failover group. Compliance archive → long-term retention policy.