Home / Azure

AzureHow-toIntermediate

Azure SQL Database high availability: service tiers, zone redundancy and what failovers look like

Understand the built-in HA architectures of General Purpose, Business Critical and Hyperscale, enable zone redundancy, and make applications handle the short reconfigurations that happen during failovers.

Deepak Bhardwaj2 min read

Applies toAzure SQL Database (vCore model), Azure SQL Managed Instance (similar concepts)

Built-in HA by tier

Tier HA architecture Typical failover
General Purpose Compute is stateless; data and log on remote premium storage. A new compute node attaches the files. Usually under a minute; cold cache afterwards
Business Critical Always On availability group with 4 replicas on local SSD (1 primary + 3 secondaries), one readable Seconds; warm replicas
Hyperscale Compute nodes + page servers + log service; optional HA secondary replicas Seconds with HA replicas; slower with 0 replicas

Zone redundancy

Spreads replicas (or compute + storage) across availability zones, so a zone outage doesn't take the database down.

az sql db update -g rg-sql -s sql-prod-cin -n Sales --zone-redundant true
az sql db show -g rg-sql -s sql-prod-cin -n Sales --query "{tier:sku.tier, zr:zoneRedundant}"
  • Supported for General Purpose (in many regions), Business Critical and Hyperscale (with HA replicas, set at creation in some cases).
  • Turning it on is an online operation, but schedule it in a quiet window.

Make the app resilient to failovers

Even with HA, failovers and planned maintenance cause brief connection drops (errors 40613, 40197, 40501, 49918, 4060 and similar transient errors).

  • Use a driver with connection resiliency (ConnectRetryCount, ConnectRetryInterval in ADO.NET).
  • Add retry with exponential backoff for transient errors around each operation, not just at connect.
  • Keep transactions short; long transactions are rolled back on failover.
  • Set MultiSubnetFailover=True in connection strings (helps redirection after failover with failover groups).
"Server=tcp:sql-prod-cin.database.windows.net,1433;Database=Sales;Authentication=Active Directory Managed Identity;Encrypt=True;Connect Timeout=30;ConnectRetryCount=3;ConnectRetryInterval=10;"

Maintenance window

Pick a maintenance window (for example weekday nights in your local time) so planned reconfigurations happen when traffic is low:

az sql db update -g rg-sql -s sql-prod-cin -n Sales --maint-config-id SQL_Centralindia_DB_1

(List available configurations with az maintenance public-configuration list.)

Read scale-out

Business Critical and Hyperscale offer readable secondaries. Route reporting with ApplicationIntent=ReadOnly to offload the primary.

Test

Trigger a user-initiated failover in a test environment and watch application errors:

az sql db failover -g rg-sql -s sql-test-cin -n Sales

Note

HA covers zone and node failures inside one region. A region outage needs geo-replication or failover groups, covered in the Azure SQL DR post.

Related fixes