change management
1 TopicEnabling Zone Redundancy on an Existing Azure SQL Database, Without Surprises
In the portal, zone redundancy is a single Yes/No setting. In production, there's a bit more to plan for. Microsoft describes enabling it as an online operation. That's true, but online still means a short disconnect, and your application has to handle it. This post is the checklist I'd go through before making the change on a database people actually depend on. What you're actually getting With zone redundancy on, your database uses availability zones to replicate data across multiple physical locations within a single Azure region. If one datacenter in the region goes down, your database stays up. It doesn't protect you if the whole region goes down. For that you need geo-replication or failover groups. Step 1: Check that you can enable it Not every combination of tier and region supports it: Service tier Where it's supported Premium All regions with availability zones Business Critical All regions with availability zones General Purpose Selected regions with availability zones Hyperscale All AZ regions, but only at creation time (see below) Check the current regional list right before you plan the change, especially for General Purpose. Regional support changes over time. Hyperscale is different. Zone redundancy support can only be specified during database creation and can't be modified once the resource is provisioned. To move an existing Hyperscale database, you'll need to transfer the data with database copy, point-in-time restore, or geo-replica, and then point your connection strings at the new database. Plan that as a separate migration, not a settings change. Elastic pools: enabling zone redundancy support for elastic pools makes all databases within the pool zone redundant. Make sure everyone who owns a database in that pool knows about the change. Step 2: Test your retry logic before the change You can't configure this part in Azure. For Premium, Business Critical and General Purpose, enabling zone redundancy is an online operation with a brief disconnect. Microsoft's guidance is that if you have implemented retry logic for standard transient errors, you won't notice the failover. The catch is the "if." Before the change, confirm that: Your apps retry transient connection errors with bounded exponential backoff, not endless instant retries Long-running transactions and batch jobs can be restarted safely Connection pools recover on their own, without anyone restarting the app A quick way to test is to run a small client against a disposable database that you scale or update while it's connected: $maxAttempts = 5 for ($i = 1; $i -le $maxAttempts; $i++) { try { Invoke-Sqlcmd -ServerInstance "<server>.database.windows.net" -Database "<db>" ` -Query "SELECT GETUTCDATE()" -AccessToken $token -ErrorAction Stop Write-Host "Attempt $i succeeded"; break } catch { $delay = [math]::Min([math]::Pow(2, $i), 30) Write-Host "Attempt $i failed, retrying in $delay s" Start-Sleep -Seconds $delay } } If your real application wouldn't get through that kind of interruption, fix that first. Step 3: Plan the window Even though it's online: Pick a quieter period. Busy workloads and long-running transactions make the disconnect more noticeable. Don't promise exact timings. How long the operation takes depends on database size and regional conditions. Tell stakeholders to expect a brief disconnect, not "X seconds." Change one thing at a time. If you use geo-replication, plan the primary and the secondary as separate, deliberate steps, and don't combine this change with a tier change or other maintenance. Tell people. Let app owners and on-call staff know the window, even when you expect no visible impact. Step 4: Make the change In the portal, go to Compute + Storage, set Would you like to make this database zone redundant? to Yes, and select Apply. Or script it: Set-AzSqlDatabase -ResourceGroupName "<rg>" -ServerName "<server>" -DatabaseName "<db>" -ZoneRedundant az sql db update --resource-group "<rg>" --server "<server>" --name "<db>" --zone-redundant While it runs, keep watch on the operation status in the portal and on your own application's connection-error logs. Step 5: Verify it worked Don't treat the completion notification as proof. Check the property directly: az sql db show --resource-group "<rg>" --server "<server>" --name "<db>" --query zoneRedundant Then confirm from the application side that error rates are back to normal and that no jobs failed quietly during the change. Step 6: Have a rollback plan You can turn zone redundancy off again with the same setting or --zone-redundant false. Write that down in the change record anyway, along with who decides to roll back and on what criteria. A simple change-record template Item Notes Tier and region supported? Verified on [date] Retry logic tested? Test result / link Change window Date, time, duration estimate Stakeholders notified Who and when Verification steps Property check + app health Rollback trigger and owner Criteria and name The short version Turning on zone redundancy is easy. Being ready for it is the real work: confirm support, test your retries, pick a sensible window and verify the result. Do that, and most of the time nobody will notice the change. Reference: Enable zone redundancy for Azure SQL Database - Azure SQL Database | Microsoft Learn