Blog Post

Running SAP Applications on the Microsoft Platform
9 MIN READ

Moving your SAP workload from regional to zonal: a migration path with Azure's migration API

dennispadia's avatar
dennispadia
Icon for Microsoft rankMicrosoft
Nov 17, 2026

High availability and resilience are essential for SAP systems that support mission-critical business processes. For years, a common Azure deployment pattern has combined availability sets with Proximity Placement Groups (PPGs) to keep an SAP system’s components close together. This helps protect against localized hardware failures, but it has limits: availability sets don’t protect against a datacenter-wide or availability-zone outage. Their placement constraints can also make it harder to change VM sizes or use newer capabilities such as on-demand capacity reservations.

Azure Availability Zones improves resiliency by allowing redundant SAP components to run across physically separate zones within a region. For existing deployments, the new availability-set-to-scale-set and regional-to-zonal VM migration APIs help existing SAP landscapes adopt this architecture through in-place migration, rather than rebuilding VMs in a parallel environment. The right path depends on how your VMs are deployed (single instance or in availability sets), and migration requires planned VM downtime. Where the configuration is supported, you can retain the VM and disk IDs, avoiding changes to associated SAP licensing or backup configuration. This blog post explores the limitations of regional deployments, the recommended zonal architecture, and how to plan a controlled migration with scheduled VM downtime.

Important

Availability-set-to-scale-set migration and in-place regional-to-zonal single-VM migration are in preview. During the preview, we don't advise customers to use the feature for production workload. In case, you want to use this procedure to migrate production workload, reach out to your Microsoft account team.

Where your SAP deployments stand today?

Two common deployment patterns illustrate why organizations are considering this transition.

Scenario 1: Availability sets with a proximity placement group

Here, every tier of a single SAP system - application servers, central services (ASCS/ERS), and the database is deployed in its own availability set. Because availability sets can otherwise be spread across the whole region, a PPG is layered on top to pin everything close together and keep inter-tier latency low.

This works, but it comes with trade-offs:

  • Capacity constraint: Pinning every VM into one tight physical location (the PPG) makes it harder for Azure to find room for all your VM sizes together, especially as you scale up or add capacity.
  • Compute/storage colocation: Storage has to keep pace with compute that's deliberately clustered in one place.
  • No capacity reservations: VMs deployed in an availability set are not eligible for on-demand capacity reservations. So, you have less of a safety net for guaranteed capacity.

Scenario 2: Availability zone + Availability set + Proximity placement group

In this pattern, the database is zonal, while central services and the application tier share a PPG. The application tier’s availability set provides fault-domain spreading within the same zone.

It's a clever workaround, but it still carries most of the same baggage: PPG-driven capacity constraints, compute/storage colocation, and no capacity reservations for the availability-set VMs. You've gained zonal protection for the database, but the rest of the architecture is still built around the PPG construct.

Target architecture: zonal deployment with a Flexible Scale Sets

For SAP workloads, Microsoft recommends Virtual Machine Scale Sets with Flexible orchestration and a platform fault domain count of 1 (FD=1).

In this model, the scale set serves as a deployment framework for the SAP system. It does not mean that SAP becomes an automatically scaling application.

A representative highly available architecture place:

  • Primary and secondary database VMs in different Availability Zones.
  • SAP ASCS and ERS VMs across zones, following the supported HA design.
  • Application servers across the selected zones, with sufficient capacity to support the intended failure scenarios.

Within each zone, the Flexible scale set spreads VMs across fault domains on a best-effort basis. Despite the setting's name, FD=1 does not mean that every VM is deliberately placed in the same physical fault domain.

For SAP, use Flexible orchestration without a scaling profile, and do not combine this deployment with a PPG. The SAP scale-set deployment guide provides the configuration requirements.

Moving away from PPG-constrained placement gives Azure a broader allocation scope within each selected zone. This can improve flexibility when resizing VMs or adopting new VM families. It does not eliminate quota or capacity constraints.

It also changes performance considerations. Zonal placement is not a substitute for latency validation. Test application-to-database communication, replication, storage access, and representative business transactions against your SAP requirements.

Most importantly, placing a VM in a zone does not make the application zone resilient. Resilience still depends on redundant SAP components, replication, clustering, fencing, networking, and enough surviving capacity.

Then the question most customers ask isn't "should we move to zonal scale sets", it's "how do we get our existing SAP systems there without rebuilding it."

From rebuilding resources to migrating in place

Moving an existing SAP system to Availability Zones is not a new requirement. But the way you get there has evolved.

Previously, a regional to zonal migration typically involved creating zonal disks from snapshots of the existing regional disks, then recreating the VM in the target zone using those disks. The process could be orchestrated using the PowerShell tooling described in How to easily migrate an existing SAP system VMs to Flexible Virtual Machine Scale Set.

While this approach preserved the SAP installation and operating-system configuration, it still required rebuilding the underlying Azure resources. Recreating a VM changes its unique VM identifier, and creating replacement disks introduces new disk resource IDs. For SAP teams, those changes could mean generating and applying a new SAP license, reviewing backup protection, and restoring identity-dependent access. If the original VM used a system-assigned managed identity, the replacement VM needed a new identity and the corresponding permissions reassigned.

In other words, the migration involved more than changing placement: it also required coordinating the operational consequences of resource recreation.

The migration APIs shift this process from resource recreation to in-place migration. For supported configurations, the existing VM and managed disks move to zonal placement while retaining their identifiers and the VM’s system-assigned managed identity. This avoids the SAP relicensing and identity-reconfiguration work triggered by recreating the VM, reducing the number of changes teams need to coordinate.

The distinction is important: you are changing where the VM runs, rather than rebuilding the VM to run somewhere else.

In-place migration still requires VM downtime, an SAP-aware migration sequence, and post-migration validation.

Two starting points, two migration paths

The appropriate workflow depends on how the VM is deployed today.

Current deployment

Migration approach

Target

VM in an availability set

Availability-set migration workflow, followed by per-VM migration with a target zone

Zonal VM in a Flexible scale set

Regional single-instance VM, outside an availability set

Deallocate the VM, update its zone assignment, and restart it

Standalone zonal VM, optionally attached to a Flexible scale set

A single-zone target can be useful for non-production or non-HA systems. For protection against a zone outage, however, the SAP architecture must distribute redundant components across zones.

How the migration works

For SAP, per-VM migration control lets you coordinate infrastructure changes with database replication, Central Services clustering, and application-server draining. Validate each migrated VM and restore the required redundancy before proceeding to the next step.

Path 1: Move availability-set VMs into a zonal Flexible scale set

The availability set migration workflow separates preparation from the disruptive per-VM operation. This allows you to validate the target configuration before entering the migration window.

  1. Prepare the target scale set: Create or choose a target VM scale set in the same resource group as the source resources. For SAP, configure the scale set as below -

    1. Flexible orchestration and FD=1.
    2. No scaling profile.
    3. The intended Availability Zones.
    4. SinglePlacementGroup set to false.

      # Example
      az vmss create --resource-group "<resource-group>" --name "<scale-set-name>" --location "<region>" --orchestration-mode Flexible --zones {1,2,3} --platform-fault-domain-count 1 --single-placement-group false
  1. Validate each availability set (no downtime): Run validateMigrationToVirtualMachineScaleSet for every availability set that participates in the SAP system.
  2. Start migration mode (no downtime): Run startMigrationToVirtualMachineScaleSet for each validated availability set.
  3. Migrate each VM to its planned zone (downtime): Invoke migrateToVirtualMachineScaleSet for an individual VM, supplying its target zone.
  4. Start and verify the migrated VM: After successful migration, the VM remains stopped and deallocated. Start it explicitly, then verify its zone assignment, scale-set membership, operating-system health, and SAP role before continuing.
  5. Clean up: Delete the empty availability sets only after all VMs have migrated and the system has been verified.

See Migrate availability sets to Virtual Machine Scale Sets for feature registration, limitations, and instruction for migrations.

Note

Choose the interface carefully: The portal migrates all VMs in an availability set together. Use CLI, PowerShell, or REST when you need per-VM control. That control enables staged execution but does not by itself guarantee uninterrupted SAP availability.

Path 2: Move a regional single-instance VM into a zone

For an eligible regional VM that is not in an availability set, the in-place migration workflow is simpler:

  1. Deallocate the VM you want to move.
  2. Update the zone property on the VM, from null to your target zone. You can remove PPG association during zone assignment.
  3. Optional: attach to a VM scale set in the same resource group, if you want the scale-set construct rather than a standalone zonal VM.
  4. Start the VM and validate the workload.

The VM can start after zone assignment completes while data migration continues transparently in the background.

If you attach the VM to a Flexible scale set, ensure the scale set includes the VM's assigned zone and meets the attachment requirements. For the SAP target architecture, retain FD=1 and no scaling profile.

See Migrate a regional virtual machine to an availability zone for feature registration, limitations, PPG handling, and attachment instructions.

Important

Do not confuse this with the Resource Mover workflow. The separately documented regional-to-zonal move experience creates a target VM and new disks, and does not retain the source VM's ARM resource ID. Its operational implications are different.

What to check before you migrate

Use the latest guidance in, Migrate availability sets to Virtual Machine Scale Sets, and Migrate a regional virtual machine to an availability zone. Before you begin, consider a few important planning points:

  • Preview status: At the time of writing, the migration APIs are in preview and shouldn't be run against production SAP systems without validating current terms and support conditions.
  • Basic SKUs aren't supported: VMs behind a Basic public IP or a Basic Load Balancer need to be upgraded to Standard SKU first.
  • Downtime is required: Each VM is deallocated as part of its migration step, so plan a maintenance window.
  • Same resource group requirement: The target scale set must live in the same resource group as the source VM or availability set.
  • Disks convert in the background: Managed disks convert from regional to zonal, keeping the same disk ID. The VM is usable immediately after migration; the underlying data transfer completes transparently afterward.

Putting it together: migrating an SAP system end to end

Consider a system with three availability sets: one for application servers, one for ASCS/ERS, and one for the database pair.

The target is one Flexible scale set with FD=1, spanning two zones. Database HA partners are separated across zones, Central Services components follow the supported cross-zone HA design, and application servers are distributed according to capacity and latency requirements.

Create the target scale set, validate all three availability sets, and place them into migration mode. Then execute the individual VM moves using an SAP-aware runbook.

For HA components, the runbook should account for replication health, controlled failover, cluster maintenance, fencing, and restoration of redundancy. For application servers, drain sessions or work as appropriate and confirm sufficient remaining capacity.

There is no universal safe migration order for every SAP system. The sequence depends on the database, operating system, cluster design, and acceptable service interruption.

Planning other VM changes?

If you are already scheduling downtime for the migration, consider whether to coordinate other planned updates in the same maintenance window as enabling Trusted Launch for existing eligible VMs, updating to a newer VM size, converting to SCSI to NVMe. These changes have their own prerequisites and support requirements, so validate the target VM and disk configuration for your SAP workload before proceeding. They may require separate steps. The migration API does not necessarily perform them as part of the migration.

References

Updated Nov 13, 2026
Version 1.0
No CommentsBe the first to comment