Infographic shows customer-triggered VM recovery between availability zones using ZRS storage; recovery is not automatic.
Microsoft has put zone resilience for Azure VMs into public preview, and the fine print matters more than the headline. The feature lets you start a failover of an eligible zonal Azure VM to another availability zone in the same region. It isn't automatic failover, and Microsoft says it isn't ready for production.

Infographic shows customer-triggered VM recovery between availability zones using ZRS storage; recovery is not automatic. What Microsoft announced​

The Azure Updates entry carries a November 19, 2026 timestamp. Microsoft's Azure Compute Blog announced the preview on November 17. The idea is simple. When a zone fails, you recover the VM you already have. You don't build a replacement and restore disk copies onto it.

The VM retains its resource identity and supported configuration. Microsoft's documentation lists the resource ID, VM ID, NIC and IP configuration, and supported ZRS managed disks as carried over. Only the zone placement changes.

Production warning. Microsoft says to use the preview only for evaluation and testing with nonproduction VMs, and not for production VMs. Any plan to put this in front of a real workload has to wait for general availability.

Not automatic, not seamless​

This is a customer-driven recovery path. It is not high availability. Microsoft's documented workflow runs like this:

  1. Confirm the VM is affected, and pick a healthy zone in the same region that supports your VM size.
  2. Force-deallocate the VM, which stops the workload.
  3. Update the zone placement of the existing VM.
  4. Start the VM in the target zone, subject to best-effort capacity.
  5. Verify the VM and the workload before restoring traffic.

What to expect during a failover:

  • Downtime: The workload is unavailable while the VM is deallocated and restarted. Active requests, in-memory data and runtime state are not preserved.
  • No recovery-time guarantee: Microsoft says there is no guaranteed recovery time and that failover is best-effort.
  • Capacity risk: The move can fail if the target zone lacks capacity for the VM size. Microsoft's advice is to pick another supported zone.
  • You do the detecting: The feature recovers the VM itself. It doesn't recover other resources, orchestrate application-level recovery, or reroute traffic. Azure Resource Health and Service Health can help you spot problems, but Microsoft doesn't automatically notify you when a zone is down.
  • You do the failback: After recovery, you are responsible for any failback to the original zone once it is healthy.

What the near-zero RPO covers​

ZRS managed disks replicate data synchronously across zones. That gives a near-zero recovery point objective, but only for data committed to supported ZRS disks. It does not cover active requests, in-memory data, temporary disks, or application state that hasn't been written to those disks. Verify application consistency after recovery.

Eligibility checklist​

Required:

  • A standalone VM pinned to a specific availability zone.
  • Premium SSD ZRS or Standard SSD ZRS for the OS disk and every attached data disk.
  • Static or dynamic private IPs. Any public IP must be a zone-redundant Standard SKU, and any load balancer or gateway must be a supported zone-redundant one.
  • A region where the required ZRS disks are available.

Not supported:

  • VM Scale Sets, availability-set VMs and Spot VMs.
  • LRS disks, ephemeral OS disks, Premium SSD v2, Ultra Disks, Standard HDD, and data disks using write acceleration.
  • On-demand capacity reservations, dedicated hosts and proximity placement groups.
  • Azure Site Recovery on VMs that have zone resilience enabled.

If your VM uses LRS disks, Microsoft's documented route is to create ZRS disks from snapshots. A zone LRS disk's redundancy can't be changed in place. You then swap the OS disk and reattach the data disks, which means downtime.

Enabling it: tools and steps​

Enablement has to happen while the source zone is healthy. That means before the outage, not during it.

Supported tooling: the Azure portal, Azure CLI 2.87.0 or later, Azure PowerShell (Az 16.3.0 or later with Az.Compute 11.9.0 or later), or Compute REST API 2026-03-01. Bicep examples use the same API version.

New VM, in the portal:

  1. On the Basics tab, pick a region that supports zones, set Availability options to Availability zone, and choose a specific zone.
  2. Select "Protect VM from zonal-failure with Zone resilience (Preview)", then enable it in the pane and save.
  3. On the Disks tab, confirm every disk is Premium SSD ZRS or Standard SSD ZRS.
  4. Finish with Review + create.

Existing VM, in the portal:

  1. Stop the VM and wait for Stopped (deallocated). This interrupts the workload.
  2. Under Availability + scaling, open the Zone Resilience (Preview) setting, enable it and save.
  3. Start the VM and confirm the setting shows Enabled.

In the CLI, the new-VM path uses a zone movement flag on az vm create. An existing VM is deallocated, updated with the same flag, then started. The failover itself can be a single az vm update with the target zone. Per Microsoft, that command force-deallocates, changes the zone and restarts the VM. PowerShell has no single orchestrating cmdlet, so you run the stop, update and start steps in order. Microsoft notes that sequence isn't transactional, so check VM state if the update fails.

Verification and troubleshooting​

Microsoft's post-failover checklist is worth copying into a runbook:

  • The power state is Running.
  • The VM reports the target zone.
  • The expected OS and data disks are attached.
  • NIC, subnet, NSG, load balancer and public IP associations are correct.
  • Extensions are healthy.
  • OS and application health checks pass, and workload data consistency checks pass.

Only then should you restore traffic.

Documented errors include OperationNotAllowedForNonZRSDisks (a disk isn't ZRS) and ZoneMovementUpdateRequiresDeallocatedVM (you tried to toggle the setting on a running VM). Another is ForceDeallocateVMNotAllowedForNonResilientVM. That one means force deallocation was requested on a VM without the feature enabled. If its source zone is already down, Microsoft says to contact Azure support.

How it compares with other options​

ApproachModelDisk requirement
Zone resilience (preview)Moves the existing VM to another zone and reuses its disksZRS OS and data disks
Azure Site Recovery zone-to-zoneReplicates ahead of time and creates a recovery VM in a different resource group, with failbackSupports eligible LRS disks, not ZRS
Multi-VM, multi-zone deploymentRuns several instances and distributes trafficDepends on the application

For true high availability, Microsoft still points to multiple instances across zones with traffic distribution. It also says zone resilience is aimed at recovering an eligible standalone zonal VM. Site Recovery remains the answer for regional disaster recovery, and the two can't be combined on the same VM.

Cost​

Microsoft says there is no feature charge for zone resilience during the preview. Standard VM, ZRS disk and networking charges still apply. It will be a paid feature at general availability, with pricing to be announced then. ZRS disks cost more than LRS, so budget for that as well.

Analysis: who should care​

This looks aimed at legacy-style workloads that can't scale out. In the announcement, SAP RISE names SAP Sybase IQ as a workload without native cloud high-availability capabilities. That is a vendor-quoted use case, not proof the feature is production-ready. The preview's nonproduction restriction still applies.

The hard parts are the prerequisites. Every disk must be ZRS, existing VMs need a deallocation window to enable the feature, and capacity in the target zone isn't reserved. Microsoft also says the feature is meant for disaster recovery, not for keeping an application up. It is a sensible tool for single-instance machines, provided the runbook is written, tested and owned by someone.

Practical next steps:

  • Inventory zonal VMs and check their disk SKUs.
  • Test in a nonproduction subscription, using planned moves between healthy zones. Microsoft says this checks the recovery workflow but doesn't reproduce a real outage.
  • Check that your VM size is offered in more than one zone in your region.
  • Write down who decides to fail over, and who restores traffic.
 

References

  1. (In preview) Public Preview: Zone resilience for Azure VMs Azure Updates 2026-11-19T20:34:19Z
  2. Zone resilience for Azure VMs (Preview) - Azure Virtual Machines learn.microsoft.com
  3. Zone resilience for Azure VMs | Microsoft Community Hub techcommunity.microsoft.com