What Microsoft says changed
Microsoft's documentation says the update improves two things for Azure-to-Azure disaster recovery of VMs in VMSS Flex:
- Target configuration.
- End-to-end recovery workflows, including failover, reprotection, and failback.
Microsoft documents two details that affect how you set this up:
- Site Recovery protects each VM in the scale set as its own replicated item. It does not protect the scale set as a single object.
- You must create the target scale set in the target resource group before you enable replication.
Microsoft's page doesn't give a changelog, an agent or portal version, a region list, or step-by-step configuration instructions. Treat any claim of that kind with suspicion. The wording also doesn't say this is the first time VMSS Flex VMs could be protected. It describes an enhancement, not an introduction.
Limits to know before you plan around it
Availability versus scalability. Microsoft's Azure-to-Azure support matrix lists VMSS Flex as supported for availability scenarios and not supported for scalability scenarios. The two pages read as complementary. The new update improves target configuration and recovery operations. The matrix still says scalability scenarios are unsupported. Don't read "generally available" as "the scale set will scale itself in the recovery region."
Scope. The documented scenario is Azure-to-Azure. The update doesn't establish anything new for on-premises-to-Azure replication.
General Site Recovery caveats. These are not new restrictions from this update, but they matter when you build a recovery plan for scale-set VMs:
- The support matrix says Azure RBAC policies on VMs are not replicated to the failover VM.
- It also says VM extensions are not replicated and have to be installed manually after failover.
The three workflows, briefly
These are standard disaster-recovery definitions. Microsoft's update confirms all three are covered but doesn't detail the conditions for each.
- Failover brings workloads up in the recovery region, either in a drill or in a real disruption.
- Reprotection sets up replication in the opposite direction after failover. Microsoft's general reprotection guidance says that after a failover, VMs boot in the secondary region in an unprotected state. It also says reprotection seeds the original site with the latest data. The reprotection page doesn't mention VMSS Flex specifically.
- Failback returns workloads to the original region once you're ready.
Microsoft's general Azure-to-Azure tutorials note that placement settings depend on target availability. They say these settings might need manual reassociation after failover or reprotection. The tutorials don't single out scale sets. Microsoft's update doesn't say whether this enhancement changes that behavior for VMSS Flex.
A practical checklist
- Create the target scale set first. Do this in the target resource group before you enable replication. This is the one explicit prerequisite Microsoft gives.
- Expect per-VM replicated items. Monitor and manage each VM individually, not the scale set as one unit.
- Check the scalability limit. If your design depends on the scale set growing during recovery, the support matrix says that isn't a supported scenario.
- Run a test failover. Use a drill to see how the target scale set behaves in your environment. Microsoft's update doesn't promise that every scale-set property carries over. It also doesn't promise application consistency or target-region capacity.
- Plan for extensions and RBAC. Reinstall extensions and reapply role assignments manually, as the support matrix describes.
- Rehearse failback. Run reprotection and failback in a drill too, since those are the stages the update specifically says it improves.
Why this matters
VMSS Flex is Azure's more flexible scale-set model. Microsoft's own wording shows that disaster recovery for it has had rough edges. Those edges include target configuration and the full cycle of recovering, reprotecting, and returning. Closing those gaps is useful, but this is an incremental change. It improves the operational workflow and doesn't remove the scalability exclusion. Teams running scale-set workloads should compare their current runbooks against the new documentation. They should also confirm in a non-production drill that the target scale set and its VMs behave as expected.
Microsoft's October 2026 feature-updates list also includes Azure Site Recovery for Azure Local, which is now generally available. It adds support for Azure Arc-enabled Windows VMware and physical servers as well.