ExplainerInformationBusiness ContinuityIT

Apply zone balance before interpreting oldest or newest scale-in selection

Why can an OldestVM scale-in policy leave the globally oldest instance running?

Paired infrastructure paths converging on a stable recovered service.
DSE visual intelligenceContinuity & recoveryExplainer · 2 min read
Executive summary

What you need to know

Why can an OldestVM scale-in policy leave the globally oldest instance running?

Potentially affected

Azure scale sets using OldestVM or NewestVM scale-in policies, particularly zone-spanning deployments.

DSE recommendation

Predict scale-in candidates from zone balance and protection first, then apply the configured age rule within that eligible population.

Source facts

For a zone-spanning scale set, OldestVM and NewestVM selection follows zone balancing rather than a global age sort alone. When zones are uneven, the age rule operates within the imbalanced zones. Balancing deletes instances; it does not relocate them between zones or fault domains. Instance protection excludes a VM from scale-in regardless of the policy, but does not prevent an explicit user deletion. Microsoft Learn.

Applicability

Investigate an unexpected age-based selection using the distribution that existed immediately before the scale-in. Identify the policy and protection state of the candidate instances. Keep this question separate from deciding when autoscale should request a smaller population.

DSE recommendation

Predict scale-in candidates from zone balance and protection first, then apply the configured age rule within that eligible population. Have the application owner check whether important per-instance work is tied to an assumed global deletion order. Do not remove protection merely to make the observed order resemble a spreadsheet sorted by creation time.

Verification

On an approved disposable deployment, capture instance creation times, zones, and protection before a bounded scale-in. Compare the actual removal with the documented selection order. Repeat the reasoning after the population changes rather than reusing the old candidate list. If the result is inconsistent, retain the complete pre-operation inventory and operation evidence for investigation instead of deleting the expected instance manually to conceal the discrepancy.

Official references

Microsoft Learn: Scale-in policies. Source reviewed September 9, 2026.

Primary reference

Review the official source

Use scale-in policies with Azure Virtual Machine Scale Sets - Azure Virtual Machine Scale Sets | Microsoft Learn · Verified September 9, 2026

Open official reference ↗
Plan the next step

Need help applying this guidance safely?

DSE can help confirm applicability, protect service continuity, and validate the result across physical security and IT systems.

Talk with DSE