Choose autoscale default capacity for a missing-metric condition

Source facts

Microsoft explains that when autoscale has a problem reading the resource metric, it increases capacity to the configured default if the current count is lower. If the current count already exceeds that default, this behavior does not scale it inward. The default is therefore a specific fallback count. Microsoft Learn.

Applicability

Use this review when selecting or revising an autoscale setting’s default capacity. Identify the resource metric and the workload expected during a monitoring interruption. Keep the default’s missing-metric purpose distinct from the profile’s minimum, maximum and ordinary rule thresholds.

DSE recommendation

DSE recommends asking the service owner to justify the default against an agreed reduced-observability scenario. Record the expected behavior when current capacity is below the default and when it is above. Assess application readiness at that capacity using approved workload evidence, and document who handles the underlying telemetry failure.

Verification

Review the deployed default and compare it with the approved scenario. In a controlled exercise where appropriate, observe the setting’s behavior when the metric is unavailable and compare lower- and higher-capacity cases. Restore normal collection and confirm fresh evaluations. Retain capacity observations separately from application-response evidence; a matching instance count alone does not prove acceptable service behavior.

Official references

Microsoft Learn: Best practices for autoscale.