Skip to content

Azure Local health and alert architecture

Azure Local health rollup

Health catalog

DomainSignalsDefault alert behavior
ComputeCluster Service, node membership, quorumAuto-resolving alerts
Compute performanceCPU, available memoryState only
StorageHealth Service faults, pool, volume/CSVAuto-resolving alerts
Storage diagnosticsIndividual physical-disk aggregateState only; Health Service owns root-cause paging
NetworkNetwork ATC intent convergenceAuto-resolving alert
Azure IntegrationRegistration/connection and local platform servicesAuto-resolving alerts
LifecycleFailed or attention-required solution updatesAuto-resolving warning/critical alert
MonitoringShared-probe completionAuto-resolving alert

Four Failover Clustering event rules cover node removal, clustered-role failure, CSV access failure, and CSV no-longer-accessible conditions. Event rules supplement current state; they do not replace state monitors.

Threshold policy

CPU, memory, and volume capacity defaults are provisional development values. A release threshold requires evidence for duration, recovery, host reserve, failover reserve, workload density, scale, maintenance, and false-positive rate.

In particular, 75 percent memory used is not a universal Azure Local failure threshold. Available host memory, host/guest split, paging, pressure, failover capacity, and trend must be considered.

DA propagation

Each node has six domain aggregates. Component dependencies consume the matching aggregate, then six service dependencies apply worst-state rollup at the DA root. A storage failure therefore degrades Storage without falsely making Network unhealthy.

Released under the MIT License.