The July 2026 Azure Monitor enhancement introduces advanced platform metrics as a paid, opt-in preview for selected providers, with Azure Storage among the supported services. For storage accounts, the immediate operational value is container-level blob count and container-level used size, which strengthens tenant-level capacity governance and chargeback. It does not replace standard Storage metrics for performance diagnosis; instead, it augments them.
Architecture & Deep Dive.
Use a three-layer observability model: advanced metrics for granular capacity, standard platform metrics for transaction/latency/throttling, and resource logs for request-level forensics. Standard Blob metrics still expose SuccessE2ELatency, SuccessServerLatency, and Transactions with useful dimensions such as ApiName, GeoType, Authentication, Tier, and ResponseType. The response-type taxonomy is especially important for distinguishing server busy states from explicit account throttling.
Step-by-Step Implementation & Scripts.
Register Microsoft.Insights, enable the advanced metrics rule on the storage account, query metrics by dimension, and route Blob resource logs into Log Analytics.
az provider register -n Microsoft.Insights
az storage advanced-platform-metric create \
-g rg-observability \
–account-name stappmetricsprod \
–enabled \
–rule-config-filter-type AllContainersFilter
RESOURCE_ID=$(az storage account show -g rg-observability -n stappmetricsprod –query id -o tsv)
az monitor metrics list \
–resource $RESOURCE_ID \
–metric SuccessE2ELatency \
–dimension ApiName
StorageBlobLogs
| where TimeGenerated > ago(30m)
| summarize count(), p95Duration=percentile(DurationMs, 95)
by OperationName, MetricResponseType, AuthenticationType, StatusCode
| order by p95Duration desc
Troubleshooting Guide & Common Edge Cases.
If advanced metrics show no data, first verify provider registration and rule scope. If end-to-end latency is high but server latency is low, the issue is likely client/network-side. If both are high and response types indicate ServerBusyError, ServerTimeoutError, or account throttling, investigate service saturation, hot paths, or scalability boundaries. If you still depend on classic Storage Analytics logs, remember that those logs can appear with up to roughly one hour of delay and are therefore not suitable as your primary real-time signal.
Best Practices.
Enable advanced metrics selectively, keep capacity and performance alerting separate, build dimension-aware alerts, and use resource logs for root-cause validation rather than relying on summary charts alone.