Azure Local CPU Utilization

Question

Azure Local CPU Utilization

Handian Sudianto 7,026

I have 3 Azure Local node and one of the node gave different value for CPU Utilization between WAC and Failover Cluster.

In the WAC the cpu utilization for node2 is 6% but why in the Failover Cluster showing 50%?

User's image

0 comments

2 answers

Your answer

Answer 1

Manish Deshpande 5,760 Microsoft External Staff Moderator

Hello Handian,

Thank you for your detailed post and for sharing the screenshot from your 3-node Azure Local (Azure Stack HCI) cluster. I appreciate you highlighting the CPU utilization difference for node2 (JKT AZLOCAL.02): 6% in Windows Admin Center (WAC) Servers inventory versus 50% in Failover Cluster Manager. This is a common and expected observation in Azure Stack HCI / Hyper-V environments and does not indicate a performance problem, resource contention, or cluster issue.

WAC reports the parent partition (management OS / host) CPU usage only — typically low (~6% here) because most processing occurs under the Hyper-V hypervisor for guest VMs.
Failover Cluster Manager reports total physical node CPU utilization (hypervisor + all guest VM activity combined), which is why it shows the higher value (~50%).
This is by design: the two tools pull from different performance counters. WAC is the recommended and most accurate monitoring tool for Azure Stack HCI.
No action is required unless you are seeing actual high CPU alerts, slow VMs, or throttling. The 50% reading in Failover Cluster Manager simply includes the full workload on the node.

Steps to work on

Verify with Performance Monitor.
- On the affected node, open perfmon.msc.
- Add these counters:
- Hyper-V Hypervisor Logical Processor(*)% Total Run Time → This matches the total CPU view shown in Failover Cluster Manager (including guests).
- Processor(*)% Processor Time → This matches the host/parent partition view shown in WAC.
- Compare the two live charts — you will see the exact same 6% vs. 50% split.
Use WAC for ongoing monitoring (preferred method)

Stay in the Servers or Virtual machines dashboard in WAC.
Switch to the Summary tab for a clear host vs. guest breakdown of CPU and memory.
- This is the modern, Azure-integrated view designed specifically for Azure Stack HCI clusters.

Optional: Enable Azure Monitor for cluster-wide insights

Connect your cluster to Azure Monitor (if not already done).
- Review the built-in metrics Percentage CPU, Percentage CPU Host, and Percentage CPU Guest for a consolidated view across all nodes.

If you would like me to walk you through the PerfMon counters remotely or review any recent VM workloads on node2, just let me know.

Link - https://learn.microsoft.com/en-us/azure/azure-monitor/reference/supported-metrics/microsoft-azurestackhci-clusters-metrics

Thanks,
Manish.

Handian Sudianto 7,026 Reputation points

2026-04-02T05:48:29.5866667+00:00
Hi..

WAC reports the parent partition (management OS / host) CPU usage only — typically low (~6% here) because most processing occurs under the Hyper-V hypervisor for guest VMs. This mean WAC Monitor CPU usaged by Management?

This is by design: the two tools pull from different performance counters. WAC is the recommended and most accurate monitoring tool for Azure Stack HCI. If Failover Cluster Manager read all CPU utilization then why you say WAC is mosy accurate monitoring tools?

No action is required unless you are seeing actual high CPU alerts, slow VMs, or throttling. The 50% reading in Failover Cluster Manager simply includes the full workload on the node. Why 50% consider as full workload, isn't full workload is 100%?
Manish Deshpande 5,760 Reputation points Microsoft External Staff Moderator

2026-05-04T18:26:36.65+00:00
Hello Handian,

Thank you for your follow-up questions they are excellent and very common among Azure Stack HCI administrators. I'll address each one clearly.

1.Does WAC only monitor CPU usage by the Management OS (parent partition)?

Yes, exactly. Windows Admin Center (in the Servers inventory or Virtual Machines > Summary tab) primarily shows the parent partition / management OS CPU usage. This is the CPU consumed by the host operating system itself (Hyper-V role, clustering services, storage stack, etc.).

This value is usually low (like your 6%) because in a well-designed Hyper-V/Azure Stack HCI environment, most of the workload runs inside the guest VMs, and the hypervisor schedules that work directly. The management OS itself does very little heavy lifting.

2.Why is WAC considered the most accurate/recommended monitoring tool when Failover Cluster Manager shows the "full" CPU?

Both tools are accurate — they just show different things:

WAC → Shows Host/Management OS CPU (what you should watch for host health, overhead, and stability).

Failover Cluster Manager → Shows Total physical node CPU (Host + Hypervisor + all Guest VMs combined).

WAC is the recommended tool for Azure Stack HCI because:

It is purpose-built for HCI with modern Azure-integrated views.

It provides a clear breakdown of Host vs Guest CPU/Memory in the Virtual Machines > Summary tab.

Microsoft recommends using WAC (or Azure Monitor) as the primary monitoring interface for Azure Stack HCI clusters.

Failover Cluster Manager is a general-purpose cluster tool and does not give you the Host vs Guest separation that HCI admins need.

3.Why is 50% considered "full workload"? Isn't full workload 100%?

Great question. The 50% you see in Failover Cluster Manager is not "half full" in the traditional sense.

100% in the total node view (Hyper-V Hypervisor Logical Processor) means all physical CPU cores on that server are fully utilized (Host + all VMs).

50% means the node is comfortably loaded but still has headroom. In production HCI clusters, it is very common and healthy to run at 40–70% total CPU during normal business hours.

Microsoft guidance on thresholds (using the correct counter):

< 60% = Healthy

60–89% = Monitor / Caution

90–100% = Critical (risk of performance impact)

So your 50% is perfectly normal and not a cause for concern.

Thanks,
Manish.
Manish Deshpande 5,760 Reputation points Microsoft External Staff Moderator

2026-05-06T02:20:20.7266667+00:00

Hello Handian

I wanted to check if my last response made sense. I’d be glad to assist further or explain anything in more detail and please accept as Yes and upvote if the answer is helpful so that it can help others in the community.

Answer 2

WAC and Failover Cluster Manager are showing different CPU values because they use different metrics and sampling/aggregation methods, not because the node is actually behaving inconsistently.

From the Azure Local monitoring perspective, there are multiple CPU-related metrics, for example:

Percentage CPU / Percentage CPU Host / Percentage CPU Guest – overall host vs. guest CPU time on the node.
VM Percentage CPU – CPU used by each VM relative to the host’s processors.

These metrics can be aggregated differently (Maximum, Average, Sum, Count) and over different time windows and dimensions (Cluster, Node, VM, LUN). For example, the same node can show:

A low average CPU utilization over the last several minutes in one tool (like WAC’s inventory view), while
Another tool (Failover Cluster Manager) shows a near instantaneous or short‑window utilization, or a different aggregation such as maximum.

Because of this, a node can legitimately show around 6% in WAC (for example, an averaged or smoothed value) and about 50% at the same moment in Failover Cluster Manager (for example, a point‑in‑time or differently aggregated host CPU metric).

To validate what is actually happening on the node, use the Azure Local performance dashboards or Azure Monitor Metrics and look at the Percentage CPU and Percentage CPU Host/Guest metrics for that specific node over the same time range and with the same aggregation (for example, Average over the last 5 minutes). This gives a consistent view across tools.

References:

Share via

Azure Local CPU Utilization

2 answers

Your answer