Host CPU overutilization may negatively impact vSAN performance
Article ID: 429468
Updated On:
Products
Issue/Introduction
CPU Ready values can be monitored via vSAN Performance Dashboard in vRealize/Aria Operations Manager.
Temporary or sustained increase of that metric seen on one (or more vSAN hosts) may manifest in increased READ/WRITE latencies for Virtual Machines.
Increase of CPU Ready times may also be accompanied by increase of 'Delayed IO Average Latency' values seen on vSAN Performance Graphs in the vSphere Web Client.
Environment
VMware vSAN 8.x, VMware vSAN 9.x
Cause
In a vSAN environment, CPU Ready should be ideally near 0% and below 1%. When vSAN modules are forced to wait for CPU (Ready state), storage latency is the direct result.
Resolution
vSAN modules run in ESXi kernel. Presence of CPU ready values over 1% (or between 0% and below 1% if sustained) does indicate heavy CPU oversubscription and can be caused by:
- High Consolidation Ratio
- Over-allocation (Oversizing)
- Noisy Neighbor VMs
Consider workload (Virtual Machine) rebalancing and review/adjust of workload characteristics (achievable vCPU allocation ratios/usage patterns) to decrease vSAN CPU Ready values.
In some cases disabling ESXi Power Management and/or Server Hardware Power Management features (as described in https://knowledge.broadcom.com/external/article/366987/) may help decrease CPU Ready values.
Note: this action may result in increased power consumption.
Additional Information
Additional Information:
Application on Virtual Machine runs slower than expected https://knowledge.broadcom.com/external/article/366987/virtual-machine-application-runs-slower.html
vSAN Performance Graphs in the vSphere Web Client Part 1 https://knowledge.broadcom.com/external/article/326953/vsan-performance-graphs-in-the-vsphere-w.html
Feedback
