Troubleshooting Mindset Series #020: When VM Storage Is Slow — How Do You Trace the Bottleneck?

A VM is experiencing slow performance, but CPU, memory and storage capacity all appear normal.

At this point, checking only the datastore free space may not tell us where the actual problem is.

The more useful question is:

Where in the storage path is the bottleneck?

Troubleshooting Approach

Instead of jumping directly to the HCI storage layer, trace the path step by step:

VM/Application → Guest OS → Virtual Disk → HCI Storage → Disk/SSD → Storage Network/Path

1. Start from the VM/Application

First identify what is actually slow.

• Application response?
  
  
• File access?
  
  
• Database operation?
  
  
• VM boot?
  
  
• General VM responsiveness?
  
  

This helps define the symptom before checking the infrastructure.

2. Check the Guest OS

Look for evidence of:

• High disk latency
  
  
• I/O wait
  
  
• Disk queue
  
  
• Unusual disk activity
  
  

The objective is to confirm whether the guest OS is actually waiting on storage.

3. Check the Virtual Disk

Compare the affected VM with a healthy VM where possible.

Look at:

• Disk latency
  
  
• IOPS
  
  
• Throughput
  
  
• I/O behaviour
  
  

A healthy comparison can help identify whether the issue is VM-specific or infrastructure-related.

4. Check HCI Storage

If multiple VMs show similar behaviour, move the investigation toward the HCI storage layer.

Check for:

• Storage latency
  
  
• I/O pressure
  
  
• Resource contention
  
  
• Storage-related alerts
  
  
• Background storage activities
  
  

5. Check Disk / SSD

If the issue points further down the stack, check the physical storage layer.

Look for:

• Disk/SSD health
  
  
• Abnormal latency
  
  
• Degraded components
  
  
• Rebuild or recovery activities
  
  

6. Check the Storage Network / Path

Don't forget the path between the compute and storage components.

Check:

• Network errors
  
  
• Link issues
  
  
• Packet loss
  
  
• Path availability
  
  
• Abnormal network latency
  
  

Common Mistake

A common mistake is to see:

“Storage usage is only 50%, so storage cannot be the problem.”

Free capacity and storage performance are two different things.

A system can have plenty of free space while still experiencing high latency, I/O pressure or resource contention.

Troubleshooting Mindset

When a VM is slow, don't immediately ask:

“How much storage space is left?”

Ask:

“Where is the I/O becoming slow?”

Then trace the complete path:

VM → Guest OS → Virtual Disk → HCI Storage → Disk/SSD → Storage Network/Path

This gives us a structured way to narrow down the bottleneck instead of changing multiple things at once.

Key Takeaway

Capacity tells you how much storage is available.
Performance tells you how well the storage is working.

When troubleshooting slow VM storage, follow the evidence through the entire I/O path.

Community Contribution

A few ideas from the community have been incorporated into this troubleshooting flow. Thanks to everyone who contributed to the discussion on TMS#019 posts. Thanks to everyone who shared their troubleshooting perspectives, especially the suggestion to trace the full path from the VM/application layer down to the storage network. Your feedback helped shape this follow-up topic.

The core idea you're highlighting is absolutely right: storage performance problems rarely live in one place, and capacity alone is almost never the cause. What matters is latency, queueing, and contention somewhere along the I/O path.

Great insights! The layered approach you explained for I/O slowness is very logical. Many of us usually only check CPU/RAM, but your point about tracing the issue from Hypervisor down to Physical Disk/SSD and Network is the right troubleshooting mindset. The common mistake about storage usage is really valuable. Thanks for sharing!

Thanks to share!

Thanks for sharing this excellent troubleshooting mindset for slow VM storage!

This is really helpful. Many engineers only check free space, but your point about tracing the full chain - from VM to Guest OS to HCI Storage to physical disks - is the correct approach to find the real bottleneck. The example about 50% free space still having performance issues is very practical and true.

Great guide for the community, especially for beginners learning HCI troubleshooting.