Server 2016 on 6.7 very laggy

[Sheridan]

I've got 3 VMs on a 6.7 ESXI host, 2 are 2016 and one is 2012

 

All were migrated from physical using the VMWare convertor and all work OK - but one is incredibly laggy to use - opening any window/command prompt takes 10-15 seconds. Its got 8 cores and 64GB RAM (which is more than the original physical server had) and the performance monitoring shows its barely ticking over, with CPU peaking at 2-5% and RAM never hitting 10%.

 

I don't think its a resource issue - but I'm not sure what would make it so laggy to use? Accessing file shares on the server is fine and no noticeable difference to any of the other VMs, but opening the DHCP console for example is painfully slow!

[RobD]

Does the VM have 8 cores and 64GB of RAM or the host?

[Sheridan]

Does the VM have 8 cores and 64GB of RAM or the host?

The VM itself does, the host has 24 cores and 196GB ram, the other two servers are only 4 core/32GB basic servers

[mavhc]

What VM integration are you using, are you using virtualised hardware drivers, or emulation?

[Jcx500]

Is your Raid array healthy? , we had this on hyper-v when a VM Array had a failing Disk

[RobD]

Whats it running as thats pretty beefy, if you oversize your VMs then there can be serious performance issues.

 

The hypervisor scheduler will have to wait for 8 physical cores to be free, I know it sound odd but try and shrink the big VMs and see if that improves performance.

[jtotheb]

The hypervisor scheduler will have to wait for 8 physical cores to be free, I know it sound odd but try and shrink the big VMs and see if that improves performance.

 

Yes, CPU Ready Time should be the counter to keep an eye on in VMware to spot CPU over-subscription. It seems counter-intuitive to reduce vCPU count to actually get more use out of the physical CPU until you see it in action.

[Sheridan]

Its the same spec as the older physical server, hence the 8 cores etc. It runs multiple roles but nothing heavy duty (DNS,NPS,DHCP and a few applications, plus lots of file server shares)

 

RAID array on the host seems healthy, all decent 10k drives with no errors. The drivers are virtualised - basically as was created using the Convertor application

 

I have noticed a slight improvement changing the 'Visual effects' to 'best performance' but its still laggy to open windows etc

[newpersn]

I would make sure you have installed the vm tools from the host. Make sure its on the same release as the host.

[Sheridan]

The VM Tools version is showing as compliant, as was installed from the host.

 

I've reduced the cpus to 4 cores now, just to see if that has any impact, but so far its about the same, obviously its not getting used as much during the holidays, but logging in sees a blank black screen for a good few minutes before the desktop appears

[bknaggs]

Is it any better/different if you RDP rather than connect through the VM console?

[Sheridan]

I tend to always use the rdp anyway, but there’s not much difference between them

[kmount]

Did you clean up removing old ghost devices from device manager after p2v? (show hidden non present devices true)

 

I've seen weird things happen when a vm thinks it's a hpe physical server when it still had hardware devices/drivers in place post migration.

 

Other thing to check is the disk controller in the vm settings, and whether you're seeing any weird traces in procmon when you feel the hang.

[RobD]

Maybe build one from new and see what the speed is like.

 

Also test the disk performance with iometer:

https://www.itprotoday.com/mobile-management-and-security/using-iometer-measure-disk-performance

 

Could test the network too with jperf:

https://turbofuture.com/computers/How-to-Measure-Network-Throughput-Using-JPerf

Edited August 5, 2020 by RobD