Sudden Performance change with VM's?

[Nexen1]

I am running version 6.1.16 r140961 on a Windows 10 computer. I have 3 VM's running different OS. Running Windows 8.2 and Windows 10 in the Vm's. In the last few days the performance has dropped to where they are not usable. I can see the drive they are running on is now showing 100% activity constantly when the VM's are running. I had not changed settings. I scanned the disk and no issues were found.

I did review the forums and tried to change a few settings to see if this impacted performance. No change was noticeable with any of the changes.

Main changes were:
1) Turned on and then off the "Use Host I/O Cache
2) Enabled and Disabled 'Enable Nested VT-x/AMD-v
3) Set Hardware Virtualization Interface to 'Default' and 'Hyper-V' and then back again.

Really not sure what changed but it affects all of the VM's. Also tried just loading one VM and had the same issue with a single VM.

Just curious if anyone has experienced this and can offer any suggestions?

Robert

[mpack]

Let's clear up a bunch of misconceptions:

None of your tweaks are relevant, except perhaps the first which would give reduced disk performance if disabled, though that doesn't seem to be what you meant.

Nested Vt-x is not the old "Vt-x enable" setting - which no longer exists in VirtualBox 6.1.x since VT-x or AMD-v is now required for all VMs.

The "Hyper-v" paravirtualization setting enables or disables presentation of a Hyper-v compatible API to the guest OS. I.e. this a comms link between the guest and VirtualBox and has nothing to do with whether Hyper-v exists on the host. I.e. Mac and Linux host owners have this exact same VM option, and they certainly don't have a Hyper-v enabled host.

As to your problem.

You most likely got Hyper-v enabled in a recent host update. See third post in FAQ: I have a 64bit host, but can't install 64bit guests to turn it off again.

Finally, this is not a "Windows Guests" question, it's a Windows Hosts problem - I will move the topic.

[Nexen1]

Thank you for your help and for guiding me to the correct board.

I did confirm Hyper-V is not enabled on my computer. I also executed this command: bcdedit /set hypervisorlaunchtype off, and then did a cold start.

Performance remains the same.

Since it was previously working very well, it does seem like a change in the Microsoft O/S features may be the cause, but it does not seem to be Hyper-V in my case.

Any other advice is apprecated.

Robert

[mpack]

Please provide a VM log file. Make sure the VM is fully shut down, then right click it in the manager UI. Select "Show Log" and save "VBox.log" (no other file) to a zip file. Attach the zip here.

[Nexen1]

I have attached the log file.

[mpack]

Reduce to 2 cores. Your host has 4 cores in total, so it can't afford to assign 4 cores to a VM (p.s. hyper-threads don't count).

I suggest that you install the host extension pack too.

Also I notice that 3D acceleration is disabled, which will slow down the graphics, especially if you go for a large display window. The log btw says that this host doesn't support 3D graphics acceleration - is that correct? I.e. what is the graphics card?

Otherwise everything looks good. I don't see anything in the VM log to indicate poor performance. Can you be more explicit about what exactly is slow? I.e. is it graphics, hard disk , network? What does the guest task list show as being the main CPU hog (don't bother doing this until you fix the CPU over-committment).

[Nexen1]

The virtual disks are on the D: drive. When I launch the VM's, the D: drive goes to 100% activity and it take 15 - 20 minutes for the VM to start. Once started, while working in the VM everything takes 5 - 15 minutes. Also I am finding running MS updates is not working or so slow it never completes. D: drive remains at 100% activity constantly.

[Nexen1]

I have attached a screenshot of the Resource Monitor showing the disk activity. This is on the Host Computer

[mpack]

You should probably test the D drive for errors.

[Nexen1]

This is the resource monitor from the Guest Computer show 100% disk activity.

[Nexen1]

I did check the D: drive for errors previously. I right clicked the drive, selected Properties and then Tools. Under Tools I clicked the 'Check' button. No errors were found. Is there a better way to test this? I did this on the Host Computer.

[mpack]

If you have attended to my other tips then I'm expecting to see a new VM log file.

[Nexen1]

Attached is the revised log file. I added the Host Extensions and reduced to 2 processors. Also enabled 3D Acceleration. No positive impact and maybe slowed it down more...not easy to tell.

[fth0]

00:00:04.212609 AIOMgr: Endpoint for file 'D:\VirtualMachines\Win82SQL\Win82SQL.vdi' (flags 000c0723) created successfully
00:00:05.364577 VD: Opening the disk took 1155025759 ns

Opening the disk took over 1 second here. I'm accustomed to values in the higher microseconds (SSD) or lower milliseconds (HDD) ranges, much lower than 1 second in any case. Your screenshots show permanent disk activity, but only with low volume data transfers around 1 MB/s. This could be caused by some security software scanning the virtual hard disk, either from the host OS or the guest OS. It could also be caused by a old worn out hard disk.

In the host OS, without the VM running, how long does it take to make a copy of a large file (e.g. Win82SQL.vdi)? Take a stop watch for measurement and don't forget to tell us the file size.

00:00:12.102843 Display::i_handleDisplayResize: uScreenId=0 pvVRAM=000000000d5e0000 w=1024 h=768 bpp=24 cbLine=0xC00 flags=0x0 origin=0,0
00:01:48.402271 GIM: HyperV: Guest OS reported ID 0x1040603002580
[...]
00:02:09.412087 AIOMgr: Async flushes not supported
00:03:09.580536 VMMDev: Guest Log: VBoxSF: g_fHostFeatures=0x8000000f g_fSfFeatures=0x1 g_uSfLastFunction=29
[...]
00:03:23.056624 NAT: Guest address guess changed from 10.0.2.15 to 169.254.112.174 by gratuitous arp
00:06:55.215579 VMMDev: Guest Log: VBoxMP::vboxWddmPickResources: found the VBE card

Booting the guest OS took at least 7 minutes this time. Take a look at the pairs of successive log messages I quoted above: The majority of the 7 minutes lie inside those pairs, and in my own Windows VMs those time intervals are less than 5 seconds each.

If the first time gap is reproducible, start the VM, close the VM window after 30 seconds using the power off method, start it again, close the VM window after 60 seconds using the power off method, and post a zip file containing VBox.log and VBox.log.1. The statistics in the log files maybe tell us something about what happens in the second 30 seconds. If the first time gap is smaller, choose smaller time intervals. The goal is that the extra time of the second run only consists of the additional 'idle' time we're after.

[Nexen1]

I am suspecting you may be right about a possible HD issue. I am ordering a new SSD and will try moving them to the new drive to see if that resolves this issue. I will follow-up once I am able to test this.

Robert