Linux host freezes with Workstation 10.0.6 and Windows 7 guest

[Kukulkano]

Hello,

I am now quite desperate as this is really stopping me from doing my work.

I deleted Windows and I installed a fresh Kubuntu 14.04.02 (LTS) with 64 bit as my Desktop. I also installed VMWare Workstation 10.0.6 to run my development machines. I'm now having serious problems:

Every now and then, my complete host is freezing if I run a virtualized Windows 7 on my Workstation. The mouse is only moving very slow and all windows are reacting very(!) slow. My CPU monitor showing all 4 CPU on 100%. If I manage to open up the system monitor (or top on console), I can see that vmware-vmx and kwin are burning most of my CPU resources. If I wait between 10 seconds and a minute, the system becomes usable again. Now CPU is going down to normal. But as soon as I start moving a Window around (no matter which), the thing is happening again.

Symptoms in short:

• HOST: Linux Kubuntu 14.04.02 (x64) on Intel i5 CPU with 8GB RAM and running on SSD (ext4).
• GUEST: Several Windows 7 (x32) running several software. 1 CPU and 1GB RAM assigned. Problem is on several guest machines.
• Every now and then, all my 4 CPU taking nearly 100%. The host CPU is used by vmware-vmx and kwin process (used top). Sometimes after 15 minutes, sometimes directly after turning on the Guest. No idea about a reason.
• Host Mouse, Windows and OS is operating very very slow.
• During this time, the virtual Windows 7 machine is idle!!! Energy-Saving or Screensaver are all off!
• If I manage to run guests Taskmanager, I can not see some action (sometimes TaskManager itself using 60% CPU!). Looks for me like the host is going down, not the guest!
• My host is running a SSD drive (256 GB).
• The VM logfile does not show something during this phase.
• vmware.log attached.

What I already tried:

• I deactivated "Accelerate 3D" for all my VM's.
• I re-installed VMWare Tools on the guest machines.
• I run VMWare workstation using vblank_mode=0 /usr/bin/vmware
• I re-created the machine with less hardware as possible and reloaded the .vmd.
• I converted the virtual disk image from previous format to current format (VMWare asked me the step before, I said 'yes').
• I inserted a separate Network Card (NIC) and turned off the onboard network adapter in BIOS.
• I turned off the onboard graphics in BIOS and added a NVIDIA GeForce 8400 PCIx16 card (using binary NVIDIA drivers offered by Kubuntu).
• I removed VM's CD-ROM, Audio, Floppy and any other unused hardware.
• I tried to use cpulimit to throttle vmware-vmx and kwin on my host, but no success. System stays unusable in that situation.

No success at all.

Is there anything else I can do to bring back my environment into a productive state? You can imagine that this always happens in the worst situations 😞

Any ideas?

Thank you!

[Kukulkano]

Update! In many cases the problems start directly if I try to move another Window on my Host. I believe there is some interaction with KDE Window manager.

[Kukulkano]

Is there a reason for not getting a single answer or suggestion? Is my question to long or to much detail?

[matthewls]

!

I don't know, but will make some guesses.

First, I'd install the latest nvidia driver 352.30 from nvidia's site. NVIDIA DRIVERS Linux x64 (AMD64/EM64T) Display Driver

You note that the host is the problem may well be true. I found kubuntu to be unstable and switched to mint kde 17.2 which is much better (17.2 is in RC now, but the official release will be available soon).

I saw similar hangups and it turned out to be problems with an older nvidia driver.

A few suggestions.

For video issues with nvidia: close workstation, open a terminal and enter

     export __GL_YIELD="USLEEP"

then

     vmware

Another workaround in a terminal:

     echo never > /sys/kernel/mm/transparent_hugepage/enabled

     echo never > /sys/kernel/mm/transparent_hugepage/defrag

I believe this has to do with an interaction between kde and some  linux kernels. You could try installing the latest stable kernel (4.1.3).

You mentioned running several win7 machines. Simultaneously? With 8GB RAM that may be pushing.

Finally, have you turned off debugging in your VMs?

Good luck and please post how you solve the issue.

matthewls

[chmick]

Hi KukulkanoKukulkano

I'm interested in your post since I have the same problem with workstation11 . The host is a debian 6 and the guests are linux and windows.

Things run smooth until i start the windows . At best I have a v-cpu0 error .  At worst the host end up frozen and i have to reboot everything .

By the way I'm looking for a good tuning guide for linux host ? any like to point me to ?

[Kukulkano]

Hello matthewls,

thank you for the tipps. Due to my holidays I've not been able to test all the settings. Sadly, it is not easy as it sometimes only happens after half an hour of VM usage. Hard to find out if a tip helped or not. It happens with Windows XP and Win 7 guests. No, I'm not running all of the the same time. Mostly only one.

@chmik

I set __GL_YIELD=USLEEP in my current environment (/etc/profile) and I also created new VM's by re-using the existing disks (no CD-ROM, no audio etc. only the basics). I also turned off "Accelerate 3D" for these machines. This seems to do the trick for me in the moment. Sadly, some machines still use way to much CPU together with kwin process. At least the VM is running at 3% while the host vmware-vmx + kwin using 200% of my system (4 CPU). But the host is no longer completely freezing.

[Kukulkano]

Ok, with some delay I recognize that this did not really help. I can see that all VM's start to consume lot of CPU together with kwin if I leave them alone for about 30 minutes. So, if I run a machine and leave VMWare Workstation in the background. I turned off any energy saving in those VM machines (guests). No screensavers, no auto-off etc. I also realized that it might be in conjunction with the dual monitors. Looks like moving windows from one screen to the other is making it even worse. Definitely some interaction effect with KDE screen manager (kwin). But maybe not the main issue.

I found other posts reporting problems similar to mine and now will try to investigate this solution:

VMWare Workstation 11 poor host performance

[Kukulkano]

Sadly, setting

mainMem.backing = "swap"

does not help, too. I consider going back to VMWare Workstation 9, as many people report that his is working better 😞

[Kukulkano · Accepted solution]

I found this page:

Disqus - Arch Linux becomes unresponsive from khugepaged

It looks like this did the trick!

I ran (as root) this:

echo never > /sys/kernel/mm/transparent_hugepage/defrag

echo madvise > /sys/kernel/mm/transparent_hugepage/defrag

Since then, everything is running fine. 🙂

[sptnx]

Hello,

after a lot of research I found this to work, with the following specs:

Host: Linux OpenSuse 13.2, x86_64, i5, 8Gb ram, with ATI/Intel Hybrid graphics, and VMware Workstation 12.

Guests: Win 7 SP1, x64, 3D accelerated.

I did:

echo never > /sys/kernel/mm/transparent_hugepage/defrag

echo madvise > /sys/kernel/mm/transparent_hugepage/defrag

and run VMware Workstation with:

vblank_mode=0 DRI_PRIME=1 vmware

Its working good with minor stops...

Thank you all!

[I1eo]

The solution works for me too - but only very temporary.

The described changes in configuration last only to a reboot, resume or even earlier (extended inactivity, etc.)

In order to make the solution permanent please consider to add those settings to the desired run-levels as described in this guide:

Often Overlooked Linux OS Tweaks

Please consider also to change swappiness as described in the link above.