Host & Guest freeze (lock up) randomly (hyperthreading?)

[robisaks]

When I start the guest it will run reliably for a seemingly random amount of time. After a while it will lock up the host and guest systems. It seems the lockup starts by preventing input (Mouse and keyboard become unresponsive, even to unplugging and replugging). If I have a command line open, the curser will continue to blink for a small amount of time before that too locks up. It requires a hard reboot of the machine to recover from this.

 http://forums.virtualbox.org/viewtopic.php?f=7&t=36753&p=177508

[moog]

I have this same problem. It seemed to start occuring after the upgrade to 4.0.4. The machine was working well for 6 months before that update.

Now the machine usually works for about 2-3 mins before locking up, though sometimes much longer (hours). I noticed that enabling the screensaver seemed to always cause the issue. In that the screensaver, when activated, would never return. I notied in another post that somebody else says turning off the screensaver cured the problem. I can confirm this is NOT the case. Turning off the screensaver makes the problem not happen when the screensaver starts, however it still happens randomly at other times. I can't be sure, but it always seems to be triggered by some graphical change.

This is always the case: I can still use the mouse to move select/move windows, but no operation on those windows work (e.g. Close) The contents of windows are not upated (e.g. top) and the keyboard seems to be ineffective.

I can't see anything in the logs (host or guest) that hint at the cause.

VBox Version 4.0.4 Guest Ubuntu 10.10 (32bit) Host Windows7 x64 Enterprise Host CPU Core I7 M620

[moog]

More info.

I got a LOT of lockups the last few days.

No screen saver active.

Turn number of cores from 4 downto 1. Turn off VT-x/AMD-V Turn off Nested Paging Leave PAE on.

No more hangs.

Therefore its something to do with graphics and the above. I note that it always seems to hang when entering text in a window - perhaps a rendering issue?

[moog]

Oh, still there in 4.0.6

[frank]

So it is probably the AMD-V setting. You could also try to restore all settings and selectively toggle each setting. Of course if you set up more than one guest CPU then the AMD-V setting must be enabled. Did this configuration ever work with older VBox versions?

[frank]

When the host freezes, could it be that there is a host kernel panic? Any chance to get access to the host log after this happen?

[robisaks]

I can check get the log file after it happens, I just need to know which log file to attach.

[vasili.zolotov]

I am experiencing what sounds like the same issue with Win7 guest and OSX 10.6.7 host. Hard freeze that requires some power button action. Locks up once an hour or perhaps more often. Running on the new Mac quad i7. Fiddled with VM settings until cows came home. Grudgingly, I have to migrate to parallels, because work can't wait. But I'd love to come back to my old favorite Virtualbox when this gets fixed.

Replying to robisaks:

When I start the guest it will run reliably for a seemingly random amount of time. After a while it will lock up the host and guest systems. It seems the lockup starts by preventing input (Mouse and keyboard become unresponsive, even to unplugging and replugging). If I have a command line open, the curser will continue to blink for a small amount of time before that too locks up. It requires a hard reboot of the machine to recover from this.

 http://forums.virtualbox.org/viewtopic.php?f=7&t=36753&p=177508

[robisaks]

I've found that disabling hyper threading dramatically increases the amount of time before the lock/freeze happens. It may resolve it entirely, I am still testing...

[frank]

robisaks, that's an interesting observation. I assume the log file you attached to this defect was with hyperthreading enabled?

[robisaks]

Replying to frank:

robisaks, that's an interesting observation. I assume the log file you attached to this defect was with hyperthreading enabled?

That is correct, the version of the original post could be upped to the latest version of vbox as well.

Here are my results so far, it seems to be working:

Fresh install of ubuntu 10.04 64bit on a fresh hard drive. Tested a myriad of settings configurations. Eventually I tried disabling hyperthreading in the bios and the vbox (OSX) seemed to be stable. I re-installed my original hard drive and tested with a fresh (and default) set up for OSX, it's been running for 12 hours and has not crashed.

A couple of things I am curious about. With hyperthreading enabled, ubuntu system monitor sees 4 CPU's and Vbox sees 8. With hyperthreading disabled, system monitor sees 2 CPUs and vbox sees 4. Could this variance be the reason vbox locks up the whole machine with hyperthreading enabled? When testing yesterday, the crash would nearly always occur in the following order:

1. CPUs are steady, trading the load back and forth, but generally at 100% (or whatever max % I set)
2. All CPUs drop to 0% for 10-15 seconds
3. Input lockout occurs, system no longer recieves input from usb attached devices, however, the mouse trackpad on the laptop continues to function, but not the keyboard. I can not interact with anything, system is frozen, except for the mouse.
4. 1 CPU jumps to 100%, seemed to mostly be CPU #2, CPU 3 and 4 were the ones doing most of the work. This CPU will jump to 100% even if I have a limiter on the % set up in the settings.
5. System lockup occurs, forcing manual shutdown.

[frank]

Could you again attach two VBox.log files, one with hyperthreading enabled and one with hyperthreading disabled (to be able to compare them)? Just only start a VM and shut it down again.

[robisaks]

Replying to frank:

Could you again attach two VBox.log files, one with hyperthreading enabled and one with hyperthreading disabled (to be able to compare them)? Just only start a VM and shut it down again.

I should be able to provide this over the weekend.

[jwythe@…]

I have same problem on two of my systems while the other two are fine. I have posted info else where on this site, and the Forum's Let me know if you would like more details.

[mrworf]

Last screen from my Ubuntu 10.04 LTS after crash

[mrworf]

I'm experiencing the same issue.

I'm running Ubuntu LTS 10.04 LTS on the host, and on that machine I run 4 Ubuntu 10.04 LTS guests. It's VBox 4.1.2 but older editions showed the same problem.

The issue at hand seems to be that whenever one of my multi-core guests run at peak performance (maxing out all four of its alotted cores) it will eventually kill the host. The host is unable to reboot (it says rebooting in 5s but it's stuck). MCE on Ubuntu claims HW issue but that is a lie, I think it gets confused by the way it crashes.

Some specs:

Host is a Core i7 with 24GB RAM, 2 raid arrays (software) and 3 network cards. The machine has been tested (Memory and CPU) for two days solid without any errors or issues. So the hardware is solid.

On this machine it looks like this:

• 1 core/2GB --> Firewall (bridged to two of the cards)
  
• 1 core/4GB --> Mail (virtual net to Firewall)
  
• 1 core/2GB --> Asterisk (briged to the one of the two cards shared with the firewall)
  
• 4 cores/8GB -> Apache/MySQL (virtual net to Firewall)
  

The web server guest is capped to 90%, all guests run a divider of 10 and nolapic_timer, this was necessary to get any kind of uptime.

As long as the web guest isn't maxing out it's system, things generally work fine.

Had to install an eRIC G4 card to remote control/reboot the host since it's located in another continent :) which is why I can provide the kernel panic info.

Othwerwise, when it crashes no info is saved anywhere nor are there any errors in the VBox logs. All looks OK.

[mrworf]

I upgraded to 4.1.14 with great hopes only to have them smashed. VirtualBox still crashes the host.

[BearFather]

Any update on this issue. I see it went dead for 3 months. Does this mean it's been fixed? I have the same issue with our setup.

[mrworf]

BearFather: It turned out that in my case, it was the CPU that was faulty. I got exceptions which said that it was a hardware issue and not software, so I replaced basically everything on my system except CPU (like, motherboard, psu, memory, etc) but in the end it seemed that everytime I used SSE or loaded down the system, it would crash, extreme cooling only seemed to delay the crash.

One quick switch of the CPU and now it's been running for 3 months straight or so.

[jwalton]

If you have docker installed, could be  https://github.com/docker/docker/issues/10031.