virtualbox.org

Hello *,

I just encountered a "Guru meditation" (repeatedly). I did quite some research, but everything I found were bugs that are supposedly already fixed for many years.

Here's my setup:

Host
OS: Ubuntu 14.04.4 LTS
Kernel: 3.13.0-86-generic
VirtualBox: 4.3.36-dfsg-1+deb8u1ubuntu1.14.04.1
CPU: Intel Xeon CPU E31270 (4 cores; hyperthreading => 8 virtual cores)
Architecture: x86_64 (64 bit)
RAM: 16 GB
HD: soft-RAID 1 (mirror) with 2 * 2 TB

Guest 1
OS: Ubuntu 16.04 LTS
Kernel: 4.4.0-22-generic
CPU: 1 core
Architecture: i686 (32 bit)
RAM: 1 GB
Acceleration: VT-x/AMD-V, Nested Paging, PAE/NX
HD: SATA 20 GB / + 500 MB /boot + 1 GB swap

Guest 2
OS: Ubuntu 16.04 LTS
Kernel: 4.4.0-22-generic
CPU: 4 core
Architecture: x86_64 (64 bit)
RAM: 4 GB
Acceleration: VT-x/AMD-V, Nested Paging
HD: SATA 200 GB / + 500 MB /boot + 4 GB swap

Please note that both guests run in separate user accounts on the host.

Attached, I send the VBox.log showing "Guru Meditation 1155 (VINF_EM_TRIPLE_FAULT)".

It seems to me that this is related to both guests running at the same time. The Guest 1 was working fine for a few weeks. I added Guest 2 today. First it seemed fine, but then I started having problems on both guests: The guests were sometimes freezing completely (nothing happened during the "Unpacking" phase of an "apt-get install" and one of the host's CPU cores was at 100%) or I encountered Guru meditations.

The newest Guru meditation (described in the attached VBox.log) happened right during the boot process, but sometimes they happened later.

Did I overlook an open bug? Should I open a new issue ticket? Is there any workaround known? Would it help to upgrade the host to Ubuntu 16.04 LTS (does this actually come with a newer VirtualBox or isn't 14.04 LTS up-to-date, too?)?

Btw. I already tried disabling the VT-x/AMD-V, Nested Paging (both or only the latter) and it made things far worse (then VirtualBox already crashed when only running one single VM instance - not even both - and usually during the booting, already).

Thanks a lot in advance!

Best regards, Marco

Your host has four CPU cores. You are trying to run a guest with one core and another with four cores = five cores for the guests...
Hyperthreading doesn't really help for virtualization.
Please try configuring the bigger guest to only two CPU cores, this way the host has a core left to handle the guests and everything else.

Thanks a lot for your *very* fast response!

The guests are most of the time idling. Hence, I thought that this should be no issue. Only sometimes, the 2nd guest has to do really a lot - and then 4 cores would be really helpful.

However, I'm pretty sure that I had the same problem already with less cores (I tried really a lot of different configurations, already). But I'm going to try this (again).

Btw., IMHO such kind of misconfiguration should cause bad performance, only, and not a crash (a.k.a. Guru meditation)...

A triple fault is the guest trying to do an impossible operation. You can shut this off but I would not advise it. As for the 4 cores going to a guest, it may work with one guest but the second guest will trigger this kind of reaction. When setting up multiple guests you should never exceed the true core count combined.

Setting "Guest 2" to only 2 cores did not help. I just encountered the same type of crash again - "Guru Meditation 1155 (VINF_EM_TRIPLE_FAULT)" - see new log file attached.

Did you also set the first guest to 2 cores?

What happens if you just run this guest by itself?

00:00:00.010182 Package type: LINUX_64BITS_GENERIC (OSE)

You are also using the Ubuntu fork and we don't support that one. If you want to continue with us you need to remove it and install the official build from virtualbox.org.

Thanks for the quick reply!

No, Guest 1 still runs with 1 core, only.

Concerning the stability of Guest 2 alone, I'd have to do some more testing. I cannot say for sure, yet, because I ran only Guest 1 alone for a while - Guest 2 was added today. I have to do these tests tomorrow - it's too late, now.

Hardware problem!

I gave up using this machine and switched to a different server (AMD, 8 real cores instead of bullshit-hyperthreading). Now, it works fine! Yabbadabbadoooh!!! Btw. with the very same OS - just installed on different hardware.

And btw. there seems to be no problem to configure more CPUs for the various VMs than there physically are. The result is as expected: If all of them are busy, the performance goes down. But this is no issue, because most of the time most VMs are idle. At least until now, I didn't encounter any error. I hope it stays this way.