sandbox VM crash with kernel:bug: soft lockup

[#1]

I downloaded the sandbox VM for virtualbox and started it on ubuntu host with virtualbox 5 and after a couple of hours it crash.

On the screen I get :

Message from syslog@sandbox at dec ....

kernel:Bug: soft lockup - CPU2 stuck for 67s! [java:30253]

and this message repeat itself on and on.

The vm is stuck and non responsive. I need to poweroff and restart it!

Here are the log vboxlog relevant lignes:

00:01:05.829053 VMMDev: Guest Log: 00:00:00.011489 main 4.3.22 r98236 started. Verbose level = 0 00:08:50.366623 TM: Giving up catch-up attempt at a 60 002 234 869 ns lag; new total: 60 002 234 869 ns 14:11:48.845736 PIIX3 ATA: Ctl#0: RESET, DevSel=0 AIOIf=0 CmdIf0=0xca (30357026 usec ago) CmdIf1=0x00 (-1 usec ago) 14:12:05.363192 PIIX3 ATA: execution time for ATA command 0xca was 46 seconds 14:12:05.366460 PIIX3 ATA: Ctl#0: finished processing RESET

The host is a dell server with 2 quad xeon 32 gig ecc ram.

I have two other VM running on the host. A debian 64 and a ubuntu lts 14.04 that run just fine.

Anybody had a luck with virtualbox 5?

Peter

[#2]

The error suggests you may not have assigned enough CPUs to the VM. How many did you give it? Also, it seems there are known issues on older Ubuntu kernels: http://serverfault.com/questions/336591/how-to-fi...

[#3]

I don't think it's related to cpu assignment, The vm has 4 cpu @ 100% cap. And doing nothing. The load is below 0.3 most of the time.

As for the kernel! The host has latest current kernel

Linux host 3.13.0-74-generic #118-Ubuntu SMP Thu Dec 17 22:52:10 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

And the known issue is related for kernel in lucid-updates not LTS 14.04

Thanks for the suggestions.

[#4]

@Peter Young: Sure makes sense.

@rmolina: have you guys encountered this before?

[#5]

@Ali Bajwa

I remember someone else posting regarding an issue with newer versions of virtualbox, but I wasn't able to reproduce. we may have to investigate further.

[#6 · Accepted solution]

After some research I was able to have the VM work for more than 12 hours now.

Looks like virtualbox 5 has some new features regarding clock.

Here is my fix:

<code>I added grub arguments on the kernel clock=pit nosmp noapic nolapic 
in order to have 

clocksource=kvm-clock

Look to check clocksouce:

Check what clocksource that you are using:

<code>$ cat /sys/devices/system/clocksource/clocksource0/current_clocksource

Check which ones that are available:

<code>$ cat /sys/devices/system/clocksource/clocksource0/available_clocksource

If you got a "kvm-clock" entry in "available_clocksource" try to use it by setting kernel parameter:

<code>clocksource=kvm-clock

From http://serverfault.com/questions/132197/best-pract...

Peter