Debian VM starts and Debian host hungs irreversible

[mabra]

Hello!

I would like to extend my infrastruture server, which is currently a fileserver only, by several services,
brought by VMs and so I installed VirtualBox (already in preparation, some weeks ago). This is
a debian host.

I exported a window guest from a windows host and imported it into the linux host without any problems.
2 seconds after start, the screen and everything including the network are frozen/dead.

Then I rebootet and created a debian guest on this debian host and configured it.
2 seconds after start, the screen and everything including the network are frozen/dead.

In this case, this would begin the install session and not guest software is in the game.
I tried with and without the root user (after changing the perms accordingly).
The logfiles on this machine, do not give a hint - there is just a hole.
I am using debian jessie on the host, 8.5, 3.16.7 and VBox 5.0.20

The VM got 1 GB RAM and 2 CPUs. The Mobo is a A1SAM-2550F from
Supermicro with 16 GB of RAM.

There are although some LXC VMs, but currently not enabled.
The network is based on OpenVswitch and the VMs are bridged to the host,
the virtual interface, the host itself is using (this is NOT a physical port).
The filesystem, the VMs are saved to, is ZFS with ACLs enabled.
At least the debian
It looks like the host is hibernated.

Any ideas would really be welcomed !! I could supply each log, but just
have no clue, where to look and for what.

Thanks anyway and best regards,
Manfred

[Perryg]

Post the guests log file ( as an attachment ). Right click on the guest in the Main Manager then click show log. Save and post as an attachment. Compress if it is too large to post.

[mabra]

Hi !

Good idea ? Forget to mention: I looked at that log first, found it empty!
If I use the VBox manager to "show log", I see the previous content!
This itself is just like follows:

Code: Select all

VirtualBox VM 5.0.20_Debian r106931 linux.amd64 (May  1 2016 18:28:45) release log
00:00:00.677566 Log opened 2016-08-28T04:48:42.777184000Z
00:00:00.677568 Build Type: release
00:00:00.677573 OS Product: Linux
00:00:00.677575 OS Release: 3.16.0-4-amd64
00:00:00.677576 OS Version: #1 SMP Debian 3.16.7-ckt25-2 (2016-04-08)
00:00:00.677610 DMI Product Name: A1SAM-2550F
00:00:00.677624 DMI Product Version: 123456789
00:00:00.677802 Host RAM: 16077MB total, 14728MB available
00:00:00.677807 Executable: /usr/lib/virtualbox/VirtualBox
00:00:00.677808 Process ID: 5741
00:00:00.677809 Package type: LINUX_64BITS_GENERIC (OSE)

It was a long night and I was probably too tired then. I'll now look again at
all logs - I'll post, if I find something interesting.

br++Manfred

[Perryg]

It would be nice to see the entire log if you want help, but with that said you are running the Debian fork of VirtualBox and you need to ask them what could be wrong as we have no idea what they change.

[mabra]

It would be nice to see the entire log if you want help

This is really ALL, what the logfile contains, like I wrote earlier!
It could help to know, how to increase the loglevel and even to disable file-write caching
for the log (which is a known problem in Linux, as I just found out with my debug scripts).

I run'ed a SystemTap script, which track all 'exec/fork/exit' call (I am not an expert on this and it might miss something)
and this just shows (excerpt), regarding just RTC<>0 in a frist step:

Code: Select all

Sun Aug 28 18:33:14 2016: EXEC: ( 6709/ 6688) nspr-2: file 00007feb2c9b6708 00007feb2c9b6705 "/usr/lib/virtualbox/VBoxNetAdpCtl hif0 speed 2>&1"
Sun Aug 28 18:33:14 2016: EXEC: ( 6710/ 6709) sh: file "/usr/lib/virtualbox/VBoxNetAdpCtl" "hif0" "speed"
Sun Aug 28 18:33:14 2016 : sh (6710) is exec'ing "/usr/lib/virtualbox/VBoxNetAdpCtl"
Sun Aug 28 18:33:14 2016: EXIT: ( 6710/ 6709) VBoxNetAdpCtl: exit code 4

This is taken from the SystemTaps log after reboot.
So I opened a console and executed the statement manually:

Code: Select all

$ /usr/lib/virtualbox/VBoxNetAdpCtl hif0 speed
VBoxNetAdpCtl: Error while retrieving link speed for hif0: VBoxNetAdpCtl: ioctl failed: Operation not supported

The configured network interface is named "hif01". If I configure "eth3" - which is currently temporarely
available - the VM starts ....

I've never tried to attach files here - I'll now attach two logfiles:
- proc1.txt: This is the version, wich hungs the machine (I think, the reason is not visible)
- proc2.txt: This is the version, where the VM at least started (nothing further, at the moment).

Without enabling VBox to a higher loglevel, I am unable to find out more.
It tells me definitively, it is a VirtualBox problem !
This tells me, the VirtualBox seems not fit in a virtualized networking world.
The interface "hif0" is working well with Linux.

So I made the next step and provided a new port on the OpenVswitch - just because
I noted, that the mentioned "eth3" is only temporarely available at the moment -
and added another port on the switch:

Code: Select all

$ ovs-vsctl add-port vbr0 vbx0 -- set Interface vbx0 type=internal

The I tried to start the VM - but the host is dead again !

Could someone offer more help, please ?
What about VDE ports from the VDE environment ? Though I hate to use too many different switches ...

BTW: The VM (which is named "debi" here - now has a logfile! If wanted, I can upload).

Thanks so far and best regards,
Manfred