VM goes unresponsive though QEMU remains responsive

I’m Leap 16.0 (just upgraded) but problem occasioned in 15.6.
Got loads of memory in (OpenSuSE 16.0) VM, even though often lots of windows open and lots of tabs in Firefox,

Blockquote
$ free

shows lots of memory available.
BUT the VM is unresponsive. Clicking on other windows or <.ALT><.TAB> doesn’t to anywhere.
The mouse pointer is still visible. I can <.LeaveFullScreen> and see particulars in the Virtual Machine Monitor GUI, and can switch to my Ubuntu VM and do something.

But the OpenSuSE VM is dead to me.
I access the top menu, which is responsive, and choose

Blockquote
Virtual Machine | shutdown | force reset

Reboots and generally recover to approximately where and what I was doing.

Where do go to start debugging what is my problem?

You probably did run free on the host system, true?

If so this does not say anything about the memory allocated to the VM. Check those settings

No. Run on VM only.

$ free -h
total used free shared buff/cache available
Mem: 15Gi 12Gi 264Mi 274Mi 3.0Gi 2.6Gi
Swap: 2.0Gi 676Mi 1.3Gi

If it matters, my host is 64Gi.

df shows my VM is 42% used wtih 86GiB available.
My host is 4TiB using software RAID so it’s presumed 2TiB

I think it is good to first to determine whether:

  • only the GUI session froze, or
  • the whole guest OS froze.

When such a hang occurs:

  • Try switching to a text console: Ctrl + Alt + F3 through F6
  • If it does not work, try to SSH into the guest to see if networking is still up
  • systemctl status qemu-guest-agent (on the host)

Also, study the log of the previous boot that did hang using “journalctl -b -1”. You can share that using susepaste.

It’s only the OpenSuse guest that freezes.
I can get to the top and see the virtual machine manager–as well as do things.
VM in question is quiescent.

And can switch to my UBUNTU guest.

Can get the openSuse guest that won’t respond, but the top-level menu, I choose “Virtual Machine” and force a restart, and essentially pick up where I left off.

$ sudo journalctl -b -1
Specifying boot ID or boot offset has no effect, no persistent journal was found.

But the next time my guest freezes, I’ll try to do this.
Thanks for the counsel and advice.

If it matters, when I’m frozen, that is, no response to mouse-click nor <.alt><.tab>, things like my VLC playing internet radio continues unabated.
And the mouse obviously is moving about, as I can get back to my hypervisor, switch to another guest, etc.

^<.alt><.F3> gets my to my host terminal, I shared those results earlier – nothing!

When I click that <.leaveFullScreen> at the top of the screen to get to the virtual machine window, I can <.sendKey> and click send ^<.alt><.F3>, to try to get to my VM to do that journalctl from it instead of my host, nothing happens and it doesn’t change my situation.

This last time, after it came back from my <.virtualMachine | shutdown | forceReset> I immediately did the sudo journalctl -b -1 you requested, it showed the same nothing as before.

If it’s of any interest, when it comes back, sound no longer works. I must do a shutdown/reboot from my terminal to get fully back.

Thanks for all the help, guys; sorry if my terminology isn’t always up to snuff, and responses aren’t always the most timely as I’m away from my computer often.

@suseParsons What do the logs on the host say about the virtual machine?

What desktop environment is on the Client (Leap 16.0), what filesystem?

Both are BTRFS
Both are Leap 16.0
Both are Gnome.

forgive my ignorance:
On host, which logs do you wish me to inspect? And looking for what ?

@suseParsons just tail the host log when the vm is running journalctl -f to see is any output related when the vm freezes. Are you using QXL for the graphics on the vm?

Cannot find host log in /var/log nor /run/log

did a test of journalctl -f and it works fine, so I’m all set there next freeze.

virsh edit <.domain>
shows under graphics that I’m using spice
Nothing anywhere says QXL

Display Spice support has been dropped, maybe still there for Leap 16.0(?)

It’s the Video section… perhaps your using Virtio, if so suggest switching.

<video>
  <model type="qxl" ram="65536" vram="65536" vgamem="16384" heads="1" primary="yes"/>
  <address type="pci" domain="0x0000" bus="0x00" slot="0x01" function="0x0"/>
</video>

Thank you very much.
Now that I know what to look for, I found it.

What I’ve got is exactly what you suggested except all my answers are single-quote where yours are double-quote.

I do have some “spice” things, such as:

Let me try to paste that again:


</sound>
    <audio id='1' type='spice'/>
<channel type='spicevmc'>
      <target type='virtio' name='com.redhat.spice.0'/>
      <address type='virtio-serial' controller='0' bus='0' port='2'/>
    </channel>

<graphics type='spice' autoport='yes'>
      <listen type='address'/>
      <image compression='off'/>
    </graphics>

which I assume doesn’t matter.

@suseParsons If not used, I would suggest removing…

Hi malcolmlewis:
In the vein of if it ain’t broke, don’t fix it; if these “spice” things aren’t causing me problems, I hesitate to “fix” them for fear of making things worse.

As I’m understanding spice, it’s involved with pass-through/pass-along between my host and my guest. (?Am I close ?)

I sometimes cut and paste between them, such as getting my .conf data on my host and pasting it in this exchange. Is that a spice thing?

i NEED audio in my VM – I’ve got the internet radio (VLC) going nearly all day. More and more things are presented audio/visual instead of text as you look up things on the internet.

I don’t exactly recall how I fixed my lack of audio in my VMsl–it was long ago and far away – but perhaps that’s when spice got introduced to my .conf. (The research I just did in trying to figure out what spice is/does seems to confirm that.)

So I think I’ll leave it alone (unless you instruct me otherwise).

btw, if it matters, my MO is on Monday morning, I check for & install updates (Myrlyn) on my SuSE guest then generally reboot.

When I get through with anything on all other VMs, I then do the same on my host/server, after which I shutdown all VMs, then reboot the host/server, after which I start VMs as needed.

It should be fine, unless you see log messages related to it and the freeze…

Try to configure the VM to never sleep and never switch off the screen.

Hi: Thanks heitormoreira.
Sorry, been mostly off the computer, and when I’m back, busy catching up with other stuff and living with this annoying problem.

I’m not sure this is my problem because it’s not going to sleep, it’s just not responding to mouse or KB action. The Host still see it and the pointer moves around, and allows me to get to both itself and other VMs. I also didn’t know how to do what you requested so it took some time to work all that through.

In all that, I did find this:

sudo virsh dompmwakeup TreeBeard1
error: Domain ‘TreeBeard1’ could not be woken up
error: internal error: unable to execute QEMU command ‘system_wakeup’: Unable to wake up: guest is not in suspended state

I think what you’re asking me to do is my normal setting :

<pm>
   <suspend-to-mem enabled='no'/>
   <suspend-to-disk enabled='no'/>
   </pm

If I’ve misunderstood what you’ve asked of me/suggested, please correct – don’t give up on me!

[malcolmlewis]
I’m thinking that since my VLC is playing while I can’t access VM, perhaps I’m not freezing as I initially stated. And like I said, the host is seeing the mouse work as it’s moving the pointer, the VM just doesn’t respond, though the host does, which enables me to get to the hypervisor and force reset, etc.

I’m back to your suspicion on SPICE. As I alluded in a previous note, I’m not sure what Spice does, but think it’s handling the coordination between Host and guest, so the VLC sound of the VM gets to the HOST which gets it to the speakers – and earlier problem I mentioned I had early on.

To “get rid of it”, since it’s no longer being supported, are suggeting I just delete references when I

sudo virsh edit TreeBeard1

I’d suspect I need to replace it.
Where do I go to begin figuring out what to do?

Spice references from my conf

<channel type='spicevmc'>
      <target type='virtio' name='com.redhat.spice.0'/>
      <address type='virtio-serial' controller='0' bus='0' port='2'/>
    </channel>

and


 <graphics type='spice' autoport='yes'>
      <listen type='address'/>
      <image compression='off'/>
    </graphics>

<audio id='1' type='spice'/>

and

<redirdev bus='usb' type='spicevmc'>
      <address type='usb' bus='0' port='2'/>
    </redirdev>
 <redirdev bus='usb' type='spicevmc'>
      <address type='usb' bus='0' port='3'/>
    </redirdev>

What to do?

@suseParsons make a backup of the xml file located in /etc/libvirt/qemu/ and edit/delete and test.

Hey: Thanks for the quick reply. btw – in the midst of my reply, my situation happened again and I had to force-reset, then reboot it when it comes up (it doesn’t “reset” clean). ha ha ha

Ok, I’m in the slow group.
Edit ?
Delete the entire <,channel> entry, the <.graphics> entry, the <.audio> entry and both <.redirdev> entries?

Or just remove the “type” references ?

<.channel l> also has a <.target> … ?