Temperamental VNC

[NoOp]

On 09/04/2010 06:31 PM, NoOp wrote:
> On 09/01/2010 03:33 PM, Michael Pavling wrote:
>> I've been getting an annoying problem on a new machine I've set up.
>> It's running Ubuntu 10.04.1 AMD64 desktop, but is "headless". If I ssh
>> onto it and launch VNCserver (I've installed TightVNC) I can log on
>> and it works fine for a time. But at some point - generally when I
>> click a button in VirtualBox, the VNC viewer closes on my PC, and in
>> my ssh window, I can see that the VNC server process has died.
>> Starting a new vncserver starts on a new port (as there are still
>> locks in /tmp from the previous process).
>> 
>> Any idea what's causing VNC to die? Or is there a better/different way
>> to administer a headless machine (I'm temporarily using remote desktop
>> to avoid the process terminations, but I have to be logged in on the
>> console for it to work, which isn't acceptable long term - I need to
>> be able to reboot without concern)
> 
> I'll forget the VRDP bit & just let you know that after trying to
> duplicate your problem, I've found that VNC and VirtualBox doesn't seem
> to want to play well together. Here is the prelim:
> 
> 1. I've two machines/desktops w/monitors side-by-side, both running
> 10.04 (gnome desktops) and VirtualBox PUEL 3.2.6 r64453 (32bit) with
> Win2KPro guests.
> 
> 2. I VNC to desktop B and start a VM Win2KPro session. The session will
> start, but then goes to 98% CPU and the session stalls (grey desktop).
> VNC doesn't die, but the VM session stalls out completely.
> 
> 3. I kill the VNC session from desktop A and get to the Win login on
> screen on B. Once I log into the Win VM session (using the keyboard
> etc., directly on machine B), I can restart VNC from desktop A and work
> in desktop B's VM guest.
> 
> I notice that it's not vino-server or VirtualBox that is eating hte cpu
> cycles, but Xorg. Xorg (when the VM is working OK w/VNC is sucking up to
> 83% CPU (2.4Ghz Intel processor). If I kill VNC on desktop A, Xorg drops
> to about 3.7% on desktop B (with the VM running. Fire VNC back up on
> desktop A to desktop B and Xorg 80+ percent CPU.
> 
> I'll see if I can find time to explore further. But so far, in this
> case, I see Xorg as being the possible culprit with both VirtualBox and
> vino-server behaving quite well. While you are ssh'ed into the machine,
> run top & see if you get the same/similar.

OK. Just want to add that if I NX into desktop B and start the
VirtualBox session, the VM comes up (albeit slowly) and Xorg stays at
about 2%. So actually is appears that the problem is Xorg combined with
VNC that is the problem.

Note: I am not using FreeNX, but instead using NomachineNX.
http://www.nomachine.com/download.php
Also note that NX uses a different screen than the original Screen 0, so
perhaps that is why Xorg doen't go CPU ballistic.