Xorg and Oom killer, i3

[cyrinux]

Hi,
I have 20gb memory, but I use a lot of VM.
Sometimes xorg is randomly kill.
I would like to protect xorg from being killed. I'm trying to play with overcommit but is it a good idea?
Must I boot dom0 minimal memory?

Bonus: after a xorg crash, I can't reconnect to guid daemon. How to do? I use i3, is it a i3 problem?
Regards

[Noor Christensen]

What version of Qubes are you running on your dom0?

By "using a lot of VM", do you mean virtual memory or virtual machines?
For clarity, I will use the term meaning the latter, and "virtual
memory" when referring to the former.

Just want to make clear since the topic makes it a bit ambiguous :-)

Is it your dom0 Xorg process that is being killed, or one that is
running in a VM?

You might want to take a look at how your resources are being used by
dom0 and any running VMs. Run "xentop" in a dom0 terminal for a nice
realtime summary.

-- noor

|_|O|_|
|_|_|O| Noor Christensen
|O|O|O| no...@fripost.org ~ 0x401DA1E0

[cyrinux]

Hi Noor,
By VM, i mean Virtual Machine (we can say Qubes), I have at minimum 15 Qubes running.
It is xorg in dom0 which is killed, I type 'dmesg' in dom0 terminal and see it is killed.
I will retry to play with xentop. In idle my dom0 use 700MB (and 2200MB cache)/~3000MB.
About reconnect to guid, before i use 'qvm-run --all true' to reconnect to them, but after this OOM this doesn't work. Do you have an idea?

Regards

[cyrinux]

Sorry, and i use last 3.2 version, fully updated, some debian/fedora/whonix qubes.

[Ilpo Järvinen]

On Sat, 15 Jul 2017, cyrinux wrote:

> By VM, i mean Virtual Machine (we can say Qubes), I have at minimum 15 Qubes running.
> It is xorg in dom0 which is killed, I type 'dmesg' in dom0 terminal and see it is killed.
> I will retry to play with xentop. In idle my dom0 use 700MB (and 2200MB cache)/~3000MB.
> About reconnect to guid, before i use 'qvm-run --all true' to reconnect to them, but after this OOM this doesn't work. Do you have an idea?

You can increase maximal dom0 memory by tweaking grub cmdline to see if it
helps. By default, dom0 is limited by dom0_mem=max:4096M. You should still
try keep it less than what you have physically available (e.g., try 6G or
8G rather than going to some large number to avoid starving appvms from
memory).


--
i.

[Noor Christensen]

On Sat, Jul 15, 2017 at 11:57:26AM -0700, cyrinux wrote:
> Le samedi 15 juillet 2017 20:20:55 UTC+2, Noor Christensen a écrit :
> > On Sat, Jul 15, 2017 at 10:55:16AM -0700, cyrinux wrote:
> > > Hi,
> > > I have 20gb memory, but I use a lot of VM.
> > > Sometimes xorg is randomly kill.
> > > I would like to protect xorg from being killed. I'm trying to play with overcommit but is it a good idea?
> > > Must I boot dom0 minimal memory?
> > >
> > > Bonus: after a xorg crash, I can't reconnect to guid daemon. How to do? I use i3, is it a i3 problem?
> > > Regards
> >
> > What version of Qubes are you running on your dom0?
> >
> > By "using a lot of VM", do you mean virtual memory or virtual machines?
> > For clarity, I will use the term meaning the latter, and "virtual
> > memory" when referring to the former.
> >
> > Just want to make clear since the topic makes it a bit ambiguous :-)
> >
> > Is it your dom0 Xorg process that is being killed, or one that is
> > running in a VM?
> >
> > You might want to take a look at how your resources are being used by
> > dom0 and any running VMs. Run "xentop" in a dom0 terminal for a nice
> > realtime summary.
>

> Hi Noor,
> By VM, i mean Virtual Machine (we can say Qubes), I have at minimum 15 Qubes running.
> It is xorg in dom0 which is killed, I type 'dmesg' in dom0 terminal and see it is killed.
> I will retry to play with xentop. In idle my dom0 use 700MB (and 2200MB cache)/~3000MB.
> About reconnect to guid, before i use 'qvm-run --all true' to reconnect to them, but after this OOM this doesn't work. Do you have an idea?

If you check the RAM and CPU allocation settings for those 15 VMs you have
running, are they reasonably set? I mean, compared to your total amount
of memory.

This will give you the current settings for a specific VM:

# qvm-prefs -l <vm-name> | egrep 'memory|maxmem'

You might have to experiment a bit with those settings, since there are
no optimal defaults for this. It really depends on what you are doing
with your VMs, and how much you allocate for each VM.