Comment # 14 on bug 1156336 from Aleksa Sarai

(In reply to seve skeis from comment #13)
> (In reply to seve skeis from comment #12)
> > (In reply to Aleksa Sarai from comment #11)
> > > (In reply to seve skeis from comment #10)
> > > > (In reply to Aleksa Sarai from comment #9)
> > > > > (In reply to seve skeis from comment #6)
> > > > > > Hi, actually, all using root, no wired structure, and on real hardware,
> > > > > > never tried on a vm. Can i ask what are the packages you installed?
> > > > > 
> > > > > I installed the same ones you did -- the ones that are in the Leap 15.1
> > > > > repos. I am running LXD on my server (on bare metal) and it also works fine,
> > > > > but the reason I tested it in a VM is to check whether there was an issue if
> > > > > you did a fresh install (I've upgraded my server incrementally from Leap
> > > > > 42.2).
> > > > > 
> > > > > Do you only have this problem on one machine, or can you replicate the
> > > > > problem on any other machines? What is the output of dmesg after LXD crashes
> > > > > (please don't paste it as a comment -- add it as an attachment)? If you run
> > > > > just 'sudo lxd` in a terminal (to start the server in your shell), what
> > > > > happens?
> > > > 
> > > > i have no issues with pc, its fresh install, no repos, no apps, just trying
> > > > to get LXD work
> > > 
> > > My question was whether this only happens on this particular machine -- do
> > > you have another PC or laptop on which you can run this test? The reason I'm
> > > asking is to figure out whether it's specific to your hardware.
> > > 
> > > > can you check the thread mentioned please
> > > 
> > > I read it after you posted comment 2.
> > > 
> > > > all this commands and dmesgs and logs are there.
> > > 
> > > Ah I missed that you ran 'sudo lxd -d' (I thought you modified the .service
> > > file the last time I skimmed through it). But there isn't a dmesg log -- the
> > > logs posted were from journalctl or from the output of LXD. dmesg will give
> > > you 
> > > 
> > > It would also be helpful to get the coredump (which gives useful debugging
> > > information to understand in which function the crash occured) -- you can
> > > get it using coredumpctl. It might be too large to upload here, but you can
> > > always upload it on a temporary sharing site and I'll download it.
> > > 
> > > > and my pc hardware details too.
> > > 
> > > (For future reference.)
> > > > CPU I7 6700K
> > > > RAM 32GB DDR4
> > > > MB: ASUS MAXIMUS EXTREME
> > > > G: NVIDIA GTX 970 TURBO
> > > > P.S: do not know if its worth mentioning, i had to boot with kernel param: acpi_enforce_resources=lax.
> > > 
> > > I notice you're using the NVIDIA drivers:
> > > 
> > > > Fri Nov 8 06:55:09 2019
> > > > +-----------------------------------------------------------------------------+
> > > > | NVIDIA-SMI 440.31 Driver Version: 440.31 CUDA Version: 10.2 |
> > > > |-------------------------------+----------------------+----------------------+
> > > 
> > > Do you have the same problem if you run with the Nouveau drivers? I ask
> > > because LXD supports GPU virutalisation (which uses a bunch of features from
> > > the GPU, GPU kernel module, and the userspace libraries).
> > 
> > Hi, i will run dumps, later and share it, i have not other machines, have
> > not tried with nouvuew, but it works with snapd with nvidia.
> 
> My system dose not have coredumpctl , or systemd-coredumpctl. and do not
> know how to use it.

First, trigger the coredump (run `lxd`) then do

  % sudo coredumpctl info lxd

which should give you the latest backtrace and a few other bits of information
from LXD. This should give us plenty of information by itself, but to get the
actual coredump you need to do:

  % sudo coredumpctl dump lxd > lxd.core

If the above give you errors about missing corefiles, then you might have to
modify your coredump.conf configuration -- but for me it just works out of the
box.

You can also disable systemd-coredump and just make a regular corefile if you
modify the kernel.core_pattern sysctl to be a simple file:

  % sudo sysctl -w kernel.core_pattern=%e.%p_%u.%g_%t.core

And then if you trigger the coredump, there will be a corefile in your current
directory with a name that looks like "lxd.$PID_$UID.$GID_$TIME.core".

> i will be using snapd, till you guys fix it.

I'm not sure how it's reasonable to think we can fix it, if we can't even
reproduce it. But sure, feel free to use whatever works for you.

> i suggest you use real hardware with fresh leap 15.1, then the issue will be clear whether its my hardware or a bug.

Unless I'm missing something, several other people in the original thread said
they tried on real hardware (just as I'm running things on my real hardware)
and didn't have issues.