Hi,
On 16/09/14 09:16, Guido Günther wrote:
However VMs create a machine.slice here:
./cpuset/machine.slice/machine-qemu\x2d<machinename>.scope/
Could you check if this gets properly removed when you shut down the domain?
Cheers,
I have a couple of bits of information that could prove useful.
In /etc/libvirt/qemu.conf I have the line:
user = "paul"
If I comment out this line and restart the machine then the qemu process
runs as user "libvirt+" and the problem goes away.
Second, I've found a work-around for the problem, which is to run (as
user paul) the following command:
kill $(cat
/sys/fs/cgroup/systemd/machine.slice/machine-qemu\\x2dxanadu.scope/tasks)
After this, the machine.slice directory no longer contains a
machine-qemu\\x2xanadu.scope directory.
The 'tasks' file contains four pids, two of which are active after the
domain has shut down. These still-running processes are:
/usr/bin/pulseaudio --start --log-target=syslog
and
/usr/lib/pulseaudio/pulse/gconf-helper
Before starting the domain, I have a pulseaudio and gconf-helper
processes running. After starting the domain, I have two pulseaudio and
two gconf-helper processes.
When I remove the user="paul" line from qemu.conf (so qemu runs as user
"libvirt+") no additional pulseaudio or gconf-helper processes are
started when starting the domain.
Therefore, it looks like the problem is caused by additional
pulseaudio/gconf-helper processes mistakenly created, or the process
responsible for killing them when the domain shuts down is unable to do
so or fails to receive the notification.
HTH,
Paul
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
