Systemd freezing execution in VMs on manually running virsh list

[Jiří Stránský]

This could well be a libvirt/qemu bug. Reporting it mostly to have some info in case someone else hits it too. Or perhaps someone might find a workaround.

Some ways of executing `virsh list` seem to cause the virsh list command to hang, and the VMs to stop working properly.

Executing as root:

sudo -u stack virsh list --all

hangs the command, and in an undercloud VM this is broadcasted:

systemd[1]: Caught <BUS>, dumped core as pid
systemd[1]: Freezing execution.

(I already lost the exact broadcast but reconstructed the above from googling.) After this, pretty much everything stops working in the undercloud VM, i guess the same would affect overcloud.

The symptom when trying to ssh from the host looks like this:

ssh -F /root/wd-containers/ssh.config.ansible undercloud
Warning: Permanently added 'dell-t5810ws-rdo-10.tpb.lab.eng.brq.redhat.com,10.40.128.52' (ECDSA) to the list of known hosts.
ssh_exchange_identification: Connection closed by remote host

[Jiří Stránský (jistr)]

Adding -i to the sudo command seems to prevent virsh list from hanging. Not sure yet if it prevents the machines from freezing too, will redeploy :)

[Jiří Stránský (jistr)]

Changed the title to make it more apparent that this doesn't block quickstart deployments, the freeze happens when running virsh list manually.

[Launchpad Janitor (janitor)]

[Expired for tripleo-quickstart because there has been no activity for 60 days.]

[Oliver Walsh (owalsh)]

The root cause appears to be that virsh is connecting via different sockets. It has been discussed previously here:
https://www.redhat.com/archives/libvirt-users/2016-March/msg00056.html

And there is this patch in oooq that attempts to workaround it, although I wonder if this makes matters worse:
https://github.com/openstack/tripleo-quickstart/commit/f21d7a6d48ac21087c9433e75dc217b85349d8d7