On 5. 3. 2020 15:46, Thierry Parmentelat wrote:
Hi there
I am using libvirt+lxc on fedora
I just upgraded a couple boxes from f29 to f31
i am running into rather puzzling glitches on one of the boxes, and am having a hard time
trying to figure what exactly is going wrong
symptom being that on the faulty box, my 3 containers startup correctly, but all of them
stop after some random time shortly after (approx. in the 5-10 minutes)
in some cases I even lose one of the containers (its name no longer shows up in `virsh
list —all` and I need to reboot the box to reinstate it
is there any known issue with libvirt/lxc on fedora 31 ?
*****
here’s the little that I’ve been able to figure out so far by myself
* I have turned on debugging of libvirtd as per the docs
log_filters="3:remote 4:event 3:util.json 3:rpc 1:*"
log_outputs="1:file:/var/log/libvirt/libvirtd.log”
and am digging in that log file, as well as the system journal, and the containers
journals, to try and find some correlation;
* I have seen an error message saying this
> virNetSocketReadWire:1850 : End of file while reading data: Input/output error
This is likely the LXC monitor disappearing. Looks like something killed
the container (either from inside or outside) and libvirtd lost
connecton to the controller process running inside and thus cleaned it
up. The question is, who is killing the container. If you run 'ps' on
the host do you see processes running inside the container/containers?
There was a problem with our event loop when we switched to glib, but
those are fixed upstream. Do you mind trying just released
libvirt-6.1.0? It can be found in the virt-preview repo [1].
https://fedoraproject.org/wiki/Virtualization_Preview_Repository
Michal