On 04/20/2015 08:45 PM, Matthew Schumacher wrote:
List,
I was under the impression that I could restart libvirtd without it
destroying my VMs, but am not finding that to be true.
If not, then something is wrong.
When I killall
libvirtd then my VM's keep running, but then when I start libvirtd it
calls qemuDomainObjEndJob:1542 : Stopping job: modify (async=none
vm=0x7fb8cc0d8510 name=test)
That message doesn't indicate that a domain is being destroyed, just
that libvirt performed some operation on the domain that modified it in
some way.
and my domain gets whacked.
If your domain is terminated, there should be a libvirtd error log (at
least one saying that it got end of file from the monitor, meaning that
the qemu process itself terminated unexpectedly. In that case there
should be some indication of qemu's problem at the end of
/var/log/libvirt/qemu/test.log).
Any way to disable this behavior?
Whatever the bug is causing this to happen, fix it :-)
Also, while I'm at it, due to issues with snapshotting, I ended up with
two domains where libvirt insists the disk source is incorrect which
causes many things to break like this:
root@wasvirt1:/etc/libvirt# virsh domblklist wasdev
Target Source
------------------------------------------------
vda /glustervol1/vm/wasdev/wasdev.reboot
hdc /dev/sr0
Yet:
root@wasvirt1:/etc/libvirt# lsof | grep wasdev
2235 /usr/bin/qemu-system-x86_64 /var/log/libvirt/qemu/wasdev.log
2235 /usr/bin/qemu-system-x86_64 /var/log/libvirt/qemu/wasdev.log
2235 /usr/bin/qemu-system-x86_64 /glustervol1/vm/wasdev/wasdev.qcow2
The reason is because a blockcommit --active --pivot doesn't work, I
blockjob --abort, then try again, it works the second time, but then the
disk is wrong.
Help would be greatly appreciated.
Thanks,
schu
_______________________________________________
libvirt-users mailing list
libvirt-users(a)redhat.com
https://www.redhat.com/mailman/listinfo/libvirt-users