| Thread Previous • Date Previous • Date Next • Thread Next |
Public bug reported:
I originally thought this was bug 1306342 but that's a different issue,
that was when not having neutron configured properly for calling back to
nova and nova would timeout on spawn waiting for a notification from
neutron that networking was setup.
This is a different issue where resize/migrate fails if you started from
a stopped instance and using neutron. In this case, the
_create_domain_and_network method in the libvirt driver passes in
power_on=False since the instance was stopped before the
resize/migration. The virtual interface isn't plugged in that case so
we're waiting on a neutron event that's not going to happen, and we hit
the eventlet timeout which then tries to destroy the non-running domain,
and that fails with a libvirtError telling you that the domain isn't
running in the first place.
The fix is to check the power_on flag in _create_domain_and_network and
if it's False, don't wait for neutron events, same as if
vifs_already_plugged=False is passed in.
** Affects: nova
Importance: Undecided
Assignee: Matt Riedemann (mriedem)
Status: In Progress
** Tags: libvirt migration network neutron resize
** Changed in: nova
Status: New => In Progress
--
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to OpenStack Compute (nova).
https://bugs.launchpad.net/bugs/1321872
Title:
Resize/migrate stopped instance fails with Neutron event timeout
Status in OpenStack Compute (Nova):
In Progress
Bug description:
I originally thought this was bug 1306342 but that's a different
issue, that was when not having neutron configured properly for
calling back to nova and nova would timeout on spawn waiting for a
notification from neutron that networking was setup.
This is a different issue where resize/migrate fails if you started
from a stopped instance and using neutron. In this case, the
_create_domain_and_network method in the libvirt driver passes in
power_on=False since the instance was stopped before the
resize/migration. The virtual interface isn't plugged in that case so
we're waiting on a neutron event that's not going to happen, and we
hit the eventlet timeout which then tries to destroy the non-running
domain, and that fails with a libvirtError telling you that the domain
isn't running in the first place.
The fix is to check the power_on flag in _create_domain_and_network
and if it's False, don't wait for neutron events, same as if
vifs_already_plugged=False is passed in.
To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1321872/+subscriptions
| Thread Previous • Date Previous • Date Next • Thread Next |
