[openstack-dev] [Nova] [nova-lxd] Broken nova-lxd

James Page james.page at ubuntu.com
Mon Jul 31 08:22:47 UTC 2017


On Thu, 27 Jul 2017 at 23:58 Michael Still <mikal at stillhq.com> wrote:

> Hi,
> I'm cc'ing openstack-dev because your email is the same as the comment you
> made on the relevant review, and I think getting visibility with the wider
> Nova team is a good idea.
> Unfortunately this is a risk of having an out of tree Nova driver, which
> has never been the recommended path for implementing drivers for Nova.
> Being out of tree isn't forbidden, but it does come with the cost of
> staying up to date with Nova and handling changes as they occur.

Agreed - and we've felt this from time-to-time since the nova-lxd driver
was started.

Maybe the Queens cycle might be a good time to review our out-of-tree-ness
and see whether the Nova team would be prepared to accept the LXD driver

I'm more comfortable now with where we are with regards to devstack tempest
testing in the gate for the LXD driver, although we do still have some
feature gaps compared to the libvirt driver (specifically boot from volume)
which results in a large-ish skip list.

> In this case, if you look at the review chain you'll see that the move is
> a pre-cursor to moving this code to use oslo.privsep. Unless lxd is going
> to move to privsep in lockstep with nova, your best bet would be to
> duplicate this utility method in the nova-lxd codebase.

We'll duplicate right now (see [0]) to unblock the current problem and take
a look at the work to move lockstep with nova to use oslo.privsep.



[0] https://review.openstack.org/#/c/488254/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20170731/95040983/attachment.html>

More information about the OpenStack-dev mailing list