[FFe] Let's get LXD 2.0 final in Xenial
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
lxd (Ubuntu) |
Fix Released
|
Undecided
|
Unassigned |
Bug Description
Hello,
LXD is currently at version 2.0.0 beta3 in the archive, our current release plan is:
- beta4 tomorrow (includes feature work)
- rc1 next week or the week after (includes feature work)
The main features we're talking about are:
- Replacement of lxcbr0 by lxdbr0
- Replacement of the standalone lxd-images script by a native simplestreams client
- Addition of container resource consumption reporting to the API
All of that should be done and landed by mid-March at the latest. Our goal is to have Xenial release with a current upstream bugfix release of LXD 2.0 so we have a good stable experience at release time.
A quick word on the other bits. LXC is currently at 2.0.0 rc2 and so hit upstream feature freeze ahead of the Ubuntu feature freeze. We don't expect needing any FFe for it. We're just holding off its final release until we know that LXD is good to go.
LXCFS is currently at version 2.0.0 beta2 but an rc should be very near as we don't expect any feature addition to it for the LTS.
The combination of LXC 2.0.0 and LXCFS 2.0.0 should mean that cgmanager will no longer be required in Xenial and so we will not be tagging a long term support release for it, we'll instead let it be demoted to universe for those users who use it directly for uses outside of lxc containers.
In summary, the LXD team would like to get a feature freeze exception to be able to land LXD 2.0.0 final in the Xenial archive.
Thanks for the heads-up, this sounds mostly fine. I take it that this is being coordinated with nova-compute-lxd and juju (whose local provider is also moving to LXD)?
> Replacement of lxcbr0 by lxdbr0
What's the rationale for this? This sounds a bit like change for change's sake -- lxcbr0 has been around for a long time including LTS releases, and by now third-party software or admins might depend on the name and use it to communicate with containers. Will the "lxc" package go away? Is this meant to not provide direct networking between "plain" LXC and LXD containers any more?