Ubuntu support for virtualization

Stéphane Graber stgraber at ubuntu.com
Thu Nov 17 15:19:27 UTC 2011


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512

On 11/14/2011 09:52 PM, Steve Langasek wrote:
> Hi Mark,
> 
> On Mon, Nov 14, 2011 at 05:59:02PM +0000, Mark Shuttleworth wrote:
>> On 11/11/11 17:13, Bodhi Zazen wrote:
>>> The issue I am raising is not maintenance of the kernel or
>>> kernel patches, the question is as to how Upstart bugs are
>>> managed on Launchpad.
> 
>> That wasn't even in the subject of your email, Bodhi ;-)
> 
>>> The common theme with this technology is that the Ubuntu guest
>>> needs to boot in a chroot environment, and that is where
>>> Upstart bugs occur.
> 
>> As of 11.10 we started to focus a lot more attention on LXC. I 
>> completely agree that Upstart will need to work smoothly in LXC,
>> and have cc'd Steve Langasek for his comments on the relationship
>> between the two, and the plan to make them work well together.
> 
> I'm not sure if I have the full context for this discussion.  Is
> there a specific bug report against upstart that has not been
> handled?  My current understanding is that upstart works correctly
> both under LXC and in a chroot, and I don't see any open upstart
> bug reports about problems in this regard.
> 
> Indeed, I know that Stéphane has been working on getting
> cross-installation of upstart working, so that he can run emulated
> *ARM* systems in LXC containers on x86.  So I'd say that upstart is
> working pretty well with LXC, all in all; and there are no current
> plans for further development in this area, because as I said there
> don't appear to be any outstanding issues.
> 
> Cheers,

Indeed, upstart works perfectly in an LXC container, the only
limitation we currently have are related to two problems:

1) Figuring out when a container is shutting down and rebooting
2) Dealing with the slightly different list of mount points at boot time

The first will be solved by a kernel fix that we worked on at our LXC
sprint in Austin and that should be in Precise's kernel.

The second will be solved by having mountall be LXC aware or by fixing
the need for a different list of mount points, I have an action item
for that one.

Currently these two problems are worked around by the "lxcguest"
package that we install by default in all LXC containers we create.
Our goal for Precise is to remove the need for that package completely
and so be able to take any Ubuntu system and run them without any
change in a container.


For anyone who needs to do LXC specific actions, we also ship a tool
called "lxc-is-container" that will return 0 if you are in a container.
LXC also passes a container=lxc environment variable to upstart that
can be used by upstart jobs to behave differently when run in a container.


As Steve mentioned, I'm also working on getting containers in a
non-native architecture to work on LXC. Letting you boot an armel or
powerpc container on your x86 system, this is currently blocked on
some APT bugs that should be solved with the next release of APT (test
packages are available in my experimental PPA).

- -- 
Stéphane Graber
Ubuntu developer
http://www.ubuntu.com
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iQIcBAEBCgAGBQJOxSX/AAoJEMY4l01keS1nLekQAJY0gvD1zJXttzmtp202ayal
i20Ftiwyw+aLjDWAicD1D4C6gQZfk/7zHVrSJyo7noGQW1ATiWhmJ/EfPWqyXY26
qlwJCuSd42hwDyqDU/8bqTFZW9mUcPcgFCaEOSVhkdLYdqHDoqflhOJ7ehvnWV/O
QuJdLg1hOco2s/VcAqL1RBNDlR2iUtp3uvHmGfOD/ieIKOMe14eSgiNK9DqM6Xs/
ntzELBED/NrrQI46b5RNRPYRix9SFWMRNae8OLUDDbDRXg8p4tK1ApbN0TieSIC2
p22oR99TU5SnkI0+KuBV+d+HzQzdH+aF7qiZ8QRr0nYQGNWwhVRy76HWf7z0u/rE
oyypV2NN4Q5IUWsfsn54pk46TC0xBAvhXc5jUK8WnfbON8GNs8JqA24mQDKIgxeX
E7/+2Fe8HvogHWvZ0OLWw9oy8+sadT7X5OoXjRHBrj8MwMHzlMk+cQTryXBnHaqs
8ZQOXGox8W5h4Xo80v59RMf4HsYOXO1IW5ATx9xTU5BUTXzRA4ppZz5kRCk+LWHV
ZaxgOCVS6JRbYMTxXDZNjL2BFMX34DoGmNNoXmHVyJL6wf4WH1iXQi4oI3HUlQCe
Lm4BVj8Um4SLwvtcgV/Aiosi7niPP0mmV3K9PcyOeVU62AsZTY++fhgyl5EtVp8V
zvlaidnl3pze5OJ2fHbJ
=f7I2
-----END PGP SIGNATURE-----



More information about the technical-board mailing list