Re: NoCloud cloud-init datasource

On Tue, Feb 9, 2016 at 9:55 PM, Cliff Rowley <cliffrowley@gmail.com> wrote:

Sorry I'm not following. I had a poke around the /etc/cloud/* of an instance to see what was there and notice that in one of the files there is a list of enabled datasources. Is this what you're referring to? As far as I'm aware (I could be completely wrong, I haven't delved too far deep into the inner workings of all of this yet - though i have read pertinent bits of the cloud-init source) the datasources can (or do?) work in series so that for example if no NoCloud drive is mounted then cloud-init will go to the next datasource. At least that's how I've seen it work on all the other distributions I've tried this with. Again, I could be completely wrong and have entirely misunderstood.

The former is right - its a list of data sources; which ideally enabling a source in a cloud image for a specific cloud that that image is not for is probably incorrect although most of the time you just see a large delay in boot or in reboot (or both) as it times out after long period of trying to hit that 169.foo. The issue sometimes is that the image is not specific to a cloud but a hypervisor or can work independent of hypervisor due to the support in the kernel (its more about the image format) - the problem there how do you cater for this unless providing a slightly different image per cloud/virtual env?

In terms of the latter, I'm not quite sure the process with our it works etc. but I assume it works with the relevant modules for each data source in each phase in kind of like a compilation.

With all that said though we still don't have a link or ref to the particular image you are referring to :)

—
Cliff Rowley
Thinker, Tinkerer, Dork
@cliffrowley

On 9 Feb 2016, at 10:29, Chris Fordham <chris@fordham-nagy.id.au> wrote:

On Tue, Feb 9, 2016 at 9:10 PM, Cliff Rowley <cliffrowley@gmail.com> wrote:
Greetings, apologies if this isn't the right place or if this has been asked and dealt with before, but are there any plans to enable the NoCloud datasource in the published cloud images?

I'm not using OpenStack but I am enjoying the benefits of cloud images. So far, my Linux of choice (Debian, of course) is the only distribution that has been giving me a few surmountable yet frustrating issues running as a guest. The ConfigDrive datasource is ok, if a little fiddly, but even once configured the cloud-init system is still trying to make contact with my (nonexistent) metadata server and somehow still thinks its being managed by OpenStack. Conversely when using the NoCloud datasource with other distributions, configuration is done and finalised with no extra considerations.

For those of us who aren't using the whole OpenStack shebang, configuring an image with NoCloud is a simple `cloud-localds` away and it'd be great if the Debian cloud images supported this too..

—
Cliff Rowley
Thinker, Tinkerer, Dork
@cliffrowley

I think that really depends on which images we are talking about here. We should be enabling only the applicable datasources per image as datasource selection afaik can only be depicted by cloud.cfg or the included config dir (baked in).