[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Adding cloud-init in the next Wheezy point release



On 05/07/2013 08:16 PM, Adam D. Barratt wrote:
> On 2013-05-06 17:19, Thomas Goirand wrote:
>> In this thread:
>> https://lists.debian.org/debian-cloud/2013/05/msg00002.html
>>
>> we have been discussing how we could have cloud-init in the Debian cloud
>> images.
>>
>> One solution is to enable wheezy-backports by default in the images.
>> Though there are some concerns that we shouldn't do that, as this isn't
>> the default in Debian right now.
>
> If it's just to pull the package in during image build, is that a
> particular problem?

The problem that was raised is that some users might want to do security
updates, which then wouldn't be possible.

A typical workflow would be:
- A user uses a pre-build Debian official image, then customizes it
- He uses the cloud, then decides to upgrade his image
- He then makes a snapshot of the image using the cloud API

If we don't provide a way so that cloud-init is also upgraded, then
we have a problem (I'm not saying there will be a problem on
cloud-init, but we should always prepare for the worst).

>
>> The other solution would be to add cloud-init in the next point release
>> of Wheezy. We all know that there's some strong rules that we shouldn't
>> add new things in the stable distribution, even more after the freeze.
>
> I assume you mean after the release? It's a little late to worry about
> being after the freeze.

A "than" is missing. I meant to write "even more than after the freeze".

> If cloud-init is so mission-critical, why was this never noticed or
> raised *before* the release?

It was. But it took some time to get it into SID. According to the
PTS, Charles Plessy uploaded it to Experimental in the 23rd of
June. Then Julien Danjou uploaded it into SID last January, but it
stayed in the NEW queue for a very long time.

The main problem is that the interest for the cloud and building
an official Debian cloud image is fairly recent. There is more and
more people (including DDs) interested in it, and not having
cloud-init in main is quite annoying for those of us who do desire
to build cloud images.

Anyway, I don't think that digging in the history of the package
will help solving anything, so let's stop! :)

We (at least Julien and I) thought it would be ok to use backports,
but on the 2nd thought, it doesn't seem to be the best solution.
I already explained that activating backports by default isn't
desirable: since that isn't the default for Wheezy. The other
argument is that cloud-init would be the only package which
would be needed from backports. Which is why I wonder what the
release team opinion was, and if we could do something about it.

If there is a strong *no* from the release team, i guess we could
live with the backports solution, though it is my opinion that it
would be best if we could make an exception in this case.
If staging cloud-init in backports for a few months is one of the
things who can ease the decision of the release team, I think
it could be a good temporary solution as well. If there are other
things which could ease the decision, let us know. Waiting for a
full release cycle (eg: 2 years) to get cloud-init in main would
IMO not be desirable.

Thomas Goirand (zigo)


Reply to: