-Re: Boxgrinder is not actively developed anymore
Sean Mackrory 2013-03-17, 22:02
>> Why don't you describe your ideas/code/efforts? They sound great and
would be delighted to learn more about them. The conversation would benefit
a lot from it.
Certainly! So I created some JSON files that roughly follow the same
structure as Boxgrinder YAML. Each one represents a specific set of
functionality, and resides in a directory with any files it will pull into
the appliance. They can list other JSON files as dependencies, and the
python script will merge dictionaries and concatenate arrays (with
descendants overriding ancestors), then write it out to a single YAML file
and a directory of the other files. It's pretty hacky at this point (having
just been a side-project done as-needed), but by generating the manifest
gives a lot of flexibility to add little hacks here and there, and it was
easier to manage multiple appliances than with Boxgrinder's inheritance
alone (I've forgotten the specific shortcomings, but it wasn't sufficient
for my purposes). I've added code to generate the .mozilla/firefox
directory with a list of bookmarks, too.
|---hbase (depends on HDFS)
Going forward, I'd like to make it more modular to support different
back-ends (perhaps kickstarter, for Live CDs, etc), and perhaps break down
the JSON files / directories further so that you can define the individual
Bigtop components and combine them however you want into. I'm putting
together a "bigtop" example for my current mechanism and will post it.
On Sun, Mar 17, 2013 at 1:33 PM, Bruno Mahé <[EMAIL PROTECTED]> wrote:
> See reply inline.
> On 03/17/2013 11:16 AM, Konstantin Boudnik wrote:
>> Oh Bruno - I am sorry if it felt too personal on you. Please accept my
>> apologies. I will make an effort not to talk about any R/C or F distros
>> in the
> I believe you missed my point.
> My point is: if not welcoming, we should have at least a civil and
> courteous community/mailing-list.
> As of my feelings, don't worry, they were not touched :)
> But are people from red hat/fedora/openstack supposed to feel welcomed
> after you just trashed them?
> However, my comment was made about this line from your original email:
>>> xml files instead of yaml. It would fit the requirements, except that
>>> for now, the documentation is pretty sparse and/or outdated. I also
>> which totally fills like a colledge kids project. If I am mistaken - I
>> take my
>> non-R hat off to these guys.
> Let's wait to have more information to make such judgment.
> As for
>>> I am not convinced by vmbuilder. Its only community is ubuntu.
>>> As expected it only handles ubuntu/debian distros, which is fine.
>> a simple http://is.gd/AmLkQe shows that not only .deb distros are
>> Pretty much all modern virtualization platforms are supported as well.
> This is pointing to a lonely github repository not updated in a year :/
> But even this is besides the point. Perhaps, we won't found one-fits-all
>> for VM creation - it's not a big deal as we can use two or more if
>> needed. I
>> don't see this is a huge problem.
> Agreed. It is not an either or type of choice. But if we can avoid
> duplication of effort, it would be great.
> And we won't know if such one-fits-all exist if we don't look for it :)
> As for
>>> interesting. I would hate to look into moderation of mailing lists
>>> at the Apache Foundation.
>> I beg you not to do this, because the discussions like this are so much
> But still, the first recommendation for mailing list on the Apache
> Foundation website is "Respectful and considerate communities are one of
> the pillars of the Apache way. Please aim to provide constructive comments
> and do not denigrate others."