[Container-tools] Recommendations for bootstrapping a local k8s cluster?

[Nick Coghlan]

On 22 May 2015 at 03:26, James Shubin <purpleidea at redhat.com> wrote:
> On Thu, 2015-05-21 at 15:34 +1000, Nick Coghlan wrote:
>> Following on from my initial experimentation with helloapache, it
>> quickly became clear to me that a local k8s cluster is going to be
>> essential for any serious exploration of nulecule's capabilities (with
>> my main aim being to see how much of
>> https://beaker-project.org/docs-develop/in-a-box/ nulecule can already
>> handle, and what would be needed to get it the rest of the way).
>>
>> James has some Oh My Vagrant instructions at
>> https://ttboj.wordpress.com/2015/05/02/kubernetes-clusters-with-oh-my-vagrant/
>>
>> k8s upstream has some Vagrant based instructions at
>> https://github.com/GoogleCloudPlatform/kubernetes/blob/master/docs/getting-started-guides/vagrant.md
>>
>> Neither of those seem particularly appealing, so if there's a way to
>> make it work I'd like something along the lines of:
> I'd love to hear some specifics of what you don't find appealing about
> Oh-My-Vagrant. If there are legitimate issues, we can fix them :)

Mainly the fact I can't install it with yum/dnf or docker/atomic - as
far as I can tell, OMV is designed to be run specifically from a
source checkout rather than using a published release artifact.

The superprivileged container approach is appealing, as it's almost as
light on the system being configured as curl+sh, but uses a payload
delivery mechanism that is both more capable (entire container, rather
than just a shell script), and offers more traceability and security
options (since you're retrieving the image via Docker rather than raw
HTTPS).

*Inside* the SPC, it would make sense to use OMV and/or the Ansible
scripts to do the actual configuration.

Cheers,
Nick.

-- 
Nick Coghlan   |   ncoghlan at gmail.com   |   Brisbane, Australia