Open Stack

On 2015-05-28 22:45:37 +0000 (+0000), Fox, Kevin M wrote:
> You could pass the cache through with a volume....

Yeah, from the "what can we do with our current CI infrastructure?"
perspective, we would just need a way to identify what bits benefit
from being cached for these particular builds and then we would bake
them into our worker base images just like we do to pre-cache all
sorts of other resources used by all our jobs.

The trick is to first understand what mechanisms we already have in
the OpenStack CI to handle performance and isolation, and then try
not to propose new solutions that redundantly reimplement them.
Hopefully as we get more project infrastructure team members
involved with this we can weed out the unnecessary additional
complexities which are being suggested in the initial brainstorming
process.

> As for what docker buys you, it would allow a vm (or my desktop)
> running centos (as an example) to build packages for multiple
> distro's using the distro's own native tools. I see that as a
> plus.

I think the question was really not "what good is isolation?" but
rather, "how is a container any better than a chroot for this
purpose?"
-- 
Jeremy Stanley