On 02/07/2016 08:29 AM, Troy Dawson wrote:
On Sun, Feb 7, 2016 at 3:22 AM, Karanbir Singh kbsingh@centos.org wrote:
hi,
Yesterday a few of us met for a face to face walk through of the CentOS Container Pipeline we've been talking about. The aim was to rescope the upstream projects we can lean on, share code with, and help - and then find the glue pieces that could bring this code base together.
The larger picture effectively boils down to :
- Find the components needed to track code git repos ( either in
git.centos.org or elsewhere )
- Find the components needed to now build that code, containerise the
code, push it through a test process, and then deliver the containers either locally into a centos transitional container registry, or to a CDN like wider centos registry or if the user so desires, to a third part registry ( provided a good process can be find to handle the credentials needed ).
- We'd want to use this pipeline both internally, for the CentOS Linux
components ( eg. a LAMP container from CentOS Linux 7 ), for SIG components via cbs output ( eg. SCLo SIG folks shipping containers for their content ), as well as open this up via a trivial UI, for anyone in the community who'd like to come and consume this pipeline.
the key piece that we didnt have clarity on was the orchestration and glue that could bind the various software components. Vaclav Pavlin, brought up that we might be able to use openshift templates in order to get the job runs done - if the input for the templates could be derived from the cccp-index, either via the jjb work already done, or writing a new filter into JJB, we might be able to execute a fairly scaleable solution without needing to own any piece of the over all code.
Additionally, there is a fair interest from the Fedora team ( ie Adam! ) working on the same problem in their space, mostly consuming the identical code stack, within and for their infra, their constraints and aims.
Over the coming days we are going to try and work though PoC's, get some infra setup and trial running through some of the user stories we want to execute on.
Finally, a quick shout out to everyone for coming together at pretty much the last minute adhoc meeting - Fabian Arrotin, Brian Stinson, Christoph Goern, Vaclav Pavlin, Aaron Weitekamp, Tomas Tomecek, Adam Miller, Dusty Mabe, Honza Horak, Radek Vokal as well as Tim Waugh, Bama Charan and Mohammed Zeeshan for dialing into the meeting.
And to everyone else, want to build containers with us ? come talk to us
- we'd love to make sure we include as many user stories to get great
scope before we start implementing bits.
regards
What do you mean by "containers"? When I hear that word everything from docker and rocket images, to virtual machine images, and more, come to mind.
Well, I think that depends on the group who wants the container.
For now, I would think we would orchestrate docker containers that work with atomic and containers that work in openshift-origin as starting points.
But I would also think that SIGs can support other container types as well, if they choose to.
Very fluid at this point, I think.