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

Re: Hosting & deploying OpenShift

I'll also note that I've got a talk at RH Summit this year that will be discussing
some of the nuts and bolts of how the Openshift PaaS is put together, along
with some of the things the operations team does to keep everything functioning

So, I'll be providing some of that information you're interested in in less
than two months. After Summit, my slides and notes will be publicly
available, so I'll cross-post them here.


On 03/05/12 09:05 -0700, Dr Nic Williams wrote:
  Matt, thanks for the answer (I hadn't seen tito before, and hopefully
  alleviates my "I've never been a RHEL person; how do I make packages?!"
  thoughts). So you package up OpenShift source code (crankcase, etc) into
  RPMs and then install them onto deployment VMs via the packaging system?
  I'll ask my question from a different direction to flesh out the other
  part of the question.
  RedHat created/sponsors BoxGrinder & DeltaCloud - to create VM images and
  to provision them on IaaS. These concepts don't appear in OpenShift. Are
  you using them to provision your own VMs for your hosted OpenShift?
  I assume that your hosted OpenShift, and others who run OpenShift on their
  own infrastructure (powered by the might of the $1B RHEL products, no
  doubt!), will have OpenShift running across a cluster of machines/VMs. A
  cluster that will grow with time, and all who's VMs/machines will be
  replaced/upgraded etc. Are you using a system built on
  BoxGrinder/DeltaCloud or something else?
  Dr Nic Williams - VP Developer Evangelism
  Engine Yard
  The Leading Platform as a Service
  Mobile: +1 415 860 2185
  Skype: nicwilliams
  Twitter: @drnic

  On Thursday, May 3, 2012 at 8:54 AM, Matt Hicks wrote:

    I'm not an expert on BOSH, but I'll try and give somewhat of a breakdown
    to how we do similar things.  We have tried to avoid building new
    tooling where we could and use the existing ecosystem where possible.

    First, I'll cover releases.  We depend heavily on RPM's for installation
    and dependency management and git for the code.  To weave the two
    together and control tagging, releasing, etc we use a project called
    tito - [1]https://github.com/dgoodwin/tito.  Tito has provided a very
    elegant way to be able to provide a consistent process between git and
    RPM.  I can go into some more details about why we use RPM and just gems
    if you are interested too (I know that's not a common thing in the Ruby

    For development, we do have some scripts that will sync a development
    repository to a machine and use tito to install the contents that we
    will have to open source.  It's a pretty simple process but we might as
    well open up what we have.

    The concept of giving VM's jobs / roles is simply covered by which RPM's
    are installed on that VM.  OpenShift uses the concept of cartridges to
    encapsulate the functionality that a VM can provide.  For example, if
    you install the JBoss cartridge on a node, it will be able to service
    requests for JBoss applications.  You might install multiple cartridges
    or selectively install them.

    Because we are very RPM centric and use standard configuration
    locations, we utilize puppet in production to pick what RPM's and config
    files get installed on the various VM's.

    For monitoring we use nagios and collectd.

    All that said, we didn't want to be too prescriptive about the tools
    that other people use to start.  Our goal has been to stay very standard
    with the installation and approach so that people could use CFEngine,
    Chef, or custom scripts to manage their infrastructure.  Having spent
    some time in the ops world, I realize there are a lot of different
    approaches that work there.

    Hope this helps!


    On 05/03/2012 11:32 AM, Dr Nic Williams wrote:

      RedHatters, what tools/infrastructure are you using for your own
      hosting/deployment of hosted OpenShift?
      As an example, the CloudFoundry team had to build BOSH in order to
      host & deploy their own [2]CloudFoundry.com (and all their personal &
      staging versions).
      Dr Nic Williams - VP Developer Evangelism
      Engine Yard
      The Leading Platform as a Service
      Mobile: +1 415 860 2185
      Skype: nicwilliams
      Twitter: @drnic

Attachment: pgpekXnSfJlo6.pgp
Description: PGP signature

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