Open Stack

To be clear the present OSAD project really has no intention to bring package based installations of OpenStack. We'd certainly not reject the idea and wouldn't mind having an implementation spec for it but all of our current tooling and design principles have been based on the fact that we've move away from distro packages and on to upstream source as it pertains to OpenStack. The system as it stands today creates an internal repository of built wheels for your environment and all of the OpenStack services are installed within LXC containers, where possible and it makes sense. The installation of these bits comes from the internal wheel repository and uses pip and all of the pre / post config happens within the Ansible playbooks.


One issue that will become a problem, for users of RedHat specifically, is the fact that RedHat has no LXC container templates (at least none that are publicly available) and even if someone were to make an official RedHat container template there'd be issues with the containers being able to connect to the satellite servers as well as other potential license problems.


I've done some experimenting with a RedHat 7.1 hosts and CentOS 7 containers and things seem to work OK but I'd not say that I have really put a lot of effort into it. That said, if its something that you'd all like to work on I'd be happy to help out to make it all go.


--

Kevin Carter
________________________________
From: Adam Young <ayoung at redhat.com>
Sent: Thursday, July 9, 2015 11:32 AM
To: Kris G. Lindgren; John Dewey
Cc: openstack-operators at lists.openstack.org
Subject: Re: [Openstack-operators] OSAD for RHEL

On 07/09/2015 02:16 AM, Kris G. Lindgren wrote:
Does OSP support running each service in an LXC container as well?  What about nova-cells? How does it handle people who need to carry local changes?  What is the upgrade path like with OSP?

So, ignoring the Hypervisor for the moment, there is no reason that the rest of the controllers can't run in separate Containers.  I think a container based deployment would be fantastic.

venv is not really sufficient, as the system level binaries can still conflict (MysQL and LDAP both require system libraries for Keystone, for example)

>From an Ansible perspective;  we need to  be able to share the HTTPD instance for Keystone and Apache, and getting that right will solve most of the issues deploying in a secure manner.  Putting Them on separate hosts or containers should be a degenerate case, and thus be supported, too.







Asking, because in Philly the general consensus, I fel,t was people want to move away from the current system level package stuff and move towards: venv's, "lightweight packages", containers.  The only reason that was brought up to keep packages around was to solve the non-python lib stuff and using a depsolver (yum/apt) that doesn't suck (pip).  So I am pretty sure my wants are inline with what other people in the community are either already doing or moving towards.
___________________________________________

Kris Lindgren
Senior Linux Systems Engineer
GoDaddy, LLC.


From: John Dewey <john at dewey.ws<mailto:john at dewey.ws>>
Date: Wednesday, July 8, 2015 at 11:43 PM
To: "Kris G. Lindgren" <klindgren at godaddy.com<mailto:klindgren at godaddy.com>>
Cc: Adam Young <ayoung at redhat.com<mailto:ayoung at redhat.com>>, "openstack-operators at lists.openstack.org<mailto:openstack-operators at lists.openstack.org>" <openstack-operators at lists.openstack.org<mailto:openstack-operators at lists.openstack.org>>
Subject: Re: [Openstack-operators] OSAD for RHEL

This would not be acceptable for those running OSP.


On Wednesday, July 8, 2015 at 10:12 PM, Kris G. Lindgren wrote:

I should be more clear. My current thought is to have a venv packaged
inside an rpm - so the rpm includes the needed init scripts, ensures the
required system level binaries are installed, adds the users - ect ect.
But would be a single deployable autonomous unit. Also, have a versioning
schema to roll forward and back between venvs for quick update/rollback.
We are already working on doing something similar to this to run kilo on
cent6 boxen, until we can finish revving the remaining parts of the fleet
to cent7.

My desire is to move away from using system level python & openstack
packages, so that I can possibly run mismatched versions if I need to. We
had a need to run kilo ceilometer and juno neutron/nova on a single
server. The conflicting python requirements between those made that task
impossible. In general I want to get away from treating Openstack as a
single system that everything needs to be upgraded in lock step (packages
force you into this). I want to move to being able to upgrade say
oslo.messaging to a newer version on just say nova on my control plane
servers. Or upgrade nova to kilo while keeping the rest of the system
(neutron) on juno. Unless I run each service in a vm/container or on a
physical piece of hardware that is pretty much impossible to do with
packages - outside of placing everything inside venv's.

However, it is my understanding that OSAD already builds its own
python-wheels and runs those inside lxc containers. So I don¹t really
follow what good throwing those into an rpm would really do?
____________________________________________
Kris Lindgren
Senior Linux Systems Engineer
GoDaddy, LLC.


On 7/8/15, 10:33 PM, "Adam Young" <ayoung at redhat.com<mailto:ayoung at redhat.com>> wrote:

On 07/07/2015 05:55 PM, Kris G. Lindgren wrote:
+1 on RHEL support. I have some interest in moving away from packages
and
am interested in the OSAD tooling as well.

I would not recommend an approach targetting RHEL that does not use
packages.

OSAD support for RHEL using packages would be an outstanding tool.

Which way are you planning on taking it?

____________________________________________
Kris Lindgren
Senior Linux Systems Engineer
GoDaddy, LLC.







On 7/7/15, 3:38 PM, "Abel Lopez" <alopgeek at gmail.com<mailto:alopgeek at gmail.com>> wrote:

Hey everyone,
I've started looking at osad, and I like much of the direction it
takes.
I'm pretty interested in developing it to run on RHEL, I just wanted to
check if anyone would be -2 opposed to that before I spend cycles on
it.

_______________________________________________
OpenStack-operators mailing list
OpenStack-operators at lists.openstack.org<mailto:OpenStack-operators at lists.openstack.org>
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators


_______________________________________________
OpenStack-operators mailing list
OpenStack-operators at lists.openstack.org<mailto:OpenStack-operators at lists.openstack.org>
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators


_______________________________________________
OpenStack-operators mailing list
OpenStack-operators at lists.openstack.org<mailto:OpenStack-operators at lists.openstack.org>
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20150710/499d1f87/attachment.html>