Open Stack

On Thu, Aug 09, 2018 at 08:00:13PM -0600, Wesley Hayutin wrote:
> On Thu, Aug 9, 2018 at 5:33 PM Alex Schultz <aschultz at redhat.com> wrote:
> 
> > On Thu, Aug 9, 2018 at 2:56 PM, Doug Hellmann <doug at doughellmann.com>
> > wrote:
> > > Excerpts from Alex Schultz's message of 2018-08-09 14:31:34 -0600:
> > >> Ahoy folks,
> > >>
> > >> I think it's time we come up with some basic rules/patterns on where
> > >> code lands when it comes to OpenStack related Ansible roles and as we
> > >> convert/export things. There was a recent proposal to create an
> > >> ansible-role-tempest[0] that would take what we use in
> > >> tripleo-quickstart-extras[1] and separate it for re-usability by
> > >> others.   So it was asked if we could work with the openstack-ansible
> > >> team and leverage the existing openstack-ansible-os_tempest[2].  It
> > >> turns out we have a few more already existing roles laying around as
> > >> well[3][4].
> > >>
> > >> What I would like to propose is that we as a community come together
> > >> to agree on specific patterns so that we can leverage the same roles
> > >> for some of the core configuration/deployment functionality while
> > >> still allowing for specific project specific customization.  What I've
> > >> noticed between all the project is that we have a few specific core
> > >> pieces of functionality that needs to be handled (or skipped as it may
> > >> be) for each service being deployed.
> > >>
> > >> 1) software installation
> > >> 2) configuration management
> > >> 3) service management
> > >> 4) misc service actions
> > >>
> > >> Depending on which flavor of the deployment you're using, the content
> > >> of each of these may be different.  Just about the only thing that is
> > >> shared between them all would be the configuration management part.
> > >
> > > Does that make the 4 things separate roles, then? Isn't the role
> > > usually the unit of sharing between playbooks?
> > >
> >
> > It can be, but it doesn't have to be.  The problem comes in with the
> > granularity at which you are defining the concept of the overall
> > action.  If you want a role to encompass all that is "nova", you could
> > have a single nova role that you invoke 5 different times to do the
> > different actions during the overall deployment. Or you could create a
> > role for nova-install, nova-config, nova-service, nova-cells, etc etc.
> > I think splitting them out into their own role is a bit too much in
> > terms of management.   In my particular openstack-ansible is already
> > creating a role to manage "nova".  So is there a way that I can
> > leverage part of their process within mine without having to duplicate
> > it.  You can pull in the task files themselves from a different so
> > technically I think you could define a ansible-role-tripleo-nova that
> > does some include_tasks: ../../os_nova/tasks/install.yaml but then
> > we'd have to duplicate the variables in our playbook rather than
> > invoking a role with some parameters.
> >
> > IMHO this structure is an issue with the general sharing concepts of
> > roles/tasks within ansible.  It's not really well defined and there's
> > not really a concept of inheritance so I can't really extend your
> > tasks with mine in more of a programming sense. I have to duplicate it
> > or do something like include a specific task file from another role.
> > Since I can't really extend a role in the traditional OO programing
> > sense, I would like to figure out how I can leverage only part of it.
> > This can be done by establishing ansible variables to trigger specific
> > actions or just actually including the raw tasks themselves.   Either
> > of these concepts needs some sort of contract to be established to the
> > other won't get broken.   We had this in puppet via parameters which
> > are checked, there isn't really a similar concept in ansible so it
> > seems that we need to agree on some community established rules.
> >
> > For tripleo, I would like to just invoke the os_nova role and pass in
> > like install: false, service: false, config_dir:
> > /my/special/location/, config_data: {...} and it spit out the configs.
> > Then my roles would actually leverage these via containers/etc.  Of
> > course most of this goes away if we had a unified (not file based)
> > configuration method across all services (openstack and non-openstack)
> > but we don't. :D
> >
> 
> I like your idea here Alex.
> So having a role for each of these steps is too much management I agree,
> however
> establishing a pattern of using tasks for each step may be a really good
> way to cleanly handle this.
> 
> Are you saying something like the following?
> 
> openstack-nova-role/
> * * /tasks/
> * * /tasks/install.yml
> * * /tasks/service.yml
> * */tasks/config.yml
> * */taks/main.yml
> ---------------------------
> # main.yml
> 
> include: install.yml
> when: nova_install|bool
> 
> include: service.yml
> when: nova_service|bool
> 
> include: config.yml
> when: nova_config.yml
> --------------------------------------------------
> 
> Interested in anything other than tags :)
> Thanks
> 
This is basically what I do with roles i write, allow the user to decide to step
over specific tasks. For example, I have created nodepool_task_manager variable
with the following:

    http://git.openstack.org/cgit/openstack/ansible-role-nodepool/tree/defaults/main.yaml#n16
    http://git.openstack.org/cgit/openstack/ansible-role-nodepool/tree/tasks/main.yaml

Been using it for a few years now, works much better then tags for me.  The
phases are pre, install, configure, service right now.

- Paul