Open Stack

Hi,

The hack-ish way is to check if the current master has a different ip than
the
swarm_api_api and based on that decide whether to swarm init or join. The
proper way is to have two resource groups (as you said) one for the primary
master and one for the secondary masters. This requires some plumping
though.

We decided two have a _v2 driver in /contrib initially. I have a prototype
working
based on fedora-25 (docker 1.12.6). I can push and work together on it, if
you
want.

Spyros

On 31 January 2017 at 20:52, Kevin Lefevre <lefevre.kevin at gmail.com> wrote:

> On Tue, 2017-01-31 at 17:01 +0100, Spyros Trigazis wrote:
> > Hi.
> >
> > I have done it by checking the ip address of the master. The current
> > state of
> > the heat drivers doesn't allow the distinction between master > 1 or
> > master=1.
> >
>
> Please, could you elaborate on this ?
>
> Also what is your opinion about starting a new swarm driver for swarm
> mode ?
>
> > Spyros
> >
> >
> >
> > On 31 January 2017 at 16:33, Kevin Lefevre <lefevre.kevin at gmail.com>
> > wrote:
> > > Hi, Docker 1.13 has been released with several improvements that
> > > brings
> > > swarm mode principles closer to Kubernetes such as docker-compose
> > > service swarm mode.
> > >
> > > I'd like to implement a v2 swarm template. I don't know if it's
> > > already
> > > been discussed.
> > >
> > > Swarm mode is a bit different but a lot simpler to deploy than
> > > Swarm
> > > Legacy.
> > >
> > > In Kubernetes you can deploy multiples masters at the same time but
> > > in
> > > swarm mode you have to:
> > > - bootstrap a first docker node
> > > - run docker swarm init
> > > - get a token (worker or manager)
> > > - bootstrap other worker
> > > - use manager or worker token depending manager count.
> > >
> > > I don't know what is the best way to do so in HEAT. I'm sure there
> > > are
> > > multiple options (I'm not an expert in HEAT i don't know if they
> > > are
> > > feasible) :
> > >
> > > - Bootstrap a first server
> > > - Wait for it to ready, run docker swarm init, get both manager and
> > > worker tokens
> > > - if manager count >1, we can bootstrap another resource group for
> > > extra managers which will use a manager token.
> > > - Bootstrap the rest of the worker and use a worker token.
> > >
> > > The difficulty is to handle multiples master properly, i'd like to
> > > hear
> > > your ideas about that.
> > >
> > >
> > > --
> > > Kevin Lefevre
> > >
> > > ___________________________________________________________________
> > > _______
> > > OpenStack Development Mailing List (not for usage questions)
> > > Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsu
> > > bscribe
> > > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
> > >
> >
> > _____________________________________________________________________
> > _____
> > OpenStack Development Mailing List (not for usage questions)
> > Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubs
> > cribe
> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
> --
> Kevin Lefevre
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20170131/d1f90ef9/attachment.html>