On Thu, Jul 19, 2018 at 6:02 AM Raoul Scarazzini <ra...@redhat.com> wrote: [...]
> Small question aside related to all-in-one: we're talking about use > cases in which we might want to go from 1 to 3 controllers, but how this > can become a thing? I always thought to all-in-one as a developer/ci > "tool", so why we should care about giving the possibility to expand? > We have a few other use-cases but 2 of them are: - PoC deployed on the field, start with one controller, scale up to 3 controllers (with compute services deployed as well). - Edge Computing, where we could think of a controller being scaled-out as well, or a remote compute note being added, with VMs in HA with pacemaker. But I agree that the first target for now is to fulfil the developer use case, and PoC use case (on one node). This question is related also to the main topic of this thread: it was > proposed to replace Keepalived with anything (instead of Pacemaker), and > one of the outcomes was that this approach would not guarantee some of > the goals, like undercloud HA and keeping 1:1 structure between > undercloud and overcloud. But what else are we supposed to control with > Pacemaker on the undercloud apart from the IPs? > Nothing, AFIK. The VIPs were the only things we wanted to managed on a single-node undercloud. -- Emilien Macchi
__________________________________________________________________________ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
