Open Stack

On Fri, Oct 2, 2015 at 10:50 PM, Alex Schultz <aschultz at mirantis.com> wrote:

> So I was working on Bug 1493520 which is about what happens when a
> controller runs out of space. For this I came up with a solution[1] to
> leverage pacemaker to migrate services away from the controller when
> it runs out of space.
>

Yesterday, while we discuss about another bug, Dmitry Iliyn propose
following changes to haproxy service checker:
Checker for each service, covered by haproxy, should check also pacemaker
status on the target node.
If pacemaker not in the operate state (maintenance mode, fencing processed,
something another) Haproxy should mark
this endpoint as bad and do not use this endpoint.

IMHO this approach may be use and here.
As soon out of space detected, pacemaker should be switched to non-operate
mode.

If all openstack services will be controlled by pacemaker -- they will be
going down automatically by pacemaker.
Advantage of moving all openstack services into pacemaker -- one control
center for each service for whole cluster.
Services, controlled by pacemaker, may be just a OS services (init,
upstart, or something another) or an containered.

/sv
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20151003/4ab3d9f9/attachment.html>