Re: [rdo-list] [tripleo] How to setup hyper-converged nodes (compute+ceph)?
On 05/18/2016 04:13 AM, Gerard Braad wrote:
Hi all,
Hereby updating the status of this message:
On Mon, May 16, 2016 at 9:55 AM, Gerard Braad <me(a)gbraad.nl> wrote:
> we would like
> to deploy Compute nodes with Ceph installed on them. This will
> probably be a change to the tripleo-heat-templates and the compute,
> and cephstorage resources
I noticed a review enabling the deployment of Ceph OSDs on the compute
node: https://review.openstack.org/#/c/273754/5
At the moment, it is marked as Workflow -1 due to possible
implementation of this feature by composable roles.
hi, that change is most probably working fine as-is but we are in the
process of migrating service defintions into more isolated roles and
that would supersede the existing submission
there is some WIP (changes Ie92b25a9c68a76b6d92abedef31e8039b16d9863 and
I1921115cb6218c7554348636c404245c79937673) to migrate ceph
mon/client/osd services into isolated roles but it's far from landing at
this point because we miss some additional machinery in the templates to
distribute roles on more
one way or another this should be possible in Newton though so I suggest
to keep tracking all three submission; we could fallback on the initial
submission should the ceph migration reveal to be problematic
--
Giulio Fidente
GPG KEY: 08D733BA | IRC: gfidente