On 3/2/19 1:29 PM, Neal Gompa wrote:
On Sat, Mar 2, 2019 at 4:17 PM Kevin Fenzi <kevin(a)scrye.com>
> On 3/1/19 4:36 AM, Miroslav Suchý wrote:
>> Dne 27. 02. 19 v 23:00 Peter Robinson napsal(a):
>>>> Hey everyone.
>>>> As you know, we currently have a RHOSP5 ancient cloud. After a bunch of
>>>> work last year, we got a RHOSP13 cloud up and mostly working, but it was
>>>> a ton of work. After hearing from the Fedora Council and our various
>>>> management chains we determined that it wouldn't really be a good use
>>>> our time moving forward to keep maintaining a OpenStack cloud.
>>>> We have not yet determined what we want to do with the hardware that we
>>>> had allocated to this, but are weighing our options. We may want to
>>>> setup OpenShift bare nodes so we can do kubevirt, we may want to just
>>>> setup a normal virthost setup managed by ansible.
>>>> For the items currently in our cloud, we will be looking at options for
>>>> them, we are definitely not shutting things off until we have plans in
>>>> Happy to answer any questions and will make sure everything is properly
>>> Do we have a documented list of services running on that cloud instance?
>> From inventory/inventory:
> Thats just what is in ansible, there are other hosts also (although
> those should all be tranistory (copr builders or scratch hosts or test
> hosts, etc).
> As for timeline, I was moving last week, so I should be able to get back
> to it this coming week, especially since we should be in freeze soon.
> Since my current plan is to make the aarch64 and ppc64le hosts
> virthosts, I am going to try and get that done first so they can be used
> before anything else needs to be done. There's some networking changes
> needed, but hopefully in the next few weeks those will be online and
> copr can look at just using them from it's existing old cloud nodes.
> Then we can move on to the rest of the setup.
Could we consider using RHEV here? That would give us some way of
doing API-based provisioning without requiring anywhere close to the
complexity of RHOSP and still let us have some niceties like live
migration across nodes as needed.
We looked at RHEV, but it only provided network isolation if you use
openstack-neutron with it. There were some other things we couldn't get
it to do easily also (but I can't recall the details, relrod evaluated it).