[openstack-dev] [tripleo] FFE request for Ceph RGW integration

Giulio Fidente gfidente at redhat.com
Thu Sep 1 00:11:05 UTC 2016


On 08/30/2016 10:50 PM, Steven Hardy wrote:
> On Tue, Aug 30, 2016 at 03:25:30PM -0400, Emilien Macchi wrote:
>> Here's my 2 cents:
>>
>> The patch in puppet-ceph has been here for long time now and it still
>> doesn't work (recent update of today, puppet-ceph is not idempotent
>> when deploying RGW service. It must be fixed in order to get
>> successful deployment).
>> Puppet CI is still not gating on Ceph RGW (scenario004 still in
>> progress and really low progress to make it working recently).
>
> This does sound concerning, Giulio, can you provide any feedback on work
> in-progress or planned to improve this?

we invested quite some time today testing and updating the patches as needed

I've a got a successful deployment where by just adding the Member role 
to my user I could use the regular swiftclient to operate against RadosGW

This is by pulling in:

https://review.openstack.org/#/c/347956/
https://review.openstack.org/#/c/363164/

https://review.openstack.org/#/c/334081/ (and its dependencies)

https://review.openstack.org/#/c/289027/

Emilien can you re-evaluate the status of the puppet-ceph and 
puppet-tripleo submissions?

>> My opinion says we should not push to have it in Newton. Work to do it
>> was not extremely pushed during the cycle I see zero reason to push
>> for it now the cycle is ending.

agreed, this might not have been pushed much during the cycle as other 
priorities needed attention too but it seems to be an interesting 
feature for those deploying Ceph and in decent state; also as per 
Steven's comment below, it'll be optional in TripleO, we'll continue to 
deploy Swift by default so it's not going to have a great impact on 
other existing work

> I agree this is being proposed too late, but given it will be disabled by
> default that does mitigate the risk somewhat.
>
> Giulio - can you confirm this will just be a new service template and
> puppet profile, and that it's not likely to require rework outside of the
> composable services interface?  If so I'm inclined to say OK even if we
> know the puppet module needs work.

no rework of the composable services interface will be needed, the tht 
submission is, in addition to adding the new service template, adding an 
output to the endpoint map for the new service, the puppet submission is 
adding a new role

https://review.openstack.org/#/c/289027/

-- 
Giulio Fidente
GPG KEY: 08D733BA | IRC: gfidente



More information about the OpenStack-dev mailing list