Re: [openstack-dev] [heat][horizon] Backward-incompatible changes to the Neutron API

2015-08-27 Thread Paul Michali
Akihiro, can you look at the developer's reference I posted (191944), where
there is the overall API plan and a proposal for handling backward
compatibility.

Thanks!

Paul Michali (pc_m)

On Thu, Aug 27, 2015 at 11:12 AM Akihiro Motoki amot...@gmail.com wrote:

 As Mathias said, Horizon worked (and in many cases works) cross releases.

 Horizon determines supported features based on keystone catalogs,
 extension list from back-end services (like nova, neutron).
 Micro-versioning support may come in future (though it is not supported).

 For backward incompatible API change in VPNaaS, Horizon can determine
 if Neutron (including VPNaaS) provides a way to determines which version
 is available.
 At now, the only way is to expose it through the extension list.

 On the other hand, it is tough to maintain multiple versions of
 implementations.
 It is reasonable to me that Horizon supports two implementation in one or
 two
 release cycle(s) and drop older implementation later.

 Akihiro


 2015-08-27 16:29 GMT+09:00 Matthias Runge mru...@redhat.com:

 On 26/08/15 23:55, James Dempsey wrote:
  Greetings Heat/Horizon Devs,
 
  There is some talk about possibly backward-incompatible changes to the
  Neutron VPNaaS API and I'd like to better understand what that means for
  Heat and Horizon.
 
  It has been proposed to change Neutron VPNService objects such that they
  reference a new resource type called an Endpoint Group instead of
  simply a Subnet.
 
  Does this mean that any version of Heat/Horizon would only be able to
  support either the old or new Neutron API, or is there some way to allow
  a version of Heat/Horizon to support both?
 
 In the past, Horizon worked cross releases.

 The way horizon works is, it looks out for a networking endpoint in
 keystone catalog. We don't really care, if it's nova or neutron
 answering. The rest should be discoverable via API.
 Horizon uses neutronclient rather than directly talking to neutron by
 using its API interface.

 If you make it discoverable, and you'd add that to neutronclient,
 horizon could support both.

 Matthias




 __
 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


 __
 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

__
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


Re: [openstack-dev] [heat][horizon] Backward-incompatible changes to the Neutron API

2015-08-27 Thread Akihiro Motoki
As Mathias said, Horizon worked (and in many cases works) cross releases.

Horizon determines supported features based on keystone catalogs,
extension list from back-end services (like nova, neutron).
Micro-versioning support may come in future (though it is not supported).

For backward incompatible API change in VPNaaS, Horizon can determine
if Neutron (including VPNaaS) provides a way to determines which version is
available.
At now, the only way is to expose it through the extension list.

On the other hand, it is tough to maintain multiple versions of
implementations.
It is reasonable to me that Horizon supports two implementation in one or
two
release cycle(s) and drop older implementation later.

Akihiro


2015-08-27 16:29 GMT+09:00 Matthias Runge mru...@redhat.com:

 On 26/08/15 23:55, James Dempsey wrote:
  Greetings Heat/Horizon Devs,
 
  There is some talk about possibly backward-incompatible changes to the
  Neutron VPNaaS API and I'd like to better understand what that means for
  Heat and Horizon.
 
  It has been proposed to change Neutron VPNService objects such that they
  reference a new resource type called an Endpoint Group instead of
  simply a Subnet.
 
  Does this mean that any version of Heat/Horizon would only be able to
  support either the old or new Neutron API, or is there some way to allow
  a version of Heat/Horizon to support both?
 
 In the past, Horizon worked cross releases.

 The way horizon works is, it looks out for a networking endpoint in
 keystone catalog. We don't really care, if it's nova or neutron
 answering. The rest should be discoverable via API.
 Horizon uses neutronclient rather than directly talking to neutron by
 using its API interface.

 If you make it discoverable, and you'd add that to neutronclient,
 horizon could support both.

 Matthias




 __
 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

__
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


Re: [openstack-dev] [heat][horizon] Backward-incompatible changes to the Neutron API

2015-08-27 Thread Matthias Runge
On 26/08/15 23:55, James Dempsey wrote:
 Greetings Heat/Horizon Devs,
 
 There is some talk about possibly backward-incompatible changes to the
 Neutron VPNaaS API and I'd like to better understand what that means for
 Heat and Horizon.
 
 It has been proposed to change Neutron VPNService objects such that they
 reference a new resource type called an Endpoint Group instead of
 simply a Subnet.
 
 Does this mean that any version of Heat/Horizon would only be able to
 support either the old or new Neutron API, or is there some way to allow
 a version of Heat/Horizon to support both?
 
In the past, Horizon worked cross releases.

The way horizon works is, it looks out for a networking endpoint in
keystone catalog. We don't really care, if it's nova or neutron
answering. The rest should be discoverable via API.
Horizon uses neutronclient rather than directly talking to neutron by
using its API interface.

If you make it discoverable, and you'd add that to neutronclient,
horizon could support both.

Matthias




__
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


Re: [openstack-dev] [heat][horizon] Backward-incompatible changes to the Neutron API

2015-08-26 Thread Paul Michali
For details on the API/resource change, see the developer's reference
document that is under review: https://review.openstack.org/#/c/191944/

There is a section at the end, where I'm proposing a possible way to
support both versions of the API and provide backward compatibility. Feel
free to comment on the review.

Regards,

Paul Michali (pc_m)


On Wed, Aug 26, 2015 at 6:10 PM James Dempsey jam...@catalyst.net.nz
wrote:

 Greetings Heat/Horizon Devs,

 There is some talk about possibly backward-incompatible changes to the
 Neutron VPNaaS API and I'd like to better understand what that means for
 Heat and Horizon.

 It has been proposed to change Neutron VPNService objects such that they
 reference a new resource type called an Endpoint Group instead of
 simply a Subnet.

 Does this mean that any version of Heat/Horizon would only be able to
 support either the old or new Neutron API, or is there some way to allow
 a version of Heat/Horizon to support both?


 Thanks,
 James

 --
 James Dempsey
 Senior Cloud Engineer
 Catalyst IT Limited
 +64 4 803 2264
 --

 __
 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

__
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