On 7/15/2015 3:24 AM, Alex Xu wrote:
2015-07-15 5:14 GMT+08:00 Matt Riedemann <mrie...@linux.vnet.ibm.com <mailto:mrie...@linux.vnet.ibm.com>>: On 7/14/2015 3:43 PM, Cale Rath wrote: Hi, I created a patch to fail on the proxy call to Neutron for used limits, found here: https://review.openstack.org/#/c/199604/ This patch was done because of this: http://docs.openstack.org/developer/nova/project_scope.html?highlight=proxy#no-more-api-proxies, where it’s stated that Nova shouldn’t be proxying API calls. That said, Matt Riedemann brings up the point that this breaks the case where Neutron is installed and we want to be more graceful, rather than just raising an exception. Here are some options: 1. fail - (the code in the patch above) 2. proxy to neutron for floating ips and security groups - that's what the original change was doing back in havana 3. return -1 or something for floatingips/security groups to indicate that we don't know, you have to get those from neutron Does anybody have an opinion on which option we should do regarding API proxies in this case? Thanks, Cale Rath __________________________________________________________________________ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe <http://openstack-dev-requ...@lists.openstack.org?subject:unsubscribe> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev I prefer the proxy option, despite that we don't want to do more proxies to other services, it's the least of all evils here in my opinion. I don't think we can do #1, that breaks anyone using those APIs and is using Neutron, so it's a non-starter. agree #3 is an API change in semantics which would at least be a microversion and is kind of clunky. agree too~ For #2 we at least have the nova.network.base_api which we didn't have in Havana when I was originally working on this, that would abstract the neutron-specific cruft out of the nova-api code. The calls to neutron were pretty simple from what I remember - we could just resurrect the old patch: https://review.openstack.org/#/c/43822/ +1, but looks like this need new microversion also. It means after 2.x version, this api value is valid for neutron, before 2.x version, don't trust this api...
I'm not exactly clear on why we couldn't implement this as a bug fix for v2.0? I guess because of the standard reason we give for all microversions which is discoverability.
I guess in the v2.0 case we could just log the warning (option 4). It's not great, but at least it's a thing that an operator could find if they are using v2.0 and expecting proper quotas/limits values for security groups and floating IPs when using neutron but talking to the nova-api.
Another option is #4, we mark the bug as won't fix and we log a warning if neutron is configured saying some of the resources aren't going to be correct, use the neutron API to get information for quotas on security groups, floating IPs, etc. That's also kind of gross IMO, but it's an option. if we plan to deprecate network proxy api in no longer future, this is easy option. -- Thanks, Matt Riedemann __________________________________________________________________________ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe <http://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
-- Thanks, Matt Riedemann __________________________________________________________________________ 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