> BTW, attached is a patch that addresses a few URL conversion problems > I > found in the mock driver; feel free to commit if it doesn't make things > worse ;)
ACK & pushed - I've confirmed things seem to not have gotten worse :) Cheers, Dies Koper > -----Original Message----- > From: David Lutterkort [mailto:lut...@redhat.com] > Sent: Tuesday, 2 April 2013 11:27 AM > To: dev@deltacloud.apache.org > Cc: mfoj...@redhat.com > Subject: Re: [PATCH] added Template subcollections to SystemTemplate > > On Thu, 2013-03-28 at 12:19 +1100, Koper, Dies wrote: > > A bit of background on this patch: > > It contains my latest work on implementing cimi systems and system > > templates for mock and fgcp. > > It is not complete and contains some debugging statements. Its main > > purpose is to show you what I have and what issues I'm seeing. > > > > It applies to master (fcffad13e66175152cf8a43d615b79727902b5ee) and > > requires "[PATCH] CIMI schema: tolerate nil hash_map attributes" to > > solve a nil error for system templates with fgcp. > > > > Issues I'm having: > > > > (1) > > With mock, even unsupported subcollections are shown when retrieving > > systems. With fgcp they are also shown, and their href urls are broken: > > This is a shortcoming of how we populate models right now - there's no > way to suppress collections that the driver doesn't support. In fact, > the code goes through great length to make sure we generate the href for > empty collections (since that href is needed to add the first element > to > an empty collection) > > The easiest way to 'fix' this for now is to comment those unsupported > collections out in system.rb. > > A better fix is to add a mechanism similar to what we do for $select to > models; call that the 'exclude' mechanism. The various generate_xxx > methods should set the that exclude parameter based on collections that > are declared in Rabbit, but not available (Rabbit knows that, Michal: > how do we get that info out of Rabbit ?) > > Similar to how CIMI::Model::Resource has @select_attrs it should also > have @exclude_attrs that are consulted in prepare and cause the > corresponding collection to be set to nil (which will remove it from the > XML/JSON output) > > > (2) > > With mock (not with fgcp), when I list a system's volumes, its id is > not > > generated correctly: > > > > d:\projects>curl --user mockuser:mockpassword > > http://localhost:3001/cimi/systems/system2/volumes?format=xml > > <Collection xmlns="http://schemas.dmtf.org/cimi/1"; > > > resourceURI="http://schemas.dmtf.org/cimi/1/SystemVolumeCollection";> > > <id>http://localhost:3001/cimi/system/system2/volumes</id> > > <count>1</count> > > <SystemVolume> > > <id>http://localhost:3001/cimi/volumes?id=sysvol1</id> > > ... > > > > That should be http://localhost:3001/cimi/volumes/sysvol1. > > Note my comment in mock_driver_cimi_methods.rb#system_volumes about > the > > 1st arg I pass to convert_cimi_mock_urls: > > > > #FIXME: with ":volumes", delete url becomes > > 'http://localhost:3001/cimi/volumes?id=sysvol1' > > #with ":system_volume" or ":system_volumes", undefined method > > `system_volume_url' for #<CIMI::Collections::Systems:0x44fe338> in > > mock_driver_cimi_methods.rb:261 > > volumes.map{|vol|convert_cimi_mock_urls(:volumes, vol, > > opts[:env])}.flatten > > This seems to be a Rabbit problem with the names of the subcollections; > somehow URL helpers for the subcollections do not get generated > correctly. Michal ? > > BTW, attached is a patch that addresses a few URL conversion problems > I > found in the mock driver; feel free to commit if it doesn't make things > worse ;) > > David
