Thanks! > > +++ b/server/lib/cimi/collections/system_templates.rb > > Nothing wrong with what you did here, but I am wondering if there is > anything we can do to reduce this sort of boilerplate. Most of our > collections look exactly the same ...
I was planning to suggest the same, after I confirmed the copying and pasting I did is actually correct. I suppose we have two types of models/collections, the ones that are implemented directly in the drivers (like system, I suppose also meter and eventlog) and the ones that are mapped to Deltacloud's models first (machines/instances, images, etc.). Especially in the former case but maybe also in the latter case most of the code is the same. I based this and system.rb on I believe machine.rb and all I did was rename machine to system. > The puts just before the else is where things go wrong: without the > puts, the value of the if branch is whatever is assigned to systems, > which in turn is what gets returned from self.find - by inserting the > puts, you are now returning nil. Put a line 'systems' just after the > puts, and things look much better (at least, you're getting a valid > response back instead of a 500, though it had no systems for me) Thanks, that worked for me. Now both systems and system_templates are returning an entry for me: <Collection xmlns="http://schemas.dmtf.org/cimi/1"; resourceURI="http://schemas.dmtf.org/cimi/1/SystemCollection";> <id>http://localhost:3001/cimi/systems</id> <count>1</count> <System> <id>http://localhost:3001/cimi/systems/system1</id> <name>system1</name> <description>the first system</description> <created>Fri Feb 08 15:25:41 EET 2013</created> <state>STOPPED</state> <systems href="http://localhost:3001/cimi/systems/system1/systems"; /> <machines href="http://cimi.example.org/systems/87654/machines"; /> <volumes href="http://localhost:3001/cimi/systems/system1/volumes"; /> <networks href="http://cimi.example.org/systems/87654/networks"; /> <addresses href="http://localhost:3001/cimi/systems/system1/addresses"; /> <forwardingGroups href="http://localhost:3001/cimi/systems/system1/forwarding_groups"; /> <operation rel="edit" href="http://localhost:3001/cimi/systems/system1"; /> <operation rel="delete" href="http://localhost:3001/cimi/systems/system1"; /> </System> </Collection> And <Collection xmlns="http://schemas.dmtf.org/cimi/1"; resourceURI="http://schemas.dmtf.org/cimi/1/SystemTemplateCollection";> <id>http://localhost:3001/cimi/system_templates</id> <count>1</count> <SystemTemplate> <id>http://localhost:3001/cimi/system_templates/system template 1</id> <name>system template 1</name> <description>A mock system template</description> <created>Fri Feb 08 12:15:15 EET 2013</created> <componentDescriptor> <name>my machine</name> <description>an inline mock machine template</description> <type>http://schemas.dmtf.org/cimi/1/Machine</type> </componentDescriptor> <componentDescriptor> <name>my network</name> <description>a reference to an existing mock network template</description> <type>http://schemas.dmtf.org/cimi/1/Network</type> </componentDescriptor> <operation rel="edit" href="http://localhost:3001/cimi/system_templates/system template 1" /> <operation rel="delete" href="http://localhost:3001/cimi/system_templates/system template 1" /> </SystemTemplate> </Collection> The first time I ran it it didn't work for me because I needed to copy the json files from mock/data to my local /var/tmp/... folder. Maybe that's why it didn't work for you? I don't believe this response XML is correct so I'll look at that first. > As a general comment, try adding as little of the System* objects in one > patch as possible, again, to make reviewing easier. I haven't looked in > detail if that makes sense here or not, but it would be nice if it did, > and there's nothing wrong with a CIMI provider that only has System, but > none of the other System* collections. It is my understanding that a System references its components (machines, etc.) through System* collections, so without the System* collections all we'd have is a System class with a name, description and state, but no containing machines, etc. So I assume leaving them out here doesn't make sense? Cheers, Dies Koper
