Open Stack

From: Maxime Belanger <mbelanger at internap.com>
Reply-To: "OpenStack Development Mailing List (not for usage questions)" <openstack-dev at lists.openstack.org>
Date: Thursday, December 22, 2016 at 9:32 AM
To: "openstack-dev at lists.openstack.org" <openstack-dev at lists.openstack.org>
Subject: [openstack-dev] [Ironic] python-ironicclient weird behaviour


Hi Ironic folks,



Yesterday I come across a weird behaviour in python-ironicclient. I discussed briefly with the problem with Julia and befor creating a bug I would like your opinion on the problem and potential fix.



When doing nodes.get(resource_id=None), I get back a Node object set with attribute nodes={nodes: [{uuid: 'blah'}, {uuid: 'bleh'}]}. I guess that is not expected which is the bug I found and I know where to fix it if we agree to not change the behavour: If resource_id=None the client is doing a GET on /v1/nodes which is expected here : https://github.com/openstack/python-ironicclient/blob/1.8.0/ironicclient/common/base.py#L55.



The thing is, returning the first item of the list while doing a get on something that doesn't exists is a bit strange. I do not want to do extra validation on my application side to not pass None to the ironicclient. I would it to behave correctly: 404 or a even a 400.



What do you think/or prefer?



A. Fixing the bug and not changing actual behaviour of the API

B. Changing the API behaviour and returning a 404 error since resource None is not found

C. Changing the API behaviour and returning a 400 error because None is not a valid request. The will be identical to the command line that is doing this:
ironic node-show None
Invalid input for field/attribute node_ident. Value: 'None'. unable to convert to uuid_or_name (HTTP 400)

Thanks
Max

Thanks for tripping over this! :-( I think this is a bug in the API. I think it is just a coincidence that it happens to return the list of nodes (because the generated HTTP request works). That code should be checking that resource_id is non-empty -- that's what the CLI does. And if it isn't, raise an exception. I vote for B, sort of. It shouldn't return an http status error since no http request should be made. Looking at the existing exceptions, I'm guessing InvalidAttribute or ValidationError?

Your C above is a different situation, 'None' is a string that is the name of a node, but that string doesn't conform to the allowed name syntax.

--ruby

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20161222/bbcb211b/attachment.html>