Open Stack

Hi, guys,

I'm working neutron network stuff in nova V3 API. We will only pass port 
ids when create server, and
Nova won't proxy any neutron call in the future. I plan to add new v3 
network extension, that only
accept port ids as parameters. And it will pass those port ids in the 
old request_networks parameters.
(line 50 in 
https://review.openstack.org/#/c/36615/13/nova/api/openstack/compute/plugins/v3/networks.py)
Then other code path is same with before. In next release or when we 
remove nova-network, we
can add neutron specific code path. But v2 and v3 api's behavior is 
different. I need change something
before adding new network api. I want to hear you guys' suggestion 
first, ensure I'm working on the right
way.


1. Disable allocate ports automatically when create server without any 
port in v3 api.
When user create server without any port, the new server shouldn't be 
created with any ports in V3.
But in v2, nova-compute will allocate port from existed networks. I plan 
to pass parameter down to
nova-compute, that told nova-compute don't allocate ports for new 
server. And also keep old behavior
for v2 api.

2. Disable delete ports from neutron when remove server in v3 api.
In v2 api, after remove server, the port that attached to that server is 
removed by nova-compute.
But in v3 api, we shoudn't proxy any neutron call. Because there are 
some periodic tasks will delete
servers, just pass a parameter down to nova-compute from api isn't 
enough. So I plan to add a parameter in instance's
metadata when create server. When remove server, it will check the 
metadata first. If the server is marked as
created by v3 api, nova-compute won't remove attached neutron ports.

3. Enable pass port ids when multiple servers creation.
Currently multiple_create didn't support pass port ids. And we won't 
allocate ports automatically in v3 api.
So my plan as below:

When request with max_count=2 and ports=[{'id': 'port_id1'},  {'id': 
'port_id2'}, {'id': 'port_id3'}, {'id': 'port_id4}]
The first server create with ports 'port_id1' and 'port_id2', the second 
server create with ports 'port_id3' and 'port_id4'

When request with max_count=2 and ports = [{'id': 'port_id1'}]
The request return fault.
The request must be len(ports) % max_count == 0

Thanks
Alex
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20140212/8bf4a93d/attachment.html>