host API has inconsistent host name attribute

[Russell Sim]

When doing a list with the os-host extension the host name attribute is host_name in all other cases the host name attribute is host.

E.g. when doing a list operation a result like this is expected:

{"hosts": [
  {"zone": "internal", "host_name": "awesome-node1", "service": "compute"}]}

When doing a describe of the same host:

{"host": [{"resource": {"project": "(total)", "memory_mb": 193278, "host": "awesome-node1", "cpu": 48, "disk_gb": 98}},
          {"resource": {"project": "(used_now)", "memory_mb": 13312, "host": "awesome-node1", "cpu": 6, "disk_gb": 20}},
          {"resource": {"project": "(used_max)", "memory_mb": 12288, "host": "awesome-node1", "cpu": 6, "disk_gb": 20}},
          {"resource": {"project": "de59ee29134b4980bbb77608347ae08a", "memory_mb": 12288, "host": "awesome-node1", "cpu": 6, "disk_gb": 20}}]}

This is confusing at best.

It has already caused some problems in the official client library.

https://github.com/openstack/python-novaclient/commit/9ce03a98eb78652fd3480cb0d8323520fd78064c
https://github.com/openstack/python-novaclient/commit/73a0e7298aeb7ff43e70a865d2350923d269db69

[Alex Xu (xuhj)]

We can't fix it now for back-compabitility. But we may can doing this when we have micro-version in API

[Russell Sim (russell-sim)]

Cool, so what is the prefered name of the attribute? host_name?

This will need to be fixed in the client for the timebeing so it would be good if there was some decision regarding which is the better name.

[jichenjc (jichenjc)]

I prefer to use 'host' because of less change

and FYI, the microversion is still under development, I plan to submit a patch when it's done

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/144995

[jichenjc (jichenjc)]

I guess it won't be fixed because
https://review.openstack.org/#/c/144995/

got following comments,

Honestly, I don't think this kind of change is worth the incompatibility. It provides no new features, only consistency, in a way that most consumers won't really care about.

I also think methodologically it does this the wrong way by decorating a private method with the api version dispatch. Which we shouldn't be doing.

[OpenStack Infra (hudson-openstack)]

Change abandoned by jichenjc (<email address hidden>) on branch: master
Review: https://review.openstack.org/144995
Reason: no need this change