Incorrect documentation for property volume_type suggest name is volumes_type

[Fredrik]

Both the help in the openstack client and web pages states that the property name for specifying a volume type is "volumes_type" and not "volume_type".

# openstack dataprocessing node group template create --help
  --volumes-type <volumes-type>
                        Type of the volumes. This parameter will be taken into
                        account only if volumes-per-node is set and non-zero

Using: "volumes_type": "my-cinder",

# openstack dataprocessing node group template create --json worker.json
create() got an unexpected keyword argument 'volumes_type'

Changing to: "volume_type": "my-cinder",

# openstack dataprocessing node group template create --json worker.json
Volume type 'my-cinder' not found

This document has 6 occurrences of the same typo.
https://docs.openstack.org/developer/python-saharaclient/cli.html

[Vitalii Gridnev (vgridnev)]

This document is auto-generated. So, volumes_type parameter exists, and it will be taken into account if volumes-per-node is present. Otherwise there is no sense to set that up.

[Fredrik (fredrik-tarnell)]

The example was abbreviated.
I am setting up volumes for my worker nodes.

[root@openstack sahara-templates]# cat worker.json
{
    "plugin_name": "ambari",
    "hadoop_version": "2.4",
    "node_processes": [
        "DataNode",
        "NodeManager"
    ],
    "name": "ambari-worker",
    "floating_ip_pool": "6ae019fe-5670-4a4d-bd00-16b9e5d1ad3c",
    "flavor_id": "a7a473b7-8e42-4b9b-a763-c95197ad0a50",
    "volumes_per_node": 4,
    "volumes_size": 1,
    "volume_local_to_instance": true,
    "volume_mount_prefix": "/volumes/disk",
    "use_autoconfig": true,
    "auto_security_group": true,
    "is_public": true,
    "is_protected": false
}

I think you might have missed that small detail "s".
volumes_type should be volume_type with out the s

If it was coming from auto-doc why is it impossible to find "volumes_type" in the git repo and 27 occurances of "volume_type"?

[Fredrik (fredrik-tarnell)]

My example in the inital bug report shows that the openstack client think so as well.
create() got an unexpected keyword argument 'volumes_type'

And example after that ones changed to "volume_type" openstack client accepts that property.

Unfortunately I pasted an older json document which is missing the type parameter altogether.

It should have been:

{
    "plugin_name": "ambari",
    "hadoop_version": "2.4",
    "node_processes": [
        "DataNode",
        "NodeManager"
    ],
    "name": "ambari-worker",
    "floating_ip_pool": "6ae019fe-5670-4a4d-bd00-16b9e5d1ad3c",
    "flavor_id": "a7a473b7-8e42-4b9b-a763-c95197ad0a50",
    "volumes_per_node": 4,
    "volumes_size": 1,
    "volume_type": "ceph.osd",
    "volume_local_to_instance": true,
    "volume_mount_prefix": "/volumes/disk",
    "use_autoconfig": true,
    "auto_security_group": true,
    "is_public": true,
    "is_protected": false
}

[Vitalii Gridnev (vgridnev)]

When you use json to create node group you should use the following:

{
  'volume_type': ..

}
because that just exposes API interface: https://github.com/openstack/python-saharaclient/blob/master/saharaclient/api/node_group_templates.py#L32

but when you don't have json, you should use something like that:

openstack .. --volumes-type ceph --volumes-size 1

that was done to have the same prefix for arguments related to volumes.

https://github.com/openstack/python-saharaclient/blob/master/saharaclient/osc/v1/node_group_templates.py#L260

So, I agreed that this probably introduces some confusion and we should take a look on this and fix that in APIv2 (to have the same pattern of arguments in API).