'openstack allocation candidate list' returns non-compute nodes

[Linda Guo]

Description
===========
'openstack allocation candidate list ' command returns a list nodes that can meet the resource allocation requests, even some of them are not compute nodes.

For example, user requests to create a VM with 50G mem, 60G disk and 6 CPUs, placement API returns two nodes:

$ openstack allocation candidate list --resource MEMORY_MB=51200 --resource DISK_GB=60 --resource VCPU=6 --os-placement-api-version 1.10
+---+-----------------------------------+--------------------------------------+-----------------------------------------------+
| # | allocation | resource provider | inventory used/capacity |
+---+-----------------------------------+--------------------------------------+-----------------------------------------------+
| 1 | VCPU=6,MEMORY_MB=51200,DISK_GB=60 | 6eccd190-1d8c-488c-9db7-0cb49d08bdc7 | VCPU=0/1280,MEMORY_MB=0/385992,DISK_GB=0/1831 |
| 2 | VCPU=6,MEMORY_MB=51200,DISK_GB=60 | 27073646-35fc-4af8-9484-c52504d30a63 | VCPU=0/1280,MEMORY_MB=0/385993,DISK_GB=0/1831 |
+---+-----------------------------------+--------------------------------------+-----------------------------------------------+

$ openstack resource provider list |grep -E "6eccd190-1d8c-488c-9db7-0cb49d08bdc7|27073646-35fc-4af8-9484-c52504d30a63"
| 6eccd190-1d8c-488c-9db7-0cb49d08bdc7 | openstack-4.maas | 1 | 6eccd190-1d8c-488c-9db7-0cb49d08bdc7 | None |
| 27073646-35fc-4af8-9484-c52504d30a63 | openstack-5.maas | 1 | 27073646-35fc-4af8-9484-c52504d30a63 | None

But openstack-4 and openstack-5 are not compute nodes. Since VM can only be hosted on compute node, I think 'allocation candidate list ' should only return the schedulable compute nodes.

Steps to reproduce
==================
see above

Expected result
===============
'allocation candidate list ' should only return the schedulable compute nodes.

Actual result
=============
'allocation candidate list ' returns non-compute nodes.

Environment
===========
1. Exact version of OpenStack:

$ dpkg -l | grep nova
ii nova-api-os-compute 2:17.0.13-0ubuntu1~cloud0 all OpenStack Compute - OpenStack Compute API frontend
ii nova-common 2:17.0.13-0ubuntu1~cloud0 all OpenStack Compute - common files
ii nova-conductor 2:17.0.13-0ubuntu1~cloud0 all OpenStack Compute - conductor service
ii nova-consoleauth 2:17.0.13-0ubuntu1~cloud0 all OpenStack Compute - Console Authenticator
ii nova-novncproxy 2:17.0.13-0ubuntu1~cloud0 all OpenStack Compute - NoVNC proxy
ii nova-placement-api 2:17.0.13-0ubuntu1~cloud0 all OpenStack Compute - placement API frontend
ii nova-scheduler 2:17.0.13-0ubuntu1~cloud0 all OpenStack Compute - virtual machine scheduler

[Artom Lifshitz (notartom)]

Placement has no concept of compute nodes, it operates in terms of the resource provider abstraction. If a resource provider has the available inventory for the requested resources (and the traits match), it can be returned as an allocation candidate.

Can I ask what openstack-4.maas and openstack-5.maas are, and how did they end up in Placement if they're not compute nodes?

[Launchpad Janitor (janitor)]

[Expired for OpenStack Compute (nova) because there has been no activity for 60 days.]