When a VM boots using non-local disk, it tries to send an allocation of 'DISK_GB': 0. This violates the default min_unit of 1 and causes an error that looks like this:
[req-858cbed4-c113-45e8-94e3-1d8ee64f9de0 488c2b05a66b441199f4c1dca7accd5b 3fa5b55ecc154427b636119f0920d252 - default default] Bad inventory
Traceback (most recent call last):
File "/usr/lib/python2.7/site-packages/nova/api/openstack/placement/handlers/allocation.py", line 253, in set_allocations
allocations.create_all()
File "/usr/lib/python2.7/site-packages/oslo_versionedobjects/base.py", line 226, in wrapper
return fn(self, *args, **kwargs)
File "/usr/lib/python2.7/site-packages/nova/objects/resource_provider.py", line 1050, in create_all
self._set_allocations(self._context, self.objects)
File "/usr/lib/python2.7/site-packages/oslo_db/sqlalchemy/enginefacade.py", line 894, in wrapper
return fn(*args, **kwargs)
File "/usr/lib/python2.7/site-packages/nova/objects/resource_provider.py", line 1011, in _set_allocations
before_gens = _check_capacity_exceeded(conn, allocs)
File "/usr/lib/python2.7/site-packages/nova/objects/resource_provider.py", line 921, in _check_capacity_exceeded
resource_provider=rp_uuid)
InvalidAllocationConstraintsViolated: Unable to create allocation for 'DISK_GB' on resource provider 'f9398126-d0e8-4cf8-ae45-9103a88aa13d'. The requested amount would violate inventory constraints.
The causing code is at https://github.com/openstack/nova/blob/474c2ef28234dacc658e9a78762cac66ef7fe334/nova/scheduler/client/report.py#L105
The correct fix is probably that whenever the value of any resource class is zero, don't send that resource class in the dict.
Yeah I think that would be the best approach for fixing that bug.