AvailabilityZoneFilter keeps all hosts in default availability zone

[Marc Koderer]

Problem:
-----------

If a host is part of an availability zone (AZ) it should not be part of the default availability zone.
Otherwise scheduling hosts by default will be spread across AZs if this is not specified in
the CLI.

Steps to reproduce:
--------------------------

 1.) Create AZ:
   nova aggregate-create test01 test01
 2.) Start a vm without the AZ parameter:
  nova boot "testvm" --flavor 1 --image 581a7392-587a-4eac-849d-0536243cd0b4
 3.) VM will (potentially) be started on compute node test01 even though it's not part of the default AZ

Hints:
------
If --availability-zone is not set in CLI the metadata dict will contain {.. availability-zone=None..}.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/87247

[Huang Zhiteng (zhiteng-huang)]

My understanding is host aggregate is different from availability zone. So this doesn't seem like a valid bug.

[Marc Koderer (m-koderer)]

Thanks Huang for pointing to the right code snippet.
There is no default setting for default_schedule_zone and that's why the AZ doesn't filter anythig if no parameter is given in the CLI.

[Marc Koderer (m-koderer)]

@Huang: setting the default_schedule_zone does not solve the problem:

Problem:
-------

 1.) Setting default_schedule_zone to nova:

cat /etc/nova/nova.conf |grep schedule_zone
default_schedule_zone=nova

 2.) The function _handle_availability_zone in nova-api will return "nova" as AZ
 3.) But function host_passes() in availability_zone_filter.py has still "None" as result of props.get('availability_zone')

[Marc Koderer (m-koderer)]

Setting default_schedule_zone helped. Seems I missed rebooting the nova-api service.

But set to invalid.