Nova does not respect default_schedule_zone `None`

[Ondrej Vasko]

Hello,

DESCRIPTION
-----------

I want Nova to behave like that when Availability Zone is not specified (`Any Availability Zone` is set in Horizon), it will create VM in random AZ.

STEPS TO REPRODUCE
------------------

Steps I do to configure that:

1. In `nova.conf` I set `default_schedule_zone = None` and restarted all nova services. I found this attribute in documentation [1] and also in Mirantis blog post [2]
2. I create 2 availability zones (2 host aggregates each with 1 hypervisor added).
3. I try to create VM In Horizon with `Any Availability Zone` and it results in following error:

    The requested availability zone is not available (HTTP 400).

The commands I executed to create AZs:

    ```
    openstack aggregate create HA-Test1 --zone AZ-Test1 --property availability_zone=AZ-Test1
    openstack aggregate create HA-Test2 --zone AZ-Test2 --property availability_zone=AZ-Test2
    openstack aggregate add host HA-Test1 os-compute-01
    openstack aggregate add host HA-Test2 os-compute-02
    ```

POINT
-----

Now this doesn't work as expected, but(!) when I remove `nova.conf` attribute `default_schedule_zone = None`, or I will configure it empty `default_schedule_zone = `, spawning of VMs works as expected and they are scheduled in random AZ.

Therefore I think that Nova doesn't handle `None` as python `None`, but as string (As if you set default schedule zone to zone `None`).

ENVIRONMENT
-----------

I am using stable Pike release with KVM + Libvirt installed via Openstack-Ansible on Ubuntu 16.04.

[1] https://docs.openstack.org/ocata/config-reference/compute/config-options.html
[2] https://www.mirantis.com/blog/the-first-and-final-word-on-openstack-availability-zones/

[Takashi Natsume (natsume-takashi)]

IMO, the issue is not an issue of nova.
Probably it is an issue of oslo.config.

[Ritesh Paiboina (rsritesh)]

Hi,

It seems to be a valid case. I tried the value as "None/none" for default_schedule_zone [DEFAULT] in section.

As per this document https://docs.openstack.org/ocata/config-reference/compute/config-options.html
we can keep None as default value.

What I see from nova api , when some value is given like 'None'
(Pdb) availability_zone
'None'

(Pdb) availability_zone
'Xyz'

So any value given to the default_schedule_zone is considered as string zone value.

So I am not sure weather we should exclude None as value for default_schedule_zone

[Ondrej Vasko (ondrej.vasko)]

I think that default value should be set to empty string or None should be treated as no zone.

It does not have a logic to set default schedule zone to zone named 'None' since default zone in Openstack is nova, it could be `nova` or nothing.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/550418

[OpenStack Infra (hudson-openstack)]

Change abandoned by Ritesh (<email address hidden>) on branch: master
Review: https://review.opendev.org/550418