AZ should be validated durign instance create at the api layer

[Phil Day]

Currently the compute API layer does not do any validation of the availability zone, so a request like:

nova boot --availability_zone foobar

will be accepted, sent to the scheduler, and the instance will go to error.

There is already code in compute/api.py which processes the availability_zone value

def _handle_availability_zone(availability_zone)
...

So it seems like there should be some basic validation added to check that the zone exists and is available

[Yang Yu (yuyangbj)]

I found if the host is not in the zone, it is still accepted and boot VM.

[Yang Yu (yuyangbj)]

Basing on the current design, we should keep the AZ check in scheduler layer because current OpenStack could force the host to boot. For example,

nova boot --availability_zone foobar:host1

The VM will be created on compute host1 even though the foobar is not existing. So if we add the AZ check in api layer, it will break current design. So mark this bug to invalid.

[Phil Day (philip-day)]

Marking this back to New because I simply don't agree with the above analysis.

Personally I doubt if the intention of the forcehost capability was to allow an invalid AZ name by design, - but if there really is a need to be able to specify an invalid AZ name when forcing to a particular host (a privileged operation) then it would be possible to make the API validation skip the check for a valid AZ if the name contains a ":"

Not a reason to reject making a significant improvement to the common use case.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/33657

[Thierry Carrez (ttx)]

fix is not merged yet

[Thierry Carrez (ttx)]

Please do not set to Public Security unless you can point to an exploitable vulnerability.