Auto allocate: when router creation fails network and subnets are not cleaned up

[Gary Kotton]

When the router allocation fails then the auto allocated networks and subnets are left dangling.
An example of this is when the external network does not have a subnet created on it (this is a validation that is done on the vmware_nsx plugin - as an edge appliance needs to be able to have a external interface on the external network)
Another reason may be the tenant exceed in the router quota they she/he has.

[Armando Migliaccio (armando-migliaccio)]

A default external network without IP space seems pretty useless. As for quota errors, they are shown as Conflicts, so there may still be gaps in the cleanup logic.

[Armando Migliaccio (armando-migliaccio)]

As for Quota errors though, I am not sure it makes an awful lot of sense to protect the existing logic, as this API is meant to be used/useful in the early life cycle of the tenant deployment.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/361699
Committed: https://git.openstack.org/cgit/openstack/neutron/commit/?id=8fdc430e62e8f575499940ea73225d842db3927a
Submitter: Jenkins
Branch: master

commit 8fdc430e62e8f575499940ea73225d842db3927a
Author: Gary Kotton <email address hidden>
Date: Sun Aug 28 00:20:56 2016 -0700

    Auto allocation: ensure that networks and subnets are cleaned up

    In the event of a router create failure we need to make sure that
    the allocated resources are cleaned up.

    When router allocation fails it is difficult for an admin to
    understand the root cause. Adding the exception here will provide
    a little additional information.

    Closes-bug: #1617707

    Change-Id: I41940bea0327ee31494bd95867d0e60d9b5a24b8

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/neutron 9.0.0.0b3 development milestone.