AllocationList.delete_all() incorrectly assumes a single consumer

[Jay Pipes]

AllocationList.delete_all() looks like this:

```
    def delete_all(self):
        # Allocations can only have a single consumer, so take advantage of
        # that fact and do an efficient batch delete
        consumer_uuid = self.objects[0].consumer.uuid
        _delete_allocations_for_consumer(self._context, consumer_uuid)
        consumer_obj.delete_consumers_if_no_allocations(
            self._context, [consumer_uuid])
```

the problem with the above is that it is based on an old assumption: that a list of allocations will only ever involve a single consumer. This hasn't been the case ever since we introduced the ability to do `POST /allocations` which was 1.12 or 1.13 IIRC.

The safety concern about the above is that if someone in code does this:

```
allocs = AllocationList.get_all_by_resource_provider(self.ctx, compute_node_provider)
allocs.delete_all()
```

they would reasonable expect all of the allocations for a provider to be deleted. However, this is not the case. Only the allocations against the "first" consumer will be deleted.

The fix is simple... check to see if there are >1 consumers in the allocation list's objects and if so, don't call _delete_allocations_for_consumer(). Instead, call _delete_allocations_by_id() and do a DELETE FROM allocations WHERE id IN (...).

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/582382

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/582382
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=11c29ae4702ac4f50baa9fe3a29e3e413ad69b09
Submitter: Zuul
Branch: master

commit 11c29ae4702ac4f50baa9fe3a29e3e413ad69b09
Author: Jay Pipes <email address hidden>
Date: Thu Jul 12 14:57:35 2018 -0400

    do not assume 1 consumer in AllocList.delete_all()

    Ever since we introduced support for setting multiple consumers in a
    single POST /allocations, the AllocationList.delete_all() method has
    been housing a latent bad assumption and bug.

    The AllocationList.delete_all() method used to assume that the
    AllocationList's Allocation objects were only ever for a single
    consumer, and took a shortcut in deleting the allocation by deleting all
    allocations with the "first" Allocation's consumer UUID:

    ```python
        def delete_all(self):
            # Allocations can only have a single consumer, so take advantage of
            # that fact and do an efficient batch delete
            consumer_uuid = self.objects[0].consumer.uuid
            _delete_allocations_for_consumer(self._context, consumer_uuid)
            consumer_obj.delete_consumers_if_no_allocations(
                self._context, [consumer_uuid])
    ```

    The problem with the above is that if you get all the allocations for a
    single resource provider, using
    AllocationList.get_all_by_resource_provider() and there are more than
    one consumer allocating resources against that provider, then calling
    AllocationList.delete_all() will only delete *some* of the resource
    provider's allocations, not all of them.

    Luckily, the handler code has never used AllocationList.delete_all()
    after calling AllocationList.get_all_by_resource_provider(), and so
    we've not hit this latent bug in production.

    However, in the next patch in this series (the reshaper DB work), we
    *do* call AllocationList.delete_all() for allocation lists for each
    provider involved in the reshape operation, which is why this fix is
    important to get done correctly.

    Note that this patch renames AllocationList.create_all() to
    AllocationList.replace_all() to make it absolutely clear that all of
    the allocations for all consumers in the list are first *deleted* by the
    codebase and then re-created. We also remove the check in
    AllocationList.create_all() that the Allocation objects in the list must
    not have an 'id' field set. The reason for that is because in order to
    properly implement AllocationList.delete_all() to call DELETE FROM
    allocations WHERE id IN (<...>) we need the list of allocation record
    internal IDs. These id field values are now properly set on the
    Allocation objects when AllocationList.get_all_by_resource_provider()
    and AllocationList.get_all_by_consumer_id() are called. This allows that
    returned object to have delete_all() called on it and the DELETE
    statement to work properly.

    Change-Id: I12393b033054683bcc3e6f20da14e6243b4d5577
    Closes-bug: #1781430

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/nova 18.0.0.0b3 development milestone.