Delete and re-add of same node to compute_nodes table is broken

[Chris Behrens]

When a compute node is deleted (or marked deleted) in the DB and another compute node is re-added with the same name, things break.

This is because the resource tracker caches the compute node object/dict and uses the 'id' to update the record. When this happens, rt.update_available_resources will raise a ComputeHostNotFound. This ends up short-circuiting the full run of the update_available_resource() periodic task.

This mostly applies when using a virt driver where a nova-compute manages more than 1 "hypervisor".

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/111837

[OpenStack Infra (hudson-openstack)]

Change abandoned by Joe Gordon (<email address hidden>) on branch: master
Review: https://review.openstack.org/111837

[Jim Rollenhagen (jim-rollenhagen)]

This seems like it's probably fixed by https://review.openstack.org/#/c/111838 but I have yet to verify.

[Jim Rollenhagen (jim-rollenhagen)]

15:15:57 edleafe | jroll: ok, got to look at that bug, and yeah, it should have closed it, but I forgot to put Closes-bug: in the commit message

And I agree it appears to fix it.

[Jim Rollenhagen (jim-rollenhagen)]

15:15:57 edleafe | jroll: ok, got to look at that bug, and yeah, it should have closed it, but I forgot to put Closes-bug: in the commit message

And I agree it appears to fix it.

This was released with Liberty-1