do_node_tear_down not allowed for node % in state None

[Robert Collins]

2014-07-07 01:54:13.450 4811 ERROR wsme.api [-] Server-side error: "RPC do_node_tear_down not allowed for node 9c3ef377-b37e-4663-8ee3-e243a383044c in state None
Traceback (most recent call last):

  File "/opt/stack/venvs/ironic/local/lib/python2.7/site-packages/ironic/openstack/common/rpc/common.py", line 423, in catch_client_exception
    return func(*args, **kwargs)

  File "/opt/stack/venvs/ironic/local/lib/python2.7/site-packages/ironic/conductor/manager.py", line 428, in do_node_tear_down
    task.release_resources()

  File "/opt/stack/venvs/ironic/local/lib/python2.7/site-packages/ironic/openstack/common/excutils.py", line 70, in __exit__
    six.reraise(self.type_, self.value, self.tb)

  File "/opt/stack/venvs/ironic/local/lib/python2.7/site-packages/ironic/conductor/manager.py", line 405, in do_node_tear_down
    % {'node': node_id, 'state': node.provision_state})

InstanceDeployFailure: RPC do_node_tear_down not allowed for node 9c3ef377-b37e-4663-8ee3-e243a383044c in state None

I see this occur with no obvious lead-in, but once it happens the nova driver refuses to delete it. Instant wedge.

[Robert Collins (lifeless)]

Having looked at the code I don't understand why do_node_tear_down has the assert at the top: the cleanup and teardown code it layers on top of is all idempotent (and has to be for robustness) - so there should be absolutely no reason not to accept this command at any stage.

[John Stafford (john-stafford)]

Hi Rakesh,

Are you still working this bug or can it be unassigned?

[Michael Davies (mrda)]

This is no longer a bug - the provision_state check was removed somewhere in juno->kilo timeframe.