Instance fails to delete while building in cells v1 due to "ObjectActionError: Object action destroy failed because: host changed"

[Matt Riedemann]

Seeing this in a cells v1 job here:

http://logs.openstack.org/55/367455/1/check/gate-tempest-dsvm-cells/1d8af83/logs/screen-n-api.txt.gz#_2016-09-08_15_59_02_553

2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions [req-0f23fed6-0aaf-461f-96d8-f4aa5055b208 tempest-ListServersNegativeTestJSON-1414534476 tempest-ListServersNegativeTestJSON-1414534476] Unexpected exception in API method
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions Traceback (most recent call last):
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions File "/opt/stack/new/nova/nova/api/openstack/extensions.py", line 338, in wrapped
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions return f(*args, **kwargs)
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions File "/opt/stack/new/nova/nova/api/openstack/compute/servers.py", line 915, in delete
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions self._delete(req.environ['nova.context'], req, id)
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions File "/opt/stack/new/nova/nova/api/openstack/compute/servers.py", line 767, in _delete
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions self.compute_api.delete(context, instance)
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions File "/opt/stack/new/nova/nova/compute/cells_api.py", line 213, in delete
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions self._handle_cell_delete(context, instance, 'delete')
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions File "/opt/stack/new/nova/nova/compute/cells_api.py", line 224, in _handle_cell_delete
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions if self._delete_while_booting(context, instance):
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions File "/opt/stack/new/nova/nova/compute/api.py", line 1648, in _delete_while_booting
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions instance.destroy()
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions File "/usr/local/lib/python2.7/dist-packages/oslo_versionedobjects/base.py", line 226, in wrapper
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions return fn(self, *args, **kwargs)
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions File "/opt/stack/new/nova/nova/objects/instance.py", line 555, in destroy
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions reason='host changed')
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions ObjectActionError: Object action destroy failed because: host changed
2016-09-08 15:59:02.553 15924 ERROR nova.api.openstack.extensions

http://logs.openstack.org/55/367455/1/check/gate-tempest-dsvm-cells/1d8af83/console.html#_2016-09-08_16_13_48_892443

logs.openstack.org/55/367455/1/check/gate-tempest-dsvm-cells/1d8af83/logs/tempest.txt.gz#_2016-09-08_15_59_02_558

In this test it creates a server and then immediately deletes it without waiting for it to be ACTIVE first.

I wonder if we're racing with any of the BuildRequest stuff that's new in Newton.

http://logstash.openstack.org/#dashboard/file/logstash.json?query=message%3A%5C%22ObjectActionError%3A%20Object%20action%20destroy%20failed%20because%3A%20host%20changed%5C%22%20AND%20message%3A%5C%22_handle_cell_delete%5C%22%20AND%20tags%3A%5C%22screen-n-api.txt%5C%22&from=7d

10 hits in the last 7 days, check and gate, master (newton) only at this point.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/367597

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/367597
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=193e71bd30bbc6dd0a22b4e07d98e8a763ddda7b
Submitter: Jenkins
Branch: master

commit 193e71bd30bbc6dd0a22b4e07d98e8a763ddda7b
Author: Andrew Laski <email address hidden>
Date: Thu Sep 8 15:42:36 2016 -0400

    Handle ObjectActionError during cells instance delete

    There was logic in place to handle an ObjectActionError being raised
    when deleting an instance with cellsv1, however a second
    instance.destroy() call was added which did not have the same logic
    applied. This adds appropriate handling for that case.

    Change-Id: I007f249f772dbb53cf24a9bf06d2962001cb8384
    Closes-Bug: 1621574

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/nova 14.0.0.0rc1 release candidate.