race between api tempest tests across classes in 'test_rules' module

[Ramana Raja]

The API tempest test run of the cephfs-native CI, 'gate-manila-tempest-dsvm-cephfs-native', fails as it encounters a race between tests across classes.

The API tests are run by the CI as,
`/bin/bash tools/pretty_tox.sh manila_tempest_tests.tests.api --concurrency=12`

The lone test that fails,
 | setUpClass (manila_tempest_tests.tests.api.test_rules.ShareRulesTest)
 | ---------------------------------------------------------------------
 |
 | Captured traceback:
 | ~~~~~~~~~~~~~~~~~~~
 | Traceback (most recent call last):
 | File "tempest/test.py", line 262, in setUpClass
 | six.reraise(etype, value, trace)
 | File "tempest/test.py", line 255, in setUpClass
 | cls.resource_setup()
 | File "/opt/stack/new/manila/manila_tempest_tests/tests/api/test_rules.py", line 320, in resource_setup
 | cls.share = cls.create_share()
 | File "/opt/stack/new/manila/manila_tempest_tests/tests/api/base.py", line 348, in create_share
 | result = cls.create_shares([{"args": args, "kwargs": kwargs}])
 | File "/opt/stack/new/manila/manila_tempest_tests/tests/api/base.py", line 383, in create_shares
 | d["share"] = cls._create_share(*d["args"], **d["kwargs"])
 | File "/opt/stack/new/manila/manila_tempest_tests/tests/api/base.py", line 329, in _create_share
 | share = client.create_share(**kwargs)
 | File "/opt/stack/new/manila/manila_tempest_tests/services/share/v2/json/shares_client.py", line 218, in create_share
 | resp, body = self.post("shares", body, version=version)
 | File "/opt/stack/new/manila/manila_tempest_tests/services/share/v2/json/shares_client.py", line 60, in post
 | return super(SharesV2Client, self).post(url, body, headers=headers)
 | File "/usr/local/lib/python2.7/dist-packages/tempest_lib/common/rest_client.py", line 259, in post
 | return self.request('POST', url, extra_headers, headers, body)
 | File "/usr/local/lib/python2.7/dist-packages/tempest_lib/common/rest_client.py", line 640, in request
 | resp, resp_body)
 | File "/usr/local/lib/python2.7/dist-packages/tempest_lib/common/rest_client.py", line 698, in _error_checker
 | raise exceptions.BadRequest(resp_body, resp=resp)
 | tempest_lib.exceptions.BadRequest: Bad request
 | Details: {u'code': 400, u'message': u"Invalid input received: Invalid share protocol provided: GLUSTERFS. It is either disabled or unsupported. Available protocols: ['CEPHFS']."}

The above test fails because the class attribute 'protocol' is incorrectly set as 'GLUSTERFS' by a tempest run of different test class in an another thread.

Solution
vponomaryov: eliminate concurrent update of shared python objects.

[Valeriy Ponomaryov (vponomaryov)]

CephFS tempest not configured properly, see http://logs.openstack.org/11/270211/11/experimental/gate-manila-tempest-dsvm-cephfs-native/116d3ab/logs/tempest.txt.gz#_2016-02-02_20_27_53_310

just set opt "enable_cert_rules_for_protocols" to empty value
it will solve this problem
right now, it has "glusterfs" value and condition on line 311 in https://review.openstack.org/#/c/270211/11/manila_tempest_tests/tests/api/test_rules.py is True

[Valeriy Ponomaryov (vponomaryov)]

Nothing to do in Manila about this bug