summaryrefslogtreecommitdiff
path: root/os_client_config/tests
Commit message (Collapse)AuthorAgeFilesLines
* Replace guts with openstack.config1.31.0Monty Taylor2018-04-274-106/+41
| | | | | | | | | | | | | | | | | | | The code has been moved into openstacksdk. Make os-client-config a thin wrapper around that code. The legacy client code needs to stay, as it was removed in openstacksdk. We also have to plumb os_client_config.defaults and os_client_config.cloud_region.CloudRegion in so that users get the os-client-config versions from the openstacksdk methods that return the relevant objects. Support for per-service region_name has been dropped. This was never used anywhere and was added for compatibility with openstacksdk Profile objects. Since Profile objects are no more, we can remove it. It never made any sense. Change-Id: I7180e20aa675c0ccbe3718493a6eff2be7e1b8b6
* Merge "Do not apply format expansions to passwords"Zuul2017-12-091-0/+60
|\
| * Do not apply format expansions to passwordsJohn Dennis2017-12-051-0/+60
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | get_one_cloud() and get_one_cloud_osc() iterate over config values and try to expand any variables in those values by calling value.format(), however some config values (e.g. password) should never have format() applied to them, not only might that change the password but it will also cause the format() function to raise an exception if it can not parse the format string. Examples would be single brace (e.g. 'foo{') which raises an ValueError because it's looking for a matching end brace or a brace pair with a key value that cannot be found (e.g. 'foo{bar}') which raises a KeyError. It is not reasonsable to try to escape any braces because: 1) Escaping all braces breaks valid use of the format string syntax. 2) Trying to determine exactly which braces should be escaped and which should be preserved is a daunting task and likely would not be robust. 3) Some strings might look like valid format syntax but still should be escaped (e.g. "foo{bar}", if this appeared in a password we wouldn't escape it and there would be a key error on the 'bar' key. 4) In general passwords should never be modified, you never want to apply formatting to them. The right approach is to maintain a list of config values which are excluded from having formatting applied to them. At the moment that list just includes 'password' but perhaps down the road other exceptions might crop up. This patch follows this approach, the list of excluded values can easily be updated if others are discovered. Change-Id: I187bdec582d4c2cc6c7fda47a1538194137c616b Closes-Bug: 1635696 Signed-off-by: John Dennis <jdennis@redhat.com>
* | Make the get_service_type() overrides tolernat of no defaultsDean Troyer2017-12-081-0/+7
|/ | | | | | | | The service type overrides in get_service_type() fail if the API version keys are not present in the config dict, which happens when CloudConfig is created without reading defaults. Change-Id: I8d035cfd1afc1cad01ceac7cd643568e94897e27
* Added nat_source flag for networks.Monty Taylor2017-10-212-0/+4
| | | | | | | | | In some more complex clouds there can not only be more than one valid network on a server that NAT can attach to, there can also be more than one valid network from which to get a NAT address. Allow flagging a network so that it can be found. Change-Id: I3d8dd6d734a1013d2d4a43e11c3538c3a345820b
* Treat clouds.yaml with one cloud like envvarsMonty Taylor2017-09-191-0/+19
| | | | | | | | | If there is only one cloud and that cloud is envvars, things work as expected. If there is only one cloud in clouds.yaml and no envvars cloud, we throw an error, even though it should be obvious which cloud was intended. Change-Id: Ia49d0fb2cc7dca36476d0e5ae3fe2b2aa1209e59
* Revert "Revert "Use interface not endpoint_type for keystoneclient""Monty Taylor2017-05-131-1/+1
| | | | | | | | | | Unrevert the endpoint_type/interface patch. But this time around, put in a check for API version 2.0 and only apply the interface arg if it's for v3. This reverts commit 4493871824839782846c3825c754895876f8d08e. Change-Id: Ib347ec686d4d01788ee943c4c4f809aad06d9ccf
* Revert "Use interface not endpoint_type for keystoneclient"Jordan Pittier2017-05-131-2/+2
| | | | | | | | | | | | | | | This reverts commit 38e5eba621e48d74c05315da2b89e6c801f4c43f. This patch introduced a bug when using Keystone v2. With this patch, the following works: python -c "import os_client_config; print(os_client_config.make_client('identity', auth_url='http://localhost/identity_admin', username='admin', project_name='admin', password='testtest', identity_api_version='3').roles.list())" But changing identity_api_version from 3 to 2.0 raises an exception. Without this patch, both 3 and 2.0 works. Change-Id: I8d2ad71ff51a08af1166d36805b740ea272939ed
* Fix interactions with keystoneauth from newtonMonty Taylor2017-04-281-4/+24
| | | | | | | | | keystoneauth in newton did not have app_name or app_version as Session parameters. Although it isn't a super common combination, user agent strings aren't a reason to break something. Add a simple workaround. Change-Id: Ib5774389fefdbc190a4b78dd6784c8006afbb270
* Make _fix_argv() somewhat compatible with Argparse action='append'Jordan Pittier2017-04-231-0/+11
| | | | | | | | | | | | | | | | | | | Python Argparse supports the 'append' action [1] which is super handy to allow a user to repeat several times the same argument, each time with different values. This doesn't work with occ that tries to "fix argv" but raises this error: os_client_config.exceptions.OpenStackConfigException: The following options were given: '--foo,--foo' which contain duplicates except that one has _ and one has -. There is no sane way for us to know what you're doing. Remove the duplicate option and try again This patch tweak the _fix_argv() function so that it doesn't explode if the duplicate option has no '_' not '-' in its name. Change-Id: I4f06b6aff8d3ab1df45637399bc3a9b4b61764a9 Related-bug: #1685630
* Add ability to pass in user_agentMonty Taylor2017-04-182-2/+18
| | | | | | | | | | | | | | keystoneauth supports adding a user_agent info to the Session and Adapter via app_name. Allow users to add app_name/app_name and versions as desired. Also, add os-client-config into additional_user_agent. As an example, once this is landed and plumbed through shade, nodepool will set app_name='nodepool' and we'll have: User-Agent: nodepool/0.4.0 os-client-config/1.26.1 shade/1.19.1 keystoneauth1/2.18.0 python-requests/2.13.0 CPython/2.7.12 Change-Id: I1eb4dbd2587dcbe297b5c060c3c34b68ef51ef5e
* Stop special-casing idenity catalog lookupsMonty Taylor2017-03-281-13/+0
| | | | | | | | | We have a special case to work around a thing we're pretty sure keystoneclient used to do but apparently doesn't do anymore. Remove the workaround. Co-Authored-By: Jamie Lennox <jamielennox@gmail.com> Change-Id: I873ad91816150b593d4aef13dcd1520e8c91b22a
* Use interface not endpoint_type for keystoneclientMonty Taylor2017-03-091-2/+2
| | | | | | keystoneclient wants the interface argument. Change-Id: I5898d8621259256f962fc006df38049d0cb059f8
* Merge "Add ability to skip yaml loading"Jenkins2017-03-071-0/+32
|\
| * Add ability to skip yaml loadingMonty Taylor2017-02-271-0/+32
| | | | | | | | | | | | | | | | | | | | Added a flag, 'load_yaml_config' that defaults to True. If set to false, no clouds.yaml files will be loaded. This is beneficial if os-client-config wants to be used inside of a service where end-user clouds.yaml files would make things more confusing. Change-Id: Idbc82bb931e9edf1bbcc575237c0e202e219c218
* | Remove the keystoneclient auth fallbackMonty Taylor2017-02-152-8/+10
| | | | | | | | | | | | | | OSC doesn't use this codepath anyway (to my knowledge) and it masks errors in exceptionally strange ways. Change-Id: I15ec5aacb037813a98ac9ea8e9504a5d1cc90837
* | Add support for overriding mistral service typeMonty Taylor2017-02-151-0/+5
|/ | | | | | | | | The mistral team copied the heinous pervsion that the cinder team propagated upon the world and appended a version to their service_type. That's ok - there is nice copy-pastable code here we can use to prevent users from feeling the pain. Change-Id: Icf280f932014e4d9abeab3e944aece125988562e
* Revert "Fix interface_key for identity clients"ocata-em1.26.0Monty Taylor2017-01-181-2/+2
| | | | | | | | This patch breaks stable/newton devstack-gate of shade. This reverts commit fa4e1bd21db4bee2a0ee779067cdd659e647d7fc. Change-Id: I31a7831693f567a0717a9b41c242453fb937d6d7
* Update swift constructor to be Session awareMonty Taylor2016-12-221-81/+15
| | | | | | | | | | python-swiftclient has Session support now! Use it, and remove a pile of customization specific for swift. We still need to put a few parameters in different places, but that's no worse than glance or trove. Change-Id: Ic51aee2bc7b535aa4b6e261fb3deb59bd921f563
* Fix interface_key for identity clientsPaulo Matias2016-12-071-2/+2
| | | | | Change-Id: I83870e8b3ee6dc7fdbb6e9d67075cc4c08646e4e Closes-Bug: #1648212
* Revert "Remove validate_auth_ksc"1.24.0Monty Taylor2016-11-143-34/+16
| | | | | | | | | This reverts commit 1f9e2cd123b38a7e744fb8a784d0ee3b523de95e. Sad as this makes me, let's revert and come back to it when we figure out the cliff thing. Change-Id: I0413d5e3b3d8652833a8e7942ba81926787ba3bf
* Merge "Remove validate_auth_ksc"Jenkins2016-11-113-16/+34
|\
| * Remove validate_auth_kscMonty Taylor2016-11-113-16/+34
| | | | | | | | | | | | | | | | This was a workaround for python-openstackclient back when it was still dependent on keystoneclient. OSC has its own workaround now, so this should no longer be needed. Change-Id: Ib1877b7978b7b016b394232235e887360b6bdf85
* | Merge "Fix a bunch of tests"Jenkins2016-10-212-3/+29
|\ \ | |/
| * Fix a bunch of testsMonty Taylor2016-10-202-3/+29
| | | | | | | | | | | | | | There is a bug in validate_auth that gets hidden by validate_auth_ksc. These tests are the test fixes for it. Change-Id: I80d558a1c794725ba2a87fbd87bf8fbdf6633bee
* | Add support for volumev3 service typeMonty Taylor2016-10-211-0/+5
|/ | | | | | | | | | | | | Words cannot begin to adequately express the disappointment and rage I felt upon learning that the cinder team had unleashed 'volumev3' upon the world. Woe betide us, the mere users, for wanting to use a 'volume' endpoint and have that choice mean something. Perhaps if we beat ourselves with leather straps while crawling for days across the country we can remove more joy from our lives. In the meantime, until we can fully appreciate the existential crisis of being, let's continue to work around it in os-client-config. Change-Id: I171e3b01497b3e3a06c3a73577f0f67e0c1e6f73
* cloud_config:get_session_endpoint: catch Keystone EndpointNotFoundJordan Pittier2016-10-061-0/+9
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The docstring of the `get_session_endpoint` says ":returns: Endpoint for the service, or None if not found" but apparently the `None` part was forgotten. This leads to this kind of spectacular traceback where the exception bubles up (even through Shade): Traceback (most recent call last): File ".tox/run/bin/ospurge", line 11, in <module> load_entry_point('ospurge', 'console_scripts', 'ospurge')() File "/path/ospurge/main.py", line 154, in main for resource in resource_manager.list(): File "/path/ospurge/resources/swift.py", line 15, in list for container in self.cloud.list_containers(): File "/path/pkg/shade/openstackcloud.py", line 4909, in list_containers full_listing=full_listing)) File "/path/pkg/shade/task_manager.py", line 244, in submit_task return task.wait(raw) File "/path/pkg/shade/task_manager.py", line 121, in wait super(Task, self).wait() File "/path/pkg/shade/task_manager.py", line 96, in wait self._traceback) File "/path/pkg/six.py", line 686, in reraise raise value File "/path/pkg/shade/task_manager.py", line 105, in run self.done(self.main(client)) File "/path/pkg/shade/_tasks.py", line 549, in main return client.swift_client.get_account(**self.args)[1] File "/path/pkg/shade/openstackcloud.py", line 849, in swift_client 'object-store', swiftclient.client.Connection) File "/path/pkg/shade/openstackcloud.py", line 343, in _get_client **kwargs) File "/path/pkg/os_client_config/cloud_config.py", line 301, in get_legacy_client return self._get_swift_client(client_class=client_class, **kwargs) File "/path/pkg/os_client_config/cloud_config.py", line 369, in _get_swift_client endpoint = self.get_session_endpoint(service_key='object-store') File "/path/pkg/os_client_config/cloud_config.py", line 253, in get_session_endpoint region_name=self.region) File "/path/pkg/keystoneauth1/session.py", line 765, in get_endpoint return auth.get_endpoint(self, **kwargs) File "/path/pkg/keystoneauth1/identity/base.py", line 216, in get_endpoint service_name=service_name) File "/path/pkg/positional/__init__.py", line 101, in inner return wrapped(*args, **kwargs) File "/path/pkg/keystoneauth1/access/service_catalog.py", line 228, in url_for raise exceptions.EndpointNotFound(msg) keystoneauth1.exceptions.catalog.EndpointNotFound: public endpoint for object-store service in RegionOne region not found Change-Id: Idbf5081117bb0a13d04a1a5cb9fd7682baaf04e5
* Merge "Fix AttributeError in `get_config`"Jenkins2016-10-011-0/+33
|\
| * Fix AttributeError in `get_config`Jordan Pittier2016-09-261-0/+33
| | | | | | | | | | Change-Id: I52bdc44800da6c1393a69c4faf96375235ef98bb Closes-Bug: #1627690
* | Using assertIsNone() instead of assertEqual(None, ...)Cao Xuan Hoang2016-09-281-5/+5
|/ | | | | | | | | | | Following OpenStack Style Guidelines[1]: [H203] Unit test assertions tend to give better messages for more specific assertions. As a result, assertIsNone(...) is preferred over assertEqual(None, ...) and assertIs(None, ...) [1] http://docs.openstack.org/developer/hacking/#unit-tests-and-assertraises Change-Id: I4ce1745a90b043ea342fb157683b01f862c1bc3d
* Add prompting for KSA options1.21.0Dean Troyer2016-08-291-0/+37
| | | | | | | | | | | | | | | | Teach OpenStackConfig to prompt the user for KSA plugin options that have no value but have a prompt string defined. * Add pw_func argument to __init__() to be used as the callback for prompting the user. The default is None which skips the prompt step. * Add option_prompt() method to perform the checks for prompting, call the callback and save the returned value. This is public to handle cases where simply passing in a callback is insufficient for the prompt mechanism. Related-Bug: #1617384 Change-Id: I5faa86e94d6f71282ac270e2acfbd3016638c780
* Merge "Precedence final solution"1.20.1Jenkins2016-08-221-0/+36
|\
| * Precedence final solutionDean Troyer2016-08-201-0/+36
| | | | | | | | | | | | | | | | | | | | * Revert most of 'fixed_argparse change' from 1.19.1 * Create a new _validate_auth_correctly() method that contains the logic from 1.19.0 * Create a new get_one_cloud_osc() method for use by OSC to get the correct argument precedence without disrupting anyone else Change-Id: Iae86cc4e267f23dbe8d010688a288db5514f329d
* | Add support for configuring split-stack networks1.20.0Monty Taylor2016-08-182-3/+21
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Some clouds, like OSIC and v1 of DreamCompute, have a split stack network. This means that a single Neutron Network has both an IPv4 and an IPv6 subnet, but that the IPv4 subnet is a private/RFC-1918 and the IPv6 subnet is a Global network. As any inferrance information is attached to the Network, it's impossible to properly categorize IP addresses that are on the Server in such a scenario. Add support for ipv4 and ipv6 versions of the current routes_externally config value, with each of them defaulting to the value of the un-specialized routes_externally. Change-Id: I1e87a1423d20eac31175f44f5f7b38dfcf3a11cb
* | Pop domain-id from the config if we infer valuesMonty Taylor2016-08-111-0/+2
|/ | | | | | | | | | | | | | | | | | | | | | | | If the user specifies a project_{name,id}, then we currently infer that a domain_{name,id} is meant to be shorthand for user_domain_{name,id} and project_domain_{name,id}. However, in other contexts, domain_id is a perfectly valid value to pass to the auth plugins - such as when doing domain-scoped activities. The problem that was uncovered by the correction of argument precedence is that we didn't pop the domain-id out of the root config dict, so if a user set OS_DOMAIN_ID in the environment, we were happily setting the user and project versions, but then leaving domain_id set as well. This then meant that when we do the pass to pull valid arguments from the root dict to the auth dict, we also pulled in domain_id - which led to the error: AuthorizationFailure: Authentication cannot be scoped to multiple targets. Pick one of: project, domain, trust or unscoped Popping the value from the root dict causes the things to work as documented. Change-Id: I6d208e5ec4115d2e72d30b2fedc90a81ea754d5a
* Add test for precedence rulesMonty Taylor2016-08-051-0/+53
| | | | | | This should cover both the OSC and the ansible incoming use cases. Change-Id: I3fdc83837692d31c5579d91892a387a5d1023785
* Pull the network settings from the actual dict1.17.0Monty Taylor2016-04-062-0/+47
| | | | | | | Turns out self._openstack_config is not the config dict. self.config is. Also, return names. Change-Id: Ib2013e737b506b3a2acd7aa7b7884240c25384c5
* Flesh out netowrk config listMonty Taylor2016-04-021-2/+34
| | | | | | | | Add support for indicating default_interface. Also, add some validation and normalization code, some interface methods and, shockingly, documentation. Change-Id: Ib45b68894585ac02821d5d2376510fd7a8e8ee40
* Change network info indication to a generic listMonty Taylor2016-03-301-0/+31
| | | | | | | | | | | Networks can have more information than just internal or external. Notably, if you have two private networks and you're trying to assign floating ips, you need to know which network should be the recipient. This should be backwards compatible with existing external_network and internal_network options. Change-Id: I0d469339ba00486683fcd3ce2995002fa0a576d1
* Send swiftclient username/password and tokenMonty Taylor2016-02-121-12/+139
| | | | | | | | | | | For longer-lived operations, tokens can timeout and we need to get new ones. While in theory we should be keystoneauth aware and passing around sessions, swiftclient does not yet support this. So, instead of passing in just a preauthtoken, also pass in credentials if we have them. However, for plugin types that swift does not know about directly, only preauthtoken will be used as before. Change-Id: If724fdcd0649d9fa3b3ee7b127e49a3f77e3b767
* Stop ignoring v2password pluginMonty Taylor2016-01-251-0/+40
| | | | | | | | | | | | | We have no codepaths that currently set v2password plugin by default. However, there are some cases, such as old clouds, where a user needs to explicitly set v2password as the auth_type to avoid version discovery because their cloud is old enough to not support it. If the user sets v2password, keep it and align the auth parameters the other direction to set tenant_name and tenant_id. Co-Authored-By: David Shrewsbury <shrewsbury.dave@gmail.com> Change-Id: Ib9eb3ae163b79b67737d01868868187b6dee1756
* Don't set project_domain if not project scopedMonty Taylor2016-01-212-0/+26
| | | | | | | | | | | | | The code to expand domain_{name,id} to {user,project}_domain_{name,id} is flawed in that it sets a project_domain_{name,id} even if a project_{name,id} is not set. There is a valid use case for not having a project_{name,id} - specifically getting a domain-scoped token. In the case where we do not set a project, check for that and don't make further assumptions that the domain input needs to be "fixed". Closes-Bug: #1535676 Change-Id: I825fe4bc375687208bb176bb5990c23fe87c8f9d
* Merge "Pass version arg by name not position"1.14.0Jenkins2016-01-131-9/+9
|\
| * Pass version arg by name not positionMonty Taylor2016-01-131-9/+9
| | | | | | | | | | | | | | | | Everyone except neutron has a first parameter called "version" - so we can pass it by name. For neutron, add a workaround, becuase YAY people being different. Change-Id: Icfd92e5e31763ffccc1ff673298f89d1888941fe
* | Merge "Fix a precedence problem with auth arguments"Jenkins2016-01-131-0/+10
|\ \
| * | Fix a precedence problem with auth argumentsMonty Taylor2016-01-131-0/+10
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | With the current code, OS_TENANT_NAME will take precednece over --os-project-name beause OS_TENANT_NAME gets early-moved to config['auth']['project_name'], then when the argparse value gets put into config['project_name'] the auth fixing sees auth['project_name'] and thinks it should win. Change-Id: I97084ea221eb963f14d98cf550a04bbd5c7d954c
* | | Merge "Return empty dict instead of None for lack of file"Jenkins2016-01-131-0/+7
|\ \ \ | |/ / | | / | |/ |/|
| * Return empty dict instead of None for lack of fileMonty Taylor2016-01-131-0/+7
| | | | | | | | | | | | | | | | We return None for the file content for non-existent files as a fallback. This is normally fine, but in the case of a person having _only_ a secure.conf file, this means that the dictionary merge fails. Change-Id: I61cc0a8c709ea3510428fc3dfce63dc254c07c83
* | Replace assertEqual(None, *) with assertIsNone in testsLiuNanke2016-01-061-2/+1
|/ | | | | | | | | | Replace assertEqual(None, *) with assertIsNone in tests to have more clear messages in case of failure. There have one more place should be modified. Change-Id: I53a8f129db0108892b8377edce2dbf19b0b95f5d Closes-bug: #1280522
* Merge "Fix token_endpoint usage"1.13.1Jenkins2015-12-301-1/+2
|\
View Lorry Controller Status