Add support for swarm DataPathAddr

[hannseman]

This PR adds support for setting the swarm DataPathAddr by the argument data_path_addr when initialising and joining a swarm. The reason why the integration tests do not check for DataPathAddr is because it isn't returned on inspect. See moby/moby#33938

I also added a test to get some coverage on join_swarm as I couldn't find any tests for this.

See:
https://docs.docker.com/engine/api/v1.39/#operation/SwarmInit
https://docs.docker.com/engine/reference/commandline/swarm_init/

[shin-]

Thanks!

Not a full review, but why is DataPathAddr becoming data_addr_path instead of the more intuitive data_path_addr?

[hannseman]

@shin- haha wow, I don't know how I ended up getting that messed up. data_path_addr absolutely makes more sense.

Fixed now. Thanks for the correction.

[shin-]

Thanks! Overall LGTM, just a couple things on the tests side.

[hannseman]

Not really sure why the Jenkins build keeps failing. Can't seem to find the cause by looking at the build logs.

[shin-]

Not really sure why the Jenkins build keeps failing. Can't seem to find the cause by looking at the build logs.

The job that was building new versions of the dockerswarm/dind image isn't detecting new versions anymore, causing our integration tests to fail when attempting to use the missing dockerswarm/dind:19.03.0 image :(

[shin-]

@hannseman CI issues should be fixed now, can you rebase against current master? Thanks!

[hannseman]

@shin- sweet! I've rebased my PR:s but this one and #2295 seems to be failing on:
tests/integration/api_container_test.py::RestartContainerTest::test_restart_with_low_timeout

10:33:43 ______________ RestartContainerTest.test_restart_with_low_timeout ______________
10:33:43 tests/integration/api_container_test.py:1167: in test_restart_with_low_timeout
10:33:43     self.client.restart(container, timeout=3)
10:33:43 docker/utils/decorators.py:19: in wrapped
10:33:43     return f(self, resource_id, *args, **kwargs)
10:33:43 docker/api/container.py:1052: in restart
10:33:43     res = self._post(url, params=params, timeout=conn_timeout)
10:33:43 docker/utils/decorators.py:46: in inner
10:33:43     return f(self, *args, **kwargs)
10:33:43 docker/api/client.py:226: in _post
10:33:43     return self.post(url, **self._set_request_timeout(kwargs))
10:33:43 /usr/local/lib/python2.7/site-packages/requests/sessions.py:572: in post
10:33:43     return self.request('POST', url, data=data, json=json, **kwargs)
10:33:43 /usr/local/lib/python2.7/site-packages/requests/sessions.py:524: in request
10:33:43     resp = self.send(prep, **send_kwargs)
10:33:43 /usr/local/lib/python2.7/site-packages/requests/sessions.py:637: in send
10:33:43     r = adapter.send(request, **kwargs)
10:33:43 /usr/local/lib/python2.7/site-packages/requests/adapters.py:529: in send
10:33:43     raise ReadTimeout(e, request=request)
10:33:43 E   ReadTimeout: HTTPConnectionPool(host='dpy-dind-6-0-py2.7-19.03.0-beta3', port=2375): Read timed out. (read timeout=4)

[shin-]

Thanks!

And yep, I'm still addressing a bunch of flakes, hopefully to get us in a state where those are rare-to-nonexistent again. The particular one you mentioned was just addressed in 1a4881a

[shin-]

💯 Thanks!

[hannseman]

@shin- thanks for the review!