[1.0] Remove args from ray start

[wuisawesome]

Why are these changes needed?

Related issue number

Checks

• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/latest/.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failure rates at https://ray-travis-tracker.herokuapp.com/.
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested (please justify below)

[ericl]

Breaks https://travis-ci.com/github/ray-project/ray/jobs/384279071

[ericl]

@wuisawesome please be more careful with the tests-ok tag, I did a double check and there were some failing tests!

[richardliaw]

Tests are failing: https://travis-ci.com/github/ray-project/ray/jobs/385303763

=================================== FAILURES ===================================
_________________________ test_calling_start_ray_head __________________________
call_ray_stop_only = None
    def test_calling_start_ray_head(call_ray_stop_only):
        # Test that we can call ray start with various command line
        # parameters. TODO(rkn): This test only tests the --head code path. We
        # should also test the non-head node code path.
    
        # Test starting Ray with no arguments.
        check_call_ray(["start", "--head"])
        check_call_ray(["stop"])
    
        # Test starting Ray with a redis port specified.
        check_call_ray(["start", "--head"])
        check_call_ray(["stop"])
    
        # Test starting Ray with a node IP address specified.
        check_call_ray(["start", "--head", "--node-ip-address", "127.0.0.1"])
        check_call_ray(["stop"])
    
        # Test starting Ray with a system config parameter set.
        check_call_ray([
            "start", "--head", "--system-config",
            "{\"metrics_report_interval_ms\":100}"
        ])
        check_call_ray(["stop"])
    
        # Test starting Ray with the object manager and node manager ports
        # specified.
        check_call_ray([
            "start", "--head", "--object-manager-port", "12345",
            "--node-manager-port", "54321"
        ])
        check_call_ray(["stop"])
    
        # Test starting Ray with the worker port range specified.
        check_call_ray([
            "start", "--head", "--min-worker-port", "50000", "--max-worker-port",
            "51000"
        ])
        check_call_ray(["stop"])
    
        # Test starting Ray with the number of CPUs specified.
        check_call_ray(["start", "--head", "--num-cpus", "2"])
        check_call_ray(["stop"])
    
        # Test starting Ray with the number of GPUs specified.
        check_call_ray(["start", "--head", "--num-gpus", "100"])
        check_call_ray(["stop"])
    
        # Test starting Ray with the max redis clients specified.
>       check_call_ray(["start", "--head", "--redis-max-clients", "100"])
/Users/travis/build/ray-project/ray/python/ray/tests/test_multi_node.py:419: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
args = ['start', '--head', '--redis-max-clients', '100'], capture_stdout = False
capture_stderr = False
    def check_call_ray(args, capture_stdout=False, capture_stderr=False):
        # We use this function instead of calling the "ray" command to work around
        # some deadlocks that occur when piping ray's output on Windows
        argv = ["ray"] + args
        if sys.platform == "win32":
            result = check_call_module(
                ray_main,
                argv,
                capture_stdout=capture_stdout,
                capture_stderr=capture_stderr)
        else:
            stdout_redir = None
            stderr_redir = None
            if capture_stdout:
                stdout_redir = subprocess.PIPE
            if capture_stderr and capture_stdout:
                stderr_redir = subprocess.STDOUT
            elif capture_stderr:
                stderr_redir = subprocess.PIPE
            proc = subprocess.Popen(argv, stdout=stdout_redir, stderr=stderr_redir)
            (stdout, stderr) = proc.communicate()
            if proc.returncode:
                raise subprocess.CalledProcessError(proc.returncode, argv, stdout,
>                                                   stderr)
E               subprocess.CalledProcessError: Command '['ray', 'start', '--head', '--redis-max-clients', '100']' returned non-zero exit status 2.

[ericl]

LINT, etc still failing. Please do not add tests ok until you have carefully checked the tests, thanks.