ceph-deploy 中使用remoto

在ceph-deploy 中使用remoto 这个第三方库来访问远程host.这个库的源码的路径为https://github.com/alfredodeza/remoto
例如在E:\ceph-deploy-master\ceph-deploy-master\ceph_deploy\hosts\common.py 中
def ceph_version(conn):
    """
    Log the remote ceph-version by calling `ceph --version`
    """
    return remoto.process.run(conn, ['ceph', '--version'])
通过remoto.process.run在远程host 上运行ceph --version
其源码分析如下：
路径为https://github.com/alfredodeza/remoto/blob/222014a3da8bb77cb462bace3a6c0e7200e5ea08/remoto/process.py
def run(conn, command, exit=False, timeout=None, **kw):
    """
    A real-time-logging implementation of a remote subprocess.Popen call where
    a command is just executed on the remote end and no other handling is done.
    :param conn: A connection oject
    :param command: The command to pass in to the remote subprocess.Popen
    :param exit: If this call should close the connection at the end
    :param timeout: How many seconds to wait after no remote data is received
                    (defaults to wait for ever)
    """
    stop_on_error = kw.pop('stop_on_error', True)
    if not kw.get('env'):
        # get the remote environment's env so we can explicitly add
        # the path without wiping out everything
        kw = extend_path(conn, kw)

    timeout = timeout or conn.global_timeout
    conn.logger.info('Running command: %s' % ' '.join(admin_command(conn.sudo, command)))
    result = conn.execute(_remote_run, cmd=command, **kw)
可见这里调用_remote_run，
def _remote_run(channel, cmd, **kw):
    import subprocess
    import sys
    from select import select
    stop_on_nonzero = kw.pop('stop_on_nonzero', True)

    process = subprocess.Popen(
        cmd,
        stdout=subprocess.PIPE,
        stderr=subprocess.PIPE,
        close_fds=True,
        **kw
    )
	
原来最后还是调用subprocess 这个第三方库来运行命令，可见remoto 只是subprocess的一个包装
第二个ceph-deploy的例子
def mon_create(distro, args, monitor_keyring):
    hostname = distro.conn.remote_module.shortname()
这里的remote_module定义如下：
路径为https://github.com/alfredodeza/remoto/blob/1e4fbd0f1af5e68ef0d824dd8b93139c3b86aa05/remoto/connection.py
class Connection(object):

    def __init__(self, hostname, logger=None, sudo=False, threads=1, eager=True,
                 detect_sudo=False, interpreter=None):
        self.sudo = sudo
        self.hostname = hostname
        self.logger = logger or FakeRemoteLogger()
        self.remote_module = None
        self.channel = None
        self.global_timeout = None  # wait for ever

        self.interpreter = interpreter or 'python%s' % sys.version_info[0]

    def _make_gateway(self, hostname):
        gateway = execnet.makegateway(
            self._make_connection_string(hostname)
        )
    def import_module(self, module):
        self.remote_module = ModuleExecute(self.gateway, module, self.logger)
        return self.remote_module
ModuleExecute 是一个类，定义如下：
		
class ModuleExecute(object):

    def __init__(self, gateway, module, logger=None):
        self.channel = gateway.remote_exec(module)
        self.module = module
        self.logger = logger
所以ceph-depoly中的remote_module.shortname()其实调用的gateway中的shortname()。而gateway的赋值如下：
    def _make_gateway(self, hostname):
        gateway = execnet.makegateway(
            self._make_connection_string(hostname)
        )
可见是通过execnet 这个第三方库完成的。这里就是调用python解释器去执行_make_connection_string产生的命令
execnet的介绍为http://codespeak.net/execnet/
execnet provides a share-nothing model with channel-send/receive communication for distributing execution across many Python interpreters across version, platform and network barriers. It has a minimal and fast API targetting the following uses:
distribute tasks to (many) local or remote CPUs
write and deploy hybrid multi-process applications
write scripts to administer multiple environments