ceph-deploy源码分析（一）——源码结构与cli-优快云博客

本文链接：https://blog.youkuaiyun.com/wei2533690761/article/details/103986105

参考博客：http://www.hl10502.com/archives/page/2/
ceph-deploy源码分析（一）——源码结构与cli
链接：https://blog.youkuaiyun.com/weixin_33739646/article/details/91644617
ceph-deploy源码分析（二）——new模块：http://www.hl10502.com/2017/06/19/ceph-deploy-new/
ceph-deploy源码分析（三）——mon模块http://www.hl10502.com/2017/06/19/ceph-deploy-mon/
ceph-deploy源码分析（四）——osd模块：http://www.hl10502.com/2017/06/21/ceph-deploy-osd/
ceph-disk源码分析：http://www.hl10502.com/2017/06/23/ceph-disk-1/

ceph-deploy源码分析（一）——源码结构与cli

ceph-deploy是部署Ceph集群的工具，可以通过SSH方式往远程主机上安装 Ceph 软件包、创建集群、增加监视器、收集（或销毁）密钥、增加 OSD 和元数据服务器、配置管理主机，甚至拆除集群。
ceph-deploy使用Python开发，GitHub为 https://github.com/ceph/ceph-deploy。本次源码分析的ceph-deploy版本为1.5.37。
源码结构

ceph-deploy-1.5.37源码目录
├── ceph_deploy                # ceph-deploy源码目录
├── ceph_deploy.egg-info       # egg-info目录
├── CONTRIBUTING.rst           # 贡献指南
├── LICENSE                    # MIT LICENSE
├── MANIFEST.in                # 打包规则
├── PKG-INFO                   # PKG-INFO文件
├── README.rst                 # ceph-deploy介绍
├── scripts                    # 启动脚本目录
├── setup.cfg                  # setup.py配置
├── setup.py                   # ceph-deploy安装脚本
├── tox.ini                    # 标准化测试
└── vendor.py

源码入口

script目录下的ceph-deploy文件是ceph-deploy的入口，安装之后是/usr/bin/ceph-deploy。

ceph-deploy的__main__函数调用ceph_deploy.cli的main函数

...
...
from ceph_deploy.cli import main
if __name__ == '__main__'
    sys.exit(main())

cli模块

cli.py是命令行操作模块。

main函数调用_main函数

 def main(args=None, namespace=None):
    try:
        _main(args=args, namespace=namespace)
    finally:
        # This block is crucial to avoid having issues with
        # Python spitting non-sense thread exceptions. We have already
        # handled what we could, so close stderr and stdout.
        if not os.environ.get('CEPH_DEPLOY_TEST'):
            try:
                sys.stdout.close()
            except:
                pass
            try:
                sys.stderr.close()
            except:
                pass

_main函数

◦设置日志：Console Logger与File Logger添加到root_logger
◦调用argparse模块，解析cli参数
◦调用conf目录下的ceph-deploy模块set_overrides函数，从当前目录的cephdeploy.conf或~/.cephdeploy.conf文件获取ceph-deploy-global、ceph-deploy-[subcmd]配置项写入args
◦调用执行subcmd相应的模块

@catches((KeyboardInterrupt, RuntimeError, exc.DeployError,), handle_all=True)
def _main(args=None, namespace=None):
    # Set console logging first with some defaults, to prevent having exceptions
    # before hitting logging configuration. The defaults can/will get overridden
    # later.

    # Console Logger
    # 命令行控制台日志
    sh = logging.StreamHandler()
    # 不同级别的日志，使用不同的颜色区别：DEBUG蓝色；WARNIN黄色；ERROR红色；INFO白色
    sh.setFormatter(log.color_format())
    # 设置日志级别为WARNING
    sh.setLevel(logging.WARNING)

    # because we're in a module already, __name__ is not the ancestor of
    # the rest of the package; use the root as the logger for everyone
    # root_logger日志
    root_logger = logging.getLogger()

    # allow all levels at root_logger, handlers control individual levels
    # 设置root_logger日志级别为DEBUG
    root_logger.setLevel(logging.DEBUG)
    # 将 sh添加到root_logger
    root_logger.addHandler(sh)

    # 获取解析cli的argparse，调用argparse模块
    parser = get_parser()
    if len(sys.argv) < 2:
        parser.print_help()
        sys.exit()
    else:
        # 解析获取sys.argv中的ceph-deploy子命令和参数
        args = parser.parse_args(args=args, namespace=namespace)

    # 设置日志级别
    console_loglevel = logging.DEBUG  # start at DEBUG for now
    if args.quiet:
        console_loglevel = logging.WARNING
    if args.verbose:
        console_loglevel = logging.DEBUG

    # Console Logger
    sh.setLevel(console_loglevel)

    # File Logger
    # 文件日志
    fh = logging.FileHandler('ceph-deploy-{cluster}.log'.format(cluster=args.cluster))
    fh.setLevel(logging.DEBUG)
    fh.setFormatter(logging.Formatter(log.FILE_FORMAT))
    # 将 fh添加到root_logger
    root_logger.addHandler(fh)

    # Reads from the config file and sets values for the global
    # flags and the given sub-command
    # the one flag that will never work regardless of the config settings is
    # logging because we cannot set it before hand since the logging config is
    # not ready yet. This is the earliest we can do.

    # 从当前目录的cephdeploy.conf或~/.cephdeploy.conf文件获取ceph-deploy配置覆盖命令行参数
    args = ceph_deploy.conf.cephdeploy.set_overrides(args)

    LOG.info("Invoked (%s): %s" % (
        ceph_deploy.__version__,
        ' '.join(sys.argv))
    )
    log_flags(args)

    # args.func为cli中的subcmd子命令，调用相应的模块
    return args.func(args)

get_parser函数

[ceph_deploy.cli]以以下方式配置：
模块名 = 模块包名:执行函数

比如： new = ceph_deploy.new:make
new作为ceph-deploy的子命令，执行ceph-deploy new命令时，执行make函数

其他的模块也类似：
mon = ceph_deploy.mon:make
osd = ceph_deploy.osd:make
rgw = ceph_deploy.rgw:make
mds = ceph_deploy.mds:make
config = ceph_deploy.config:make

 def get_parser():
    # 调用argparse模块
    parser = argparse.ArgumentParser(
        prog='ceph-deploy',
        formatter_class=argparse.RawDescriptionHelpFormatter,
        description='Easy Ceph deployment\n\n%s' % __header__,
        )
    verbosity = parser.add_mutually_exclusive_group(required=False)
    verbosity.add_argument(
        '-v', '--verbose',
        action='store_true', dest='verbose', default=False,
        help='be more verbose',
        )
    verbosity.add_argument(
        '-q', '--quiet',
        action='store_true', dest='quiet',
        help='be less verbose',
        )
    parser.add_argument(
        '--version',
        action='version',
        version='%s' % ceph_deploy.__version__,
        help='the current installed version of ceph-deploy',
        )
    parser.add_argument(
        '--username',
        help='the username to connect to the remote host',
        )
    parser.add_argument(
        '--overwrite-conf',
        action='store_true',
        help='overwrite an existing conf file on remote host (if present)',
        )
    parser.add_argument(
        '--cluster',
        metavar='NAME',
        help='name of the cluster',
        type=validate.alphanumeric,
        )
    parser.add_argument(
        '--ceph-conf',
        dest='ceph_conf',
        help='use (or reuse) a given ceph.conf file',
    )
    sub = parser.add_subparsers(
        title='commands',
        metavar='COMMAND',
        help='description',
        )
    sub.required = True
    # 获取ceph_deploy.cli下的entry_points
    entry_points = [
        (ep.name, ep.load())
        for ep in pkg_resources.iter_entry_points('ceph_deploy.cli')
        ]
    # 根据priority排序
    entry_points.sort(
        key=lambda name_fn: getattr(name_fn[1], 'priority', 100),
        )
    # 将模块加入到子命令
    for (name, fn) in entry_points:
        p = sub.add_parser(
            name,
            description=fn.__doc__,
            help=fn.__doc__,
            )
        if not os.environ.get('CEPH_DEPLOY_TEST'):
            p.set_defaults(cd_conf=ceph_deploy.conf.cephdeploy.load())

        # flag if the default release is being used
        p.set_defaults(default_release=False)
        fn(p)
        p.required = True
    parser.set_defaults(
        cluster='ceph',
        )

    return parser

argparse模块

cli命令的解析使用了argparse.py模块。argparse是Python标准库中命令行选项、参数和子命令的解析器，其是为替代已经过时的optparse模块，argparse在Python2.7中被引入。

argparse模块请参考
https://docs.python.org/2.7/library/argparse.html
http://python.usyiyi.cn/translate/python_278/library/argparse.html

set_overrides函数

conf目录下的ceph-deploy模块set_overrides函数
◦调用load()函数
◦判断ceph-deploy配置文件ceph-deploy-global、ceph-deploy-[subcommand]配置项，调用override_subcommand()函数写入args。

def set_overrides(args, _conf=None):
    """
    Read the configuration file and look for ceph-deploy sections
    to set flags/defaults from the values found. This will alter the
    ``args`` object that is created by argparse.
    """
    # Get the subcommand name to avoid overwritting values from other
    # subcommands that are not going to be used
    subcommand = args.func.__name__
    command_section = 'ceph-deploy-%s' % subcommand
    # 加载ceph-deploy配置
    conf = _conf or load()

    for section_name in conf.sections():
        if section_name in ['ceph-deploy-global', command_section]:
            # ceph-deploy-global、ceph-deploy-[subcommand]配置项写入args
            override_subcommand(
                section_name,
                conf.items(section_name),
                args
            )
    return args

load函数，调用location()函数

def load():
    parser = Conf()
    # 读取解析ceph-deploy配置文件
    parser.read(location())
    return parser

location函数，调用_locate_or_create()函数

 def location():
    """
    Find and return the location of the ceph-deploy configuration file. If this
    file does not exist, create one in a default location.
    """
    return _locate_or_create()

_locate_or_create函数，判断当前目录的cephdeploy.conf或~/.cephdeploy.conf文件是否存在。
◦如果都不存在则调用create_stub函数创建一个~/.cephdeploy.conf文件。这个文件是根据模板创建的，内容为空。
◦如果存在（提前创建）cephdeploy.conf或~/.cephdeploy.conf文件，可以在文件中配置public_network、cluster_network、overwrite-conf等。

 def _locate_or_create():
    home_config = path.expanduser('~/.cephdeploy.conf')
    # With order of importance
    locations = [
        path.join(os.getcwd(), 'cephdeploy.conf'),
        home_config,
    ]

    for location in locations:
        if path.exists(location):
            logger.debug('found configuration file at: %s' % location)
            return location
    logger.info('could not find configuration file, will create one in $HOME')
    create_stub(home_config)
    return home_config