#!/usr/bin/python3

DEFAULT_IMAGE='ceph/daemon-base:latest-master-devel'  # FIXME when octopus is ready!!!
DATA_DIR='/var/lib/ceph'
LOG_DIR='/var/log/ceph'
LOCK_DIR='/run/cephadm'
LOGROTATE_DIR='/etc/logrotate.d'
UNIT_DIR='/etc/systemd/system'
LOG_DIR_MODE=0o770
DATA_DIR_MODE=0o700
CONTAINER_PREFERENCE = ['podman', 'docker']  # prefer podman to docker
CUSTOM_PS1=r'[ceph: \u@\h \W]\$ '
DEFAULT_TIMEOUT=None # in seconds

"""
You can invoke cephadm in two ways:

1. The normal way, at the command line.

2. By piping the script to the python3 binary.  In this latter case, you should
   prepend one or more lines to the beginning of the script.

   For arguments,

       injected_argv = [...]

   e.g.,

       injected_argv = ['ls']

   For reading stdin from the '--config-and-json -' argument,

       injected_stdin = '...'
"""

import argparse
import fcntl
import json
import logging
import os
import random
import re
import select
import shutil
import socket
import string
import subprocess
import sys
import tempfile
import time
import errno
try:
    from typing import Dict, List, Tuple, Optional, Union, Any
except ImportError:
    pass
import uuid

from distutils.spawn import find_executable
from functools import wraps
from glob import glob
from threading import Thread

if sys.version_info >= (3, 0):
    from io import StringIO
else:
    from StringIO import StringIO

if sys.version_info >= (3, 2):
    from configparser import ConfigParser
else:
    from ConfigParser import SafeConfigParser

container_path = ''

class Error(Exception):
    pass

class TimeoutExpired(Error):
    pass

##################################


class Ceph(object):
    daemons = ('mon', 'mgr', 'mds', 'osd', 'rgw', 'rbd-mirror')


class Monitoring(object):
    """Define the configs for the monitoring containers"""

    port_map = {
        "prometheus": 9095,  # Avoid default 9090, due to conflict with cockpit UI
        "node-exporter": 9100,
        "grafana": 3000,
    }

    components = {
        "prometheus": {
            "image": {
                "image": "prom/prometheus:latest",
                "cpus": '2',
                "memory": '4GB',
                "args": [
                    "--config.file=/etc/prometheus/prometheus.yml",
                    "--storage.tsdb.path=/prometheus",
                    "--web.listen-address=:{}".format(port_map['prometheus']),
                ]
            },
            "config-json": [
                "prometheus.yml",
            ],
        },
        "node-exporter": {
            "image": {
                "image": "prom/node-exporter",
                "cpus": "1",
                "memory": "1GB",
                "args": [
                    "--no-collector.timex",
                ],
            }
        },
        "grafana": {
            "image": {
                "image": "pcuzner/ceph-grafana-el8:latest",
                "cpus": "2",
                "memory": "4GB",
                "args": [],
            },
            "config-json": [
                "grafana.ini",
                "provisioning/datasources/ceph-dashboard.yml",
                "certs/cert_file",
                "certs/cert_key",
            ],
        }
    }

def attempt_bind(s, address, port):
    # type (str) -> None
    try:
        s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
        s.bind((address, port))
    except (socket.error, OSError) as e:  # py2 and py3
        msg = 'Cannot bind to IP %s port %d: %s' % (address, port, e)
        logger.warning(msg)
        if e.errno == errno.EADDRINUSE:
            raise OSError(msg)
        elif e.errno == errno.EADDRNOTAVAIL:
            pass
    finally:
        s.close()

def port_in_use(port_num):
    # type (int) -> bool
    """Detect whether a port is in use on the local machine - IPv4 and IPv6"""

    try:
        s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
        attempt_bind(s, '0.0.0.0', port_num)

        s = socket.socket(socket.AF_INET6, socket.SOCK_STREAM)
        attempt_bind(s, '::', port_num)
    except OSError:
        return True
    else:
        return False

def check_ip_port(ip, port):
    if not args.skip_ping_check:
        logger.info('Verifying IP %s port %d ...' % (ip, port))
        if ip.startswith('[') or '::' in ip:
            s = socket.socket(socket.AF_INET6, socket.SOCK_STREAM)
        else:
            s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
        try:
            attempt_bind(s, ip, port)
        except OSError as e:
            raise Error(e)

##################################

# this is an abbreviated version of
# https://github.com/benediktschmitt/py-filelock/blob/master/filelock.py
# that drops all of the compatibility (this is Unix/Linux only).

try:
    TimeoutError
except NameError:
    TimeoutError = OSError

class Timeout(TimeoutError):
    """
    Raised when the lock could not be acquired in *timeout*
    seconds.
    """

    def __init__(self, lock_file):
        """
        """
        #: The path of the file lock.
        self.lock_file = lock_file
        return None

    def __str__(self):
        temp = "The file lock '{}' could not be acquired."\
               .format(self.lock_file)
        return temp


class _Acquire_ReturnProxy(object):
    def __init__(self, lock):
        self.lock = lock
        return None

    def __enter__(self):
        return self.lock

    def __exit__(self, exc_type, exc_value, traceback):
        self.lock.release()
        return None


class FileLock(object):
    def __init__(self, name, timeout = -1):
        if not os.path.exists(LOCK_DIR):
            os.mkdir(LOCK_DIR, 0o700)
        self._lock_file = os.path.join(LOCK_DIR, name + '.lock')

        # The file descriptor for the *_lock_file* as it is returned by the
        # os.open() function.
        # This file lock is only NOT None, if the object currently holds the
        # lock.
        self._lock_file_fd = None
        self.timeout = timeout
        # The lock counter is used for implementing the nested locking
        # mechanism. Whenever the lock is acquired, the counter is increased and
        # the lock is only released, when this value is 0 again.
        self._lock_counter = 0
        return None

    @property
    def is_locked(self):
        return self._lock_file_fd is not None

    def acquire(self, timeout=None, poll_intervall=0.05):
        """
        Acquires the file lock or fails with a :exc:`Timeout` error.
        .. code-block:: python
            # You can use this method in the context manager (recommended)
            with lock.acquire():
                pass
            # Or use an equivalent try-finally construct:
            lock.acquire()
            try:
                pass
            finally:
                lock.release()
        :arg float timeout:
            The maximum time waited for the file lock.
            If ``timeout < 0``, there is no timeout and this method will
            block until the lock could be acquired.
            If ``timeout`` is None, the default :attr:`~timeout` is used.
        :arg float poll_intervall:
            We check once in *poll_intervall* seconds if we can acquire the
            file lock.
        :raises Timeout:
            if the lock could not be acquired in *timeout* seconds.
        .. versionchanged:: 2.0.0
            This method returns now a *proxy* object instead of *self*,
            so that it can be used in a with statement without side effects.
        """
        # Use the default timeout, if no timeout is provided.
        if timeout is None:
            timeout = self.timeout

        # Increment the number right at the beginning.
        # We can still undo it, if something fails.
        self._lock_counter += 1

        lock_id = id(self)
        lock_filename = self._lock_file
        start_time = time.time()
        try:
            while True:
                if not self.is_locked:
                    logger.debug('Acquiring lock %s on %s', lock_id,
                                 lock_filename)
                    self._acquire()

                if self.is_locked:
                    logger.debug('Lock %s acquired on %s', lock_id,
                                 lock_filename)
                    break
                elif timeout >= 0 and time.time() - start_time > timeout:
                    logger.warning('Timeout acquiring lock %s on %s', lock_id,
                                   lock_filename)
                    raise Timeout(self._lock_file)
                else:
                    logger.debug(
                        'Lock %s not acquired on %s, waiting %s seconds ...',
                        lock_id, lock_filename, poll_intervall
                    )
                    time.sleep(poll_intervall)
        except:
            # Something did go wrong, so decrement the counter.
            self._lock_counter = max(0, self._lock_counter - 1)

            raise
        return _Acquire_ReturnProxy(lock = self)

    def release(self, force = False):
        """
        Releases the file lock.
        Please note, that the lock is only completly released, if the lock
        counter is 0.
        Also note, that the lock file itself is not automatically deleted.
        :arg bool force:
            If true, the lock counter is ignored and the lock is released in
            every case.
        """
        if self.is_locked:
            self._lock_counter -= 1

            if self._lock_counter == 0 or force:
                lock_id = id(self)
                lock_filename = self._lock_file

                logger.debug('Releasing lock %s on %s', lock_id, lock_filename)
                self._release()
                self._lock_counter = 0
                logger.debug('Lock %s released on %s', lock_id, lock_filename)

        return None

    def __enter__(self):
        self.acquire()
        return self

    def __exit__(self, exc_type, exc_value, traceback):
        self.release()
        return None

    def __del__(self):
        self.release(force = True)
        return None


    def _acquire(self):
        open_mode = os.O_RDWR | os.O_CREAT | os.O_TRUNC
        fd = os.open(self._lock_file, open_mode)

        try:
            fcntl.flock(fd, fcntl.LOCK_EX | fcntl.LOCK_NB)
        except (IOError, OSError):
            os.close(fd)
        else:
            self._lock_file_fd = fd
        return None

    def _release(self):
        # Do not remove the lockfile:
        #
        #   https://github.com/benediktschmitt/py-filelock/issues/31
        #   https://stackoverflow.com/questions/17708885/flock-removing-locked-file-without-race-condition
        fd = self._lock_file_fd
        self._lock_file_fd = None
        fcntl.flock(fd, fcntl.LOCK_UN)
        os.close(fd)
        return None


##################################
# Popen wrappers, lifted from ceph-volume

def call(command,  # type: List[str]
         desc=None,  # type: Optional[str]
         verbose=False,  # type: bool
         verbose_on_failure=True,  # type: bool
         timeout=DEFAULT_TIMEOUT,  # type: Optional[int]
         **kwargs):
    """
    Wrap subprocess.Popen to

    - log stdout/stderr to a logger,
    - decode utf-8
    - cleanly return out, err, returncode

    If verbose=True, log at info (instead of debug) level.

    :param verbose_on_failure: On a non-zero exit status, it will forcefully set
                               logging ON for the terminal
    :param timeout: timeout in seconds
    """
    if not desc:
        desc = command[0]
    timeout = timeout or args.timeout

    logger.debug("Running command: %s" % ' '.join(command))
    process = subprocess.Popen(
        command,
        stdout=subprocess.PIPE,
        stderr=subprocess.PIPE,
        close_fds=True,
        **kwargs
    )
    # get current p.stdout flags, add O_NONBLOCK
    stdout_flags = fcntl.fcntl(process.stdout, fcntl.F_GETFL)
    stderr_flags = fcntl.fcntl(process.stderr, fcntl.F_GETFL)
    fcntl.fcntl(process.stdout, fcntl.F_SETFL, stdout_flags | os.O_NONBLOCK)
    fcntl.fcntl(process.stderr, fcntl.F_SETFL, stderr_flags | os.O_NONBLOCK)

    out = ''
    err = ''
    reads = None
    stop = False
    out_buffer = ''   # partial line (no newline yet)
    err_buffer = ''   # partial line (no newline yet)
    start_time = time.time()
    end_time = None
    if timeout:
        end_time = start_time + timeout
    while not stop:
        if end_time and (time.time() >= end_time):
            logger.info(desc + ':timeout after %s seconds' % timeout)
            stop = True
            process.kill()
        if reads and process.poll() is not None:
            # we want to stop, but first read off anything remaining
            # on stdout/stderr
            stop = True
        else:
            reads, _, _ = select.select(
                [process.stdout.fileno(), process.stderr.fileno()],
                [], [], timeout
            )
        for fd in reads:
            try:
                message_b = os.read(fd, 1024)
                if isinstance(message_b, bytes):
                    message = message_b.decode('utf-8')
                if isinstance(message_b, str):
                    message = message_b
                if fd == process.stdout.fileno():
                    out += message
                    message = out_buffer + message
                    lines = message.split('\n')
                    out_buffer = lines.pop()
                    for line in lines:
                        if verbose:
                            logger.info(desc + ':stdout ' + line)
                        else:
                            logger.debug(desc + ':stdout ' + line)
                elif fd == process.stderr.fileno():
                    err += message
                    message = err_buffer + message
                    lines = message.split('\n')
                    err_buffer = lines.pop()
                    for line in lines:
                        if verbose:
                            logger.info(desc + ':stderr ' + line)
                        else:
                            logger.debug(desc + ':stderr ' + line)
                else:
                    assert False
            except (IOError, OSError):
                pass

    returncode = process.wait()

    if out_buffer != '':
        if verbose:
            logger.info(desc + ':stdout ' + out_buffer)
        else:
            logger.debug(desc + ':stdout ' + out_buffer)
    if err_buffer != '':
        if verbose:
            logger.info(desc + ':stderr ' + err_buffer)
        else:
            logger.debug(desc + ':stderr ' + err_buffer)

    if returncode != 0 and verbose_on_failure and not verbose:
        # dump stdout + stderr
        logger.info('Non-zero exit code %d from %s' % (returncode, ' '.join(command)))
        for line in out.splitlines():
            logger.info(desc + ':stdout ' + line)
        for line in err.splitlines():
            logger.info(desc + ':stderr ' + line)

    return out, err, returncode


def call_throws(command, **kwargs):
    # type: (List[str], Any) -> Tuple[str, str, int]
    out, err, ret = call(command, **kwargs)
    if ret:
        raise RuntimeError('Failed command: %s' % ' '.join(command))
    return out, err, ret


def call_timeout(command, timeout):
    #type (List[str], int) -> int

    logger.debug('Running command (timeout=%s): %s'
            % (timeout, ' '.join(command)))

    def raise_timeout(command, timeout):
        msg = 'Command \'%s\' timed out after %s seconds' % (command, timeout)
        logger.debug(msg)
        raise TimeoutExpired(msg)

    def call_timeout_py2(command, timeout):
        #type (List[str], int) -> int
        proc = subprocess.Popen(command)
        thread = Thread(target=proc.wait)
        thread.start()
        thread.join(timeout)
        if thread.is_alive():
            proc.kill()
            thread.join()
            raise_timeout(command, timeout)
        return proc.returncode

    def call_timeout_py3(command, timeout):
        #type (List[str], int) -> int
        try:
            return subprocess.call(command, timeout=timeout)
        except subprocess.TimeoutExpired as e:
            raise_timeout(command, timeout)

    ret = 1
    if sys.version_info >= (3, 3):
        ret = call_timeout_py3(command, timeout)
    else:
        # py2 subprocess has no timeout arg
        ret = call_timeout_py2(command, timeout)
    return ret

##################################

def is_available(what, func, retry_max=5):
    # type (str, func, Optional[int]) -> func
    """
    Wait for a service to become available

    :param what: the name of the service
    :param func: the callable object that determines availability
    :param retry_max: max number of retry invocations of func
    """
    @wraps(func)
    def func_wrapper(*args, **kwargs):
        logger.info('Waiting for %s...' % (what))
        retry_num = 1
        while True:
            if func(*args, **kwargs):
                break
            elif retry_num > retry_max:
                raise Error('%s not available after %s tries'
                        % (what, retry_max))

            logger.info('%s not available, waiting (%s/%s)...'
                    % (what, retry_num, retry_max))

            retry_num += 1
            time.sleep(1)
    return func_wrapper

def read_config(fn):
    # type: (Optional[str]) -> ConfigParser
    # bend over backwards here because py2's ConfigParser doesn't like
    # whitespace before config option names (e.g., '\n foo = bar\n').
    # Yeesh!
    if sys.version_info >= (3, 2):
        cp = ConfigParser()
    else:
        cp = SafeConfigParser()

    if fn:
        with open(fn, 'r') as f:
            raw_conf = f.read()
        nice_conf = re.sub(r'\n(\s)+', r'\n', raw_conf)
        s_io = StringIO(nice_conf)
        if sys.version_info >= (3, 2):
            cp.read_file(s_io)
        else:
            cp.readfp(s_io)

    return cp

def pathify(p):
    # type: (str) -> str
    if not p.startswith('/'):
        return os.path.join(os.getcwd(), p)
    return p


def get_podman_version():
    # type: () -> Tuple[int, ...]
    if 'podman' not in container_path:
        raise ValueError('not using podman')
    out, _, _ = call_throws([container_path, '--version'])
    return _parse_podman_version(out)

def _parse_podman_version(out):
    # type: (str) -> Tuple[int, ...]
    _, _, version_str = out.strip().split()

    def to_int(val, org_e=None):
        if not val and org_e:
            raise org_e
        try:
            return int(val)
        except ValueError as e:
            return to_int(val[0:-1], org_e or e)

    return tuple(map(to_int, version_str.split('.')))


def get_hostname():
    # type: () -> str
    return socket.gethostname()

def get_fqdn():
    # type: () -> str
    return socket.getfqdn() or socket.gethostname()

def generate_service_id():
    # type: () -> str
    return ''.join(random.choice(string.ascii_lowercase)
                   for _ in range(6))

def generate_password():
    # type: () -> str
    return ''.join(random.choice(string.ascii_lowercase + string.digits)
                   for i in range(10))

def normalize_container_id(i):
    # docker adds the sha256: prefix, but AFAICS both
    # docker (18.09.7 in bionic at least) and podman
    # both always use sha256, so leave off the prefix
    # for consistency.
    prefix = 'sha256:'
    if i.startswith(prefix):
        i = i[len(prefix):]
    return i

def make_fsid():
    # type: () -> str
    return str(uuid.uuid1())

def is_fsid(s):
    # type: (str) -> bool
    try:
        uuid.UUID(s)
    except ValueError:
        return False
    return True

def infer_fsid(func):
    """
    If we only find a single fsid in /var/lib/ceph/*, use that
    """
    @wraps(func)
    def _infer_fsid():
        if args.fsid:
            logger.debug('Using specified fsid: %s' % args.fsid)
            return func()

        fsids = set()
        daemon_list = list_daemons(detail=False)
        for daemon in daemon_list:
            if 'name' not in args or not args.name:
                fsids.add(daemon['fsid'])
            elif daemon['name'] == args.name:
                fsids.add(daemon['fsid'])
        fsids = list(fsids)

        if not fsids:
            # some commands do not always require an fsid
            pass
        elif len(fsids) == 1:
            logger.info('Inferring fsid %s' % fsids[0])
            args.fsid = fsids[0]
        else:
            raise Error('Cannot infer an fsid, one must be specified: %s' % fsids)
        return func()

    return _infer_fsid

def write_tmp(s, uid, gid):
    tmp_f = tempfile.NamedTemporaryFile(mode='w',
                                        prefix='ceph-tmp')
    os.fchown(tmp_f.fileno(), uid, gid)
    tmp_f.write(s)
    tmp_f.flush()

    return tmp_f

def makedirs(dir, uid, gid, mode):
    # type: (str, int, int, int) -> None
    if not os.path.exists(dir):
        os.makedirs(dir, mode=mode)
    else:
        os.chmod(dir, mode)
    os.chown(dir, uid, gid)
    os.chmod(dir, mode)   # the above is masked by umask...

def get_data_dir(fsid, t, n):
    # type: (str, str, Union[int, str]) -> str
    return os.path.join(args.data_dir, fsid, '%s.%s' % (t, n))

def get_log_dir(fsid):
    # type: (str) -> str
    return os.path.join(args.log_dir, fsid)

def make_data_dir_base(fsid, uid, gid):
    # type: (str, int, int) -> str
    data_dir_base = os.path.join(args.data_dir, fsid)
    makedirs(data_dir_base, uid, gid, DATA_DIR_MODE)
    makedirs(os.path.join(data_dir_base, 'crash'), uid, gid, DATA_DIR_MODE)
    makedirs(os.path.join(data_dir_base, 'crash', 'posted'), uid, gid,
             DATA_DIR_MODE)
    return data_dir_base

def make_data_dir(fsid, daemon_type, daemon_id, uid=None, gid=None):
    # type: (str, str, Union[int, str], int, int) -> str
    if not uid or not gid:
        (uid, gid) = extract_uid_gid()
    make_data_dir_base(fsid, uid, gid)
    data_dir = get_data_dir(fsid, daemon_type, daemon_id)
    makedirs(data_dir, uid, gid, DATA_DIR_MODE)
    return data_dir

def make_log_dir(fsid, uid=None, gid=None):
    # type: (str, int, int) -> str
    if not uid or not gid:
        (uid, gid) = extract_uid_gid()
    log_dir = get_log_dir(fsid)
    makedirs(log_dir, uid, gid, LOG_DIR_MODE)
    return log_dir

def make_var_run(fsid, uid, gid):
    # type: (str, int, int) -> None
    call_throws(['install', '-d', '-m0770', '-o', str(uid), '-g', str(gid),
                 '/var/run/ceph/%s' % fsid])

def copy_files(src, dst, uid=None, gid=None):
    # type: (List[str], str, int, int) -> None
    """
    Copy a files from src to dst
    """
    if not uid or not gid:
        (uid, gid) = extract_uid_gid()

    for src_file in src:
        dst_file = dst
        if os.path.isdir(dst):
            dst_file = os.path.join(dst, os.path.basename(src_file))

        logger.debug('copy file \'%s\' -> \'%s\'' % (src_file, dst_file))
        shutil.copyfile(src_file, dst_file)

        logger.debug('chown %s:%s \'%s\'' % (uid, gid, dst_file))
        os.chown(dst_file, uid, gid)

def move_files(src, dst, uid=None, gid=None):
    # type: (List[str], str, int, int) -> None
    """
    Move files from src to dst
    """
    if not uid or not gid:
        (uid, gid) = extract_uid_gid()

    for src_file in src:
        dst_file = dst
        if os.path.isdir(dst):
            dst_file = os.path.join(dst, os.path.basename(src_file))

        if os.path.islink(src_file):
            # shutil.move() in py2 does not handle symlinks correctly
            src_rl = os.readlink(src_file)
            logger.debug("symlink '%s' -> '%s'" % (dst_file, src_rl))
            os.symlink(src_rl, dst_file)
            os.unlink(src_file)
        else:
            logger.debug("move file '%s' -> '%s'" % (src_file, dst_file))
            shutil.move(src_file, dst_file)
            logger.debug('chown %s:%s \'%s\'' % (uid, gid, dst_file))
            os.chown(dst_file, uid, gid)

def find_program(filename):
    # type: (str) -> str
    name = find_executable(filename)
    if name is None:
        raise ValueError('%s not found' % filename)
    return name

def get_unit_name(fsid, daemon_type, daemon_id=None):
    # type (str, str, Optional[Union[int, str]]) -> str
    # accept either name or type + id
    if daemon_id is not None:
        return 'ceph-%s@%s.%s' % (fsid, daemon_type, daemon_id)
    else:
        return 'ceph-%s@%s' % (fsid, daemon_type)

def check_unit(unit_name):
    # type: (str) -> Tuple[bool, str]
    # NOTE: we ignore the exit code here because systemctl outputs
    # various exit codes based on the state of the service, but the
    # string result is more explicit (and sufficient).
    enabled = False
    try:
        out, err, code = call(['systemctl', 'is-enabled', unit_name],
                              verbose_on_failure=False)
        if code == 0:
            enabled = True
    except Exception as e:
        logger.warning('unable to run systemctl: %s' % e)
        enabled = False

    state = 'unknown'
    try:
        out, err, code = call(['systemctl', 'is-active', unit_name],
                              verbose_on_failure=False)
        out = out.strip()
        if out in ['active']:
            state = 'running'
        elif out in ['inactive']:
            state = 'stopped'
        elif out in ['failed', 'auto-restart']:
            state = 'error'
        else:
            state = 'unknown'
    except Exception as e:
        logger.warning('unable to run systemctl: %s' % e)
        state = 'unknown'
    return (enabled, state)

def get_legacy_config_fsid(cluster, legacy_dir=None):
    # type: (str, str) -> Optional[str]
    config_file = '/etc/ceph/%s.conf' % cluster
    if legacy_dir is not None:
        config_file = os.path.abspath(legacy_dir + config_file)

    if os.path.exists(config_file):
        config = read_config(config_file)
        if config.has_section('global') and config.has_option('global', 'fsid'):
            return config.get('global', 'fsid')
    return None

def get_legacy_daemon_fsid(cluster, daemon_type, daemon_id, legacy_dir=None):
    # type: (str, str, Union[int, str], str) -> Optional[str]
    fsid = None
    if daemon_type == 'osd':
        try:
            fsid_file = os.path.join(args.data_dir,
                                     daemon_type,
                                     'ceph-%s' % daemon_id,
                                     'ceph_fsid')
            if legacy_dir is not None:
                fsid_file = os.path.abspath(legacy_dir + fsid_file)
            with open(fsid_file, 'r') as f:
                fsid = f.read().strip()
        except IOError:
            pass
    if not fsid:
        fsid = get_legacy_config_fsid(cluster, legacy_dir=legacy_dir)
    return fsid

def get_daemon_args(fsid, daemon_type, daemon_id):
    # type: (str, str, Union[int, str]) -> List[str]
    r = list()  # type: List[str]

    if daemon_type in Ceph.daemons:
        r += [
            '--setuser', 'ceph',
            '--setgroup', 'ceph',
            '--default-log-to-file=false',
            '--default-log-to-stderr=true',
            '--default-log-stderr-prefix="debug "',
        ]
        if daemon_type == 'mon':
            r += [
                '--default-mon-cluster-log-to-file=false',
                '--default-mon-cluster-log-to-stderr=true',
            ]
    elif daemon_type in Monitoring.components:
        component = Monitoring.components[daemon_type]  # type: ignore
        metadata = component.get('image', list())  # type: ignore
        r += metadata.get('args', list())  # type: ignore
    return r

def create_daemon_dirs(fsid, daemon_type, daemon_id, uid, gid,
                       config=None, keyring=None,
                       reconfig=False):
    # type: (str, str, Union[int, str], int, int, Optional[str], Optional[str], Optional[bool]) ->  None
    data_dir = make_data_dir(fsid, daemon_type, daemon_id, uid=uid, gid=gid)
    make_log_dir(fsid, uid=uid, gid=gid)

    if config:
        with open(data_dir + '/config', 'w') as f:
            os.fchown(f.fileno(), uid, gid)
            os.fchmod(f.fileno(), 0o600)
            f.write(config)
    if keyring:
        with open(data_dir + '/keyring', 'w') as f:
            os.fchmod(f.fileno(), 0o600)
            os.fchown(f.fileno(), uid, gid)
            f.write(keyring)

    if daemon_type in Monitoring.components.keys():

        received_config = get_parm(args.config_json)
        required_config = Monitoring.components[daemon_type].get('config-json', list())  # type: ignore
        if required_config:
            if not received_config or not all(c in received_config.keys() for c in required_config):
                raise Error("{} deployment requires config-json which must "
                            "contain settings for {}".format(daemon_type.capitalize(), ', '.join(required_config)))

        # Set up directories specific to the monitoring component
        config_dir = ''
        if daemon_type == 'prometheus':
            data_dir_root = get_data_dir(fsid, daemon_type, daemon_id)
            config_dir = 'etc/prometheus'
            makedirs(os.path.join(data_dir_root, config_dir), uid, gid, 0o755)
            makedirs(os.path.join(data_dir_root, config_dir, 'alerting'), uid, gid, 0o755)
            makedirs(os.path.join(data_dir_root, 'data'), uid, gid, 0o755)
        elif daemon_type == 'grafana':
            data_dir_root = get_data_dir(fsid, daemon_type, daemon_id)
            config_dir = 'etc/grafana'
            makedirs(os.path.join(data_dir_root, config_dir), uid, gid, 0o755)
            makedirs(os.path.join(data_dir_root, config_dir, 'certs'), uid, gid, 0o755)
            makedirs(os.path.join(data_dir_root, config_dir, 'provisioning/datasources'), uid, gid, 0o755)
            makedirs(os.path.join(data_dir_root, 'data'), uid, gid, 0o755)

        # populate the config directory for the component from the config-json
        for fname in required_config:
            if isinstance(received_config[fname], list):
                content = '\n'.join(received_config[fname])
            else:
                content = received_config[fname]

            with open(os.path.join(data_dir_root, config_dir, fname), 'w') as f:
                os.fchown(f.fileno(), uid, gid)
                os.fchmod(f.fileno(), 0o600)
                f.write(content)

def get_parm(option):
    # type: (str) -> Dict[str, str]

    if not option:
        return dict()

    if option == '-':
        try:
            j = injected_stdin  # type: ignore
        except NameError:
            j = sys.stdin.read()
    else:
        # inline json string
        if option[0] == '{' and option[-1] == '}':
            j = option
        # json file
        elif os.path.exists(option):
            with open(option, 'r') as f:
                j = f.read()
        else:
            raise Error("Config file {} not found".format(option))

    try:
        js = json.loads(j)
    except ValueError:
        raise Error("Invalid JSON in {}".format(option))
    else:
        return js

def get_config_and_keyring():
    # type: () -> Tuple[str, str]
    if args.config_and_keyring:
        if args.config_and_keyring == '-':
            try:
                j = injected_stdin # type: ignore
            except NameError:
                j = sys.stdin.read()
        else:
            with open(args.config_and_keyring, 'r') as f:
                j = f.read()
        d = json.loads(j)
        config = d.get('config')
        keyring = d.get('keyring')
    else:
        if args.key:
            keyring = '[%s]\n\tkey = %s\n' % (args.name, args.key)
        elif args.keyring:
            with open(args.keyring, 'r') as f:
                keyring = f.read()
        else:
            raise Error('no keyring provided')
        with open(args.config, 'r') as f:
            config = f.read()
    return (config, keyring)

def get_config_and_both_keyrings():
    # type: () -> Tuple[str, str, Optional[str]]
    if args.config_and_keyrings:
        if args.config_and_keyrings == '-':
            try:
                j = injected_stdin # type: ignore
            except NameError:
                j = sys.stdin.read()
        else:
            with open(args.config_and_keyrings, 'r') as f:
                j = f.read()
        d = json.loads(j)
        return (d.get('config'), d.get('keyring'), d.get('crash_keyring'))
    else:
        if args.key:
            keyring = '[%s]\n\tkey = %s\n' % (args.name, args.key)
        elif args.keyring:
            with open(args.keyring, 'r') as f:
                keyring = f.read()
        else:
            raise Error('no keyring provided')
        crash_keyring = None
        if args.crash_keyring:
            with open(args.crash_keyring, 'r') as f:
                crash_keyring = f.read()
        with open(args.config, 'r') as f:
            config = f.read()
        return (config, keyring, crash_keyring)

def get_container_mounts(fsid, daemon_type, daemon_id,
                         no_config=False):
    # type: (str, str, Union[int, str, None], Optional[bool]) -> Dict[str, str]
    mounts = dict()

    if daemon_type in Ceph.daemons:
        if fsid:
            run_path = os.path.join('/var/run/ceph', fsid);
            if os.path.exists(run_path):
                mounts[run_path] = '/var/run/ceph:z'
            log_dir = get_log_dir(fsid)
            mounts[log_dir] = '/var/log/ceph:z'
            crash_dir = '/var/lib/ceph/%s/crash' % fsid
            if os.path.exists(crash_dir):
                mounts[crash_dir] = '/var/lib/ceph/crash:z'

    if daemon_type in Ceph.daemons and daemon_id:
        data_dir = get_data_dir(fsid, daemon_type, daemon_id)
        if daemon_type == 'rgw':
            cdata_dir = '/var/lib/ceph/radosgw/ceph-rgw.%s' % (daemon_id)
        else:
            cdata_dir = '/var/lib/ceph/%s/ceph-%s' % (daemon_type, daemon_id)
        mounts[data_dir] = cdata_dir + ':z'
        if not no_config:
            mounts[data_dir + '/config'] = '/etc/ceph/ceph.conf:z'
        if daemon_type == 'rbd-mirror':
            # rbd-mirror does not search for its keyring in a data directory
            mounts[data_dir + '/keyring'] = '/etc/ceph/ceph.client.rbd-mirror.%s.keyring' % daemon_id

    if daemon_type in ['mon', 'osd']:
        mounts['/dev'] = '/dev'  # FIXME: narrow this down?
        mounts['/run/udev'] = '/run/udev'
    if daemon_type == 'osd':
        mounts['/sys'] = '/sys'  # for numa.cc, pick_address, cgroups, ...
        mounts['/run/lvm'] = '/run/lvm'
        mounts['/run/lock/lvm'] = '/run/lock/lvm'

    if daemon_type in Monitoring.components and daemon_id:
        data_dir = get_data_dir(fsid, daemon_type, daemon_id)
        if daemon_type == 'prometheus':
            mounts[os.path.join(data_dir, 'etc/prometheus')] = '/etc/prometheus:Z'
            mounts[os.path.join(data_dir, 'data')] = '/prometheus:Z'
        elif daemon_type == 'node-exporter':
            mounts['/proc'] = '/host/proc:ro'
            mounts['/sys'] = '/host/sys:ro'
            mounts['/'] = '/rootfs:ro'
        elif daemon_type == "grafana":
            mounts[os.path.join(data_dir, 'etc/grafana/grafana.ini')] = '/etc/grafana/grafana.ini:Z'
            mounts[os.path.join(data_dir, 'etc/grafana/provisioning/datasources')] = '/etc/grafana/provisioning/datasources:Z'
            mounts[os.path.join(data_dir, 'etc/grafana/certs')] = '/etc/grafana/certs:Z'

    return mounts

def get_container(fsid, daemon_type, daemon_id, privileged=False,
                  container_args=[]):
    # type: (str, str, Union[int, str], bool, List[str]) -> CephContainer
    if daemon_type in ['mon', 'osd'] or privileged:
        # mon and osd need privileged in order for libudev to query devices
        container_args += ['--privileged']
    if daemon_type == 'rgw':
        entrypoint = '/usr/bin/radosgw'
        name = 'client.rgw.%s' % daemon_id
    elif daemon_type == 'rbd-mirror':
        entrypoint = '/usr/bin/rbd-mirror'
        name = 'client.rbd-mirror.%s' % daemon_id
    elif daemon_type in ['mon', 'mgr', 'mds', 'osd']:
        entrypoint = '/usr/bin/ceph-' + daemon_type
        name = '%s.%s' % (daemon_type, daemon_id)
    elif daemon_type in Monitoring.components:
        entrypoint = ''
        name = ''

    ceph_args = ['-n', name, '-f']

    if daemon_type in Monitoring.components:
        ceph_args = []

    return CephContainer(
        image=args.image,
        entrypoint=entrypoint,
        args=ceph_args + get_daemon_args(fsid, daemon_type, daemon_id),
        container_args=container_args,
        volume_mounts=get_container_mounts(fsid, daemon_type, daemon_id),
        cname='ceph-%s-%s.%s' % (fsid, daemon_type, daemon_id),
    )

def extract_uid_gid(img='', file_path='/var/lib/ceph'):
    # type: (str, str) -> Tuple[int, int]

    if not img:
        img = args.image

    out = CephContainer(
        image=img,
        entrypoint='stat',
        args=['-c', '%u %g', file_path]
    ).run()
    (uid, gid) = out.split(' ')
    return (int(uid), int(gid))

def deploy_daemon(fsid, daemon_type, daemon_id, c, uid, gid,
                  config=None, keyring=None,
                  osd_fsid=None,
                  reconfig=False):
    # type: (str, str, Union[int, str], CephContainer, int, int, Optional[str], Optional[str], Optional[str], Optional[bool]) -> None
    data_dir = get_data_dir(fsid, daemon_type, daemon_id)
    if reconfig and not os.path.exists(data_dir):
        raise Error('cannot reconfig, data path %s does not exist' % data_dir)
    if daemon_type == 'mon' and not os.path.exists(data_dir):
        assert config
        assert keyring
        # tmp keyring file
        tmp_keyring = write_tmp(keyring, uid, gid)

        # tmp config file
        tmp_config = write_tmp(config, uid, gid)

        # --mkfs
        create_daemon_dirs(fsid, daemon_type, daemon_id, uid, gid)
        mon_dir = get_data_dir(fsid, 'mon', daemon_id)
        log_dir = get_log_dir(fsid)
        out = CephContainer(
            image=args.image,
            entrypoint='/usr/bin/ceph-mon',
            args=['--mkfs',
                  '-i', str(daemon_id),
                  '--fsid', fsid,
                  '-c', '/tmp/config',
                  '--keyring', '/tmp/keyring',
            ] + get_daemon_args(fsid, 'mon', daemon_id),
            volume_mounts={
                log_dir: '/var/log/ceph:z',
                mon_dir: '/var/lib/ceph/mon/ceph-%s:z' % (daemon_id),
                tmp_keyring.name: '/tmp/keyring:z',
                tmp_config.name: '/tmp/config:z',
            },
        ).run()

        # write conf
        with open(mon_dir + '/config', 'w') as f:
            os.fchown(f.fileno(), uid, gid)
            os.fchmod(f.fileno(), 0o600)
            f.write(config)
    else:
        # dirs, conf, keyring
        create_daemon_dirs(
            fsid, daemon_type, daemon_id,
            uid, gid,
            config, keyring)

    if not reconfig:
        deploy_daemon_units(fsid, uid, gid, daemon_type, daemon_id, c,
                            osd_fsid=osd_fsid)

    update_firewalld(daemon_type)

    if reconfig and daemon_type not in Ceph.daemons:
        # ceph daemons do not need a restart; others (presumably) do to pick
        # up the new config
        call_throws(['systemctl', 'restart',
                     get_unit_name(fsid, daemon_type, daemon_id)])

def deploy_daemon_units(fsid, uid, gid, daemon_type, daemon_id, c,
                        enable=True, start=True,
                        osd_fsid=None):
    # type: (str, int, int, str, Union[int, str], CephContainer, bool, bool, Optional[str]) -> None
    # cmd
    data_dir = get_data_dir(fsid, daemon_type, daemon_id)
    with open(data_dir + '/unit.run', 'w') as f:
        if daemon_type == 'osd':
            # osds have a pre-start step
            assert osd_fsid
            prestart = CephContainer(
                image=args.image,
                entrypoint='/usr/sbin/ceph-volume',
                args=[
                    'lvm', 'activate',
                    str(daemon_id), osd_fsid,
                    '--no-systemd'
                ],
                container_args=['--privileged'],
                volume_mounts=get_container_mounts(fsid, daemon_type, daemon_id),
                cname='ceph-%s-%s.%s-activate' % (fsid, daemon_type, daemon_id),
            )
            f.write(' '.join(prestart.run_cmd()) + '\n')
        f.write(' '.join(c.run_cmd()) + '\n')
        os.fchmod(f.fileno(), 0o600)
    with open(data_dir + '/unit.poststop', 'w') as f:
        if daemon_type == 'osd':
            assert osd_fsid
            poststop = CephContainer(
                image=args.image,
                entrypoint='/usr/sbin/ceph-volume',
                args=[
                    'lvm', 'deactivate',
                    str(daemon_id), osd_fsid,
                ],
                container_args=['--privileged'],
                volume_mounts=get_container_mounts(fsid, daemon_type, daemon_id),
                cname='ceph-%s-%s.%s-deactivate' % (fsid, daemon_type,
                                                    daemon_id),
            )
            f.write(' '.join(poststop.run_cmd()) + '\n')
        os.fchmod(f.fileno(), 0o600)

    # systemd
    install_base_units(fsid)
    unit = get_unit_file(fsid, uid, gid)
    unit_file = 'ceph-%s@.service' % (fsid)
    with open(args.unit_dir + '/' + unit_file + '.new', 'w') as f:
        f.write(unit)
        os.rename(args.unit_dir + '/' + unit_file + '.new',
                  args.unit_dir + '/' + unit_file)
    call_throws(['systemctl', 'daemon-reload'])

    unit_name = get_unit_name(fsid, daemon_type, daemon_id)
    call(['systemctl', 'stop', unit_name],
         verbose_on_failure=False)
    call(['systemctl', 'reset-failed', unit_name],
         verbose_on_failure=False)
    if enable:
        call_throws(['systemctl', 'enable', unit_name])
    if start:
        call_throws(['systemctl', 'start', unit_name])

def update_firewalld(daemon_type):
    if args.skip_firewalld:
        return
    cmd = find_executable('firewall-cmd')
    if not cmd:
        logger.debug('firewalld does not appear to be present')
        return
    (enabled, state) = check_unit('firewalld.service')
    if not enabled:
        logger.debug('firewalld.service is not enabled')
        return

    fw_services = []
    fw_ports = []
    if daemon_type == 'mon':
        fw_services.append('ceph-mon')
    elif daemon_type in ['mgr', 'mds', 'osd']:
        fw_services.append('ceph')
    if daemon_type == 'mgr':
        fw_ports.append(8080)  # dashboard
        fw_ports.append(8443)  # dashboard
        fw_ports.append(9283)  # mgr/prometheus exporter
    elif daemon_type in Monitoring.port_map.keys():
        fw_ports.append(Monitoring.port_map[daemon_type])  # prometheus etc

    for svc in fw_services:
        out, err, ret = call([cmd, '--permanent', '--query-service', svc])
        if ret:
            logger.info('Enabling firewalld service %s in current zone...' % svc)
            out, err, ret = call([cmd, '--permanent', '--add-service', svc])
            if ret:
                raise RuntimeError(
                    'unable to add service %s to current zone: %s' % (svc, err))
        else:
            logger.debug('firewalld service %s is enabled in current zone' % svc)
    for port in fw_ports:
        port = str(port) + '/tcp'
        out, err, ret = call([cmd, '--permanent', '--query-port', port])
        if ret:
            logger.info('Enabling firewalld port %s in current zone...' % port)
            out, err, ret = call([cmd, '--permanent', '--add-port', port])
            if ret:
                raise RuntimeError('unable to add port %s to current zone: %s' %
                                   (port, err))
        else:
            logger.debug('firewalld port %s is enabled in current zone' % port)
    call_throws([cmd, '--reload'])

def install_base_units(fsid):
    # type: (str) -> None
    """
    Set up ceph.target and ceph-$fsid.target units.
    """
    # global unit
    existed = os.path.exists(args.unit_dir + '/ceph.target')
    with open(args.unit_dir + '/ceph.target.new', 'w') as f:
        f.write('[Unit]\n'
                'Description=All Ceph clusters and services\n'
                '\n'
                '[Install]\n'
                'WantedBy=multi-user.target\n')
        os.rename(args.unit_dir + '/ceph.target.new',
                  args.unit_dir + '/ceph.target')
    if not existed:
        # we disable before enable in case a different ceph.target
        # (from the traditional package) is present; while newer
        # systemd is smart enough to disable the old
        # (/lib/systemd/...) and enable the new (/etc/systemd/...),
        # some older versions of systemd error out with EEXIST.
        call_throws(['systemctl', 'disable', 'ceph.target'])
        call_throws(['systemctl', 'enable', 'ceph.target'])
        call_throws(['systemctl', 'start', 'ceph.target'])

    # cluster unit
    existed = os.path.exists(args.unit_dir + '/ceph-%s.target' % fsid)
    with open(args.unit_dir + '/ceph-%s.target.new' % fsid, 'w') as f:
        f.write('[Unit]\n'
                'Description=Ceph cluster {fsid}\n'
                'PartOf=ceph.target\n'
                'Before=ceph.target\n'
                '\n'
                '[Install]\n'
                'WantedBy=multi-user.target ceph.target\n'.format(
                    fsid=fsid)
        )
        os.rename(args.unit_dir + '/ceph-%s.target.new' % fsid,
                  args.unit_dir + '/ceph-%s.target' % fsid)
    if not existed:
        call_throws(['systemctl', 'enable', 'ceph-%s.target' % fsid])
        call_throws(['systemctl', 'start', 'ceph-%s.target' % fsid])

    # logrotate for the cluster
    with open(args.logrotate_dir + '/ceph-%s' % fsid, 'w') as f:
        """
        This is a bit sloppy in that the killall/pkill will touch all ceph daemons
        in all containers, but I don't see an elegant way to send SIGHUP *just* to
        the daemons for this cluster.  (1) systemd kill -s will get the signal to
        podman, but podman will exit.  (2) podman kill will get the signal to the
        first child (bash), but that isn't the ceph daemon.  This is simpler and
        should be harmless.
        """
        f.write("""# created by cephadm
/var/log/ceph/%s/*.log {
    rotate 7
    daily
    compress
    sharedscripts
    postrotate
        killall -q -1 ceph-mon ceph-mgr ceph-mds ceph-osd ceph-fuse radosgw rbd-mirror || pkill -1 -x "ceph-mon|ceph-mgr|ceph-mds|ceph-osd|ceph-fuse|radosgw|rbd-mirror" || true
    endscript
    missingok
    notifempty
    su root root
}
""" % fsid)

def deploy_crash(fsid, uid, gid, config, keyring):
    # type: (str, int, int, str, str) -> None
    crash_dir = os.path.join(args.data_dir, fsid, 'crash')
    makedirs(crash_dir, uid, gid, DATA_DIR_MODE)

    with open(os.path.join(crash_dir, 'keyring'), 'w') as f:
        os.fchmod(f.fileno(), 0o600)
        os.fchown(f.fileno(), uid, gid)
        f.write(keyring)
    with open(os.path.join(crash_dir, 'config'), 'w') as f:
        os.fchmod(f.fileno(), 0o600)
        os.fchown(f.fileno(), uid, gid)
        f.write(config)

    # ceph-crash unit
    mounts = {
        crash_dir: '/var/lib/ceph/crash:z',
        os.path.join(crash_dir, 'config'): '/etc/ceph/ceph.conf:z',
        os.path.join(crash_dir, 'keyring'): '/etc/ceph/ceph.keyring:z',
    }
    c = CephContainer(
        image=args.image,
        entrypoint='/usr/bin/ceph-crash',
        args=['-n', 'client.crash.%s' % get_hostname()],
        volume_mounts=mounts,
        cname='ceph-%s-crash' % (fsid),
    )
    unit_name = 'ceph-%s-crash.service' % fsid
    with open(os.path.join(args.unit_dir, unit_name + '.new'), 'w') as f:
        f.write('[Unit]\n'
                'Description=Ceph cluster {fsid} crash dump collector\n'
                'PartOf=ceph-{fsid}.target\n'
                'Before=ceph-{fsid}.target\n'
                '\n'
                '[Service]\n'
                'Type=simple\n'
                'ExecStartPre=-{container_path} rm ceph-{fsid}-crash\n'
                'ExecStart={cmd}\n'
                'ExecStop=-{container_path} stop ceph-{fsid}-crash\n'
                'Restart=always\n'
                'RestartSec=10\n'
                'StartLimitInterval=10min\n'
                'StartLimitBurst=10\n'
                '\n'
                '[Install]\n'
                'WantedBy=ceph-{fsid}.target\n'.format(
                    container_path=container_path,
                    fsid=fsid,
                    cmd=' '.join(c.run_cmd()))
        )
        os.rename(os.path.join(args.unit_dir, unit_name + '.new'),
                  os.path.join(args.unit_dir, unit_name))
    subprocess.check_output(['systemctl', 'enable', unit_name])
    subprocess.check_output(['systemctl', 'start', unit_name])

def get_unit_file(fsid, uid, gid):
    # type: (str, int, int) -> str
    install_path = find_program('install')
    u = """[Unit]
Description=Ceph daemon for {fsid}

# According to:
#   http://www.freedesktop.org/wiki/Software/systemd/NetworkTarget
# these can be removed once ceph-mon will dynamically change network
# configuration.
After=network-online.target local-fs.target time-sync.target
Wants=network-online.target local-fs.target time-sync.target

PartOf=ceph-{fsid}.target
Before=ceph-{fsid}.target

[Service]
LimitNOFILE=1048576
LimitNPROC=1048576
EnvironmentFile=-/etc/environment
ExecStartPre=-{container_path} rm ceph-{fsid}-%i
ExecStartPre=-{install_path} -d -m0770 -o {uid} -g {gid} /var/run/ceph/{fsid}
ExecStart=/bin/bash {data_dir}/{fsid}/%i/unit.run
ExecStop=-{container_path} stop ceph-{fsid}-%i
ExecStopPost=-/bin/bash {data_dir}/{fsid}/%i/unit.poststop
Restart=on-failure
RestartSec=10s
TimeoutStartSec=120
TimeoutStopSec=15
StartLimitInterval=30min
StartLimitBurst=5

[Install]
WantedBy=ceph-{fsid}.target
""".format(
    container_path=container_path,
    install_path=install_path,
    fsid=fsid,
    uid=uid,
    gid=gid,
    data_dir=args.data_dir)
    return u

##################################

class CephContainer:
    def __init__(self,
                 image,
                 entrypoint,
                 args=[],
                 volume_mounts={},
                 cname='',
                 container_args=[]):
        # type: (str, str, List[str], Dict[str, str], str, List[str]) -> None
        self.image = image
        self.entrypoint = entrypoint
        self.args = args
        self.volume_mounts = volume_mounts
        self.cname = cname
        self.container_args = container_args

    def run_cmd(self):
        # type: () -> List[str]
        vols = [] # type: List[str]
        envs = [] # type: List[str]
        cname = [] # type: List[str]
        entrypoint = [] # type: List[str]
        if self.entrypoint:
            entrypoint = ['--entrypoint', self.entrypoint]

        vols = sum(
            [['-v', '%s:%s' % (host_dir, container_dir)]
             for host_dir, container_dir in self.volume_mounts.items()], [])
        envs = [
            '-e', 'CONTAINER_IMAGE=%s' % self.image,
            '-e', 'NODE_NAME=%s' % get_hostname(),
        ]
        cname = ['--name', self.cname] if self.cname else []
        return [
            str(container_path),
            'run',
            '--rm',
            '--net=host',
        ] + self.container_args + \
        cname + envs + \
        vols + entrypoint + \
        [
            self.image
        ] + self.args # type: ignore

    def shell_cmd(self, cmd):
        # type: (List[str]) -> List[str]
        vols = [] # type: List[str]
        vols = sum(
            [['-v', '%s:%s' % (host_dir, container_dir)]
             for host_dir, container_dir in self.volume_mounts.items()], [])
        envs = [
            '-e', 'CONTAINER_IMAGE=%s' % self.image,
            '-e', 'NODE_NAME=%s' % get_hostname(),
        ]
        cmd_args = [] # type: List[str]
        if cmd:
            cmd_args = ['-c'] + cmd
        return [
            str(container_path),
            'run',
            '--rm',
            '--net=host',
        ] + self.container_args + envs + vols + [
            '--entrypoint', cmd[0],
            self.image
        ] + cmd[1:]

    def exec_cmd(self, cmd):
        # type: (List[str]) -> List[str]
        return [
            str(container_path),
            'exec',
        ] + self.container_args + [
            self.cname,
        ] + cmd

    def run(self, timeout=DEFAULT_TIMEOUT):
        # type: (Optional[int]) -> str
        logger.debug(self.run_cmd())
        out, _, _ = call_throws(
                self.run_cmd(), desc=self.entrypoint, timeout=timeout)
        return out

##################################

def command_version():
    # type: () -> int
    out = CephContainer(args.image, 'ceph', ['--version']).run()
    print(out.strip())
    return 0

##################################

def command_pull():
    # type: () -> None
    logger.info('Pulling latest %s...' % args.image)
    call_throws([container_path, 'pull', args.image])
    out, err, ret = call_throws([
        container_path, 'inspect',
        '--format', '{{.Id}}',
        args.image])
    ver = CephContainer(args.image, 'ceph', ['--version']).run()
    image_id = normalize_container_id(out.strip())
    print(json.dumps({
        'ceph_version': ver.strip(),
        'image_id': image_id,
        }, indent=4, sort_keys=True))


##################################

def command_bootstrap():
    # type: () -> int

    # verify output files
    if not args.allow_overwrite:
        for f in [args.output_config, args.output_keyring, args.output_pub_ssh_key]:
            if os.path.exists(f):
                raise Error('%s already exists; delete or pass '
                              '--allow-overwrite to overwrite' % f)

    # initial vars
    fsid = args.fsid or make_fsid()
    hostname = get_hostname()
    mon_id = args.mon_id or hostname
    mgr_id = args.mgr_id or generate_service_id()
    logging.info('Cluster fsid: %s' % fsid)

    l = FileLock(fsid)
    l.acquire()

    # ip
    r = re.compile(r':(\d+)$')
    if args.mon_ip:
        hasport = r.findall(args.mon_ip)
        if hasport:
            port = int(hasport[0])
            if port == 6789:
                addr_arg = '[v1:%s]' % args.mon_ip
            elif port == 3300:
                addr_arg = '[v2:%s]' % args.mon_ip
            else:
                logger.warning('Using msgr2 protocol for unrecognized port %d' %
                               port)
                addr_arg = '[v2:%s]' % args.mon_ip
            check_ip_port(args.mon_ip[0:-(len(str(port)))-1], port)
        else:
            addr_arg = '[v2:%s:3300,v1:%s:6789]' % (args.mon_ip, args.mon_ip)
            check_ip_port(args.mon_ip, 3300)
            check_ip_port(args.mon_ip, 6789)
    elif args.mon_addrv:
        addr_arg = args.mon_addrv
        if addr_arg[0] != '[' or addr_arg[-1] != ']':
            raise Error('--mon-addrv value %s must use square backets' %
                        addr_arg)
        for addr in addr_arg[1:-1].split(','):
            hasport = r.findall(addr)
            if not hasport:
                raise Error('--mon-addrv value %s must include port number' %
                            addr_arg)
            port = int(hasport[0])
            # strip off v1: or v2: prefix
            addr = re.sub(r'^\w+:', '', addr)
            check_ip_port(addr[0:-(len(str(port)))-1], port)
    else:
        raise Error('must specify --mon-ip or --mon-addrv')
    logger.debug('Final addrv is %s' % addr_arg)

    # config
    cp = read_config(args.config)
    if not cp.has_section('global'):
        cp.add_section('global')
    cp.set('global', 'fsid', fsid);
    cp.set('global', 'mon host', addr_arg)
    cp.set('global', 'container_image', args.image)
    cpf = StringIO()
    cp.write(cpf)
    config = cpf.getvalue()

    if not args.skip_pull:
        logger.info('Pulling latest %s container...' % args.image)
        call_throws([container_path, 'pull', args.image])

    logger.info('Extracting ceph user uid/gid from container image...')
    (uid, gid) = extract_uid_gid()

    # create some initial keys
    logger.info('Creating initial keys...')
    mon_key = CephContainer(
        image=args.image,
        entrypoint='/usr/bin/ceph-authtool',
        args=['--gen-print-key'],
    ).run().strip()
    admin_key = CephContainer(
        image=args.image,
        entrypoint='/usr/bin/ceph-authtool',
        args=['--gen-print-key'],
    ).run().strip()
    mgr_key = CephContainer(
        image=args.image,
        entrypoint='/usr/bin/ceph-authtool',
        args=['--gen-print-key'],
    ).run().strip()
    crash_key = CephContainer(
        image=args.image,
        entrypoint='/usr/bin/ceph-authtool',
        args=['--gen-print-key'],
    ).run().strip()

    keyring = ('[mon.]\n'
               '\tkey = %s\n'
               '\tcaps mon = allow *\n'
               '[client.admin]\n'
               '\tkey = %s\n'
               '\tcaps mon = allow *\n'
               '\tcaps mds = allow *\n'
               '\tcaps mgr = allow *\n'
               '\tcaps osd = allow *\n'
               '[mgr.%s]\n'
               '\tkey = %s\n'
               '\tcaps mon = profile mgr\n'
               '\tcaps mds = allow *\n'
               '\tcaps osd = allow *\n'
               '[client.crash.%s]\n'
               '\tkey = %s\n'
               '\tcaps mon = profile crash\n'
               '\tcaps mgr = profile crash\n'
               % (mon_key, admin_key, mgr_id, mgr_key, hostname, crash_key))

    # tmp keyring file
    tmp_bootstrap_keyring = write_tmp(keyring, uid, gid)

    # create initial monmap, tmp monmap file
    logger.info('Creating initial monmap...')
    tmp_monmap = write_tmp('', 0, 0)
    out = CephContainer(
        image=args.image,
        entrypoint='/usr/bin/monmaptool',
        args=['--create',
              '--clobber',
              '--fsid', fsid,
              '--addv', mon_id, addr_arg,
              '/tmp/monmap'
        ],
        volume_mounts={
            tmp_monmap.name: '/tmp/monmap:z',
        },
    ).run()

    # pass monmap file to ceph user for use by ceph-mon --mkfs below
    os.fchown(tmp_monmap.fileno(), uid, gid)

    # create mon
    logger.info('Creating mon...')
    create_daemon_dirs(fsid, 'mon', mon_id, uid, gid)
    mon_dir = get_data_dir(fsid, 'mon', mon_id)
    log_dir = get_log_dir(fsid)
    out = CephContainer(
        image=args.image,
        entrypoint='/usr/bin/ceph-mon',
        args=['--mkfs',
              '-i', mon_id,
              '--fsid', fsid,
              '-c', '/dev/null',
              '--monmap', '/tmp/monmap',
              '--keyring', '/tmp/keyring',
        ] + get_daemon_args(fsid, 'mon', mon_id),
        volume_mounts={
            log_dir: '/var/log/ceph:z',
            mon_dir: '/var/lib/ceph/mon/ceph-%s:z' % (mon_id),
            tmp_bootstrap_keyring.name: '/tmp/keyring:z',
            tmp_monmap.name: '/tmp/monmap:z',
        },
    ).run()

    with open(mon_dir + '/config', 'w') as f:
        os.fchown(f.fileno(), uid, gid)
        os.fchmod(f.fileno(), 0o600)
        f.write(config)

    make_var_run(fsid, uid, gid)
    mon_c = get_container(fsid, 'mon', mon_id)
    deploy_daemon(fsid, 'mon', mon_id, mon_c, uid, gid,
                  config=None, keyring=None)

    # client.admin key + config to issue various CLI commands
    tmp_admin_keyring = write_tmp('[client.admin]\n'
                                  '\tkey = ' + admin_key + '\n',
                                       uid, gid)
    tmp_config = write_tmp(config, uid, gid)

    # a CLI helper to reduce our typing
    def cli(cmd, extra_mounts={}, timeout=DEFAULT_TIMEOUT):
        # type: (List[str], Dict[str, str], Optional[int]) -> str
        mounts = {
            log_dir: '/var/log/ceph:z',
            tmp_admin_keyring.name: '/etc/ceph/ceph.client.admin.keyring:z',
            tmp_config.name: '/etc/ceph/ceph.conf:z',
        }
        for k, v in extra_mounts.items():
            mounts[k] = v
        timeout = timeout or args.timeout
        return CephContainer(
            image=args.image,
            entrypoint='/usr/bin/ceph',
            args=cmd,
            volume_mounts=mounts,
        ).run(timeout=timeout)

    logger.info('Waiting for mon to start...')
    c = CephContainer(
        image=args.image,
        entrypoint='/usr/bin/ceph',
        args=[
            'status'],
        volume_mounts={
            mon_dir: '/var/lib/ceph/mon/ceph-%s:z' % (mon_id),
            tmp_admin_keyring.name: '/etc/ceph/ceph.client.admin.keyring:z',
            tmp_config.name: '/etc/ceph/ceph.conf:z',
        },
    )
    def is_mon_available():
        out, err, ret = call(c.run_cmd(), desc=c.entrypoint, timeout=30)
        return ret == 0
    is_available('mon', is_mon_available)()

    # assimilate and minimize config
    if not args.no_minimize_config:
        logger.info('Assimilating anything we can from ceph.conf...')
        cli([
            'config', 'assimilate-conf',
            '-i', '/var/lib/ceph/mon/ceph-%s/config' % mon_id
        ], {
            mon_dir: '/var/lib/ceph/mon/ceph-%s:z' % mon_id
        })
        logger.info('Generating new minimal ceph.conf...')
        cli([
            'config', 'generate-minimal-conf',
            '-o', '/var/lib/ceph/mon/ceph-%s/config' % mon_id
        ], {
            mon_dir: '/var/lib/ceph/mon/ceph-%s:z' % mon_id
        })
        # re-read our minimized config
        with open(mon_dir + '/config', 'r') as f:
            config = f.read()
        logger.info('Restarting the monitor...')
        call_throws([
            'systemctl',
            'restart',
            get_unit_name(fsid, 'mon', mon_id)
        ])

    # create mgr
    logger.info('Creating mgr...')
    mgr_keyring = '[mgr.%s]\n\tkey = %s\n' % (mgr_id, mgr_key)
    mgr_c = get_container(fsid, 'mgr', mgr_id)
    deploy_daemon(fsid, 'mgr', mgr_id, mgr_c, uid, gid, config, mgr_keyring)

    # crash unit
    logger.info('Creating crash agent...')
    deploy_crash(fsid, uid, gid, config,
                 '[client.crash.%s]\n\tkey = %s\n' % (hostname, crash_key))

    # output files
    with open(args.output_keyring, 'w') as f:
        os.fchmod(f.fileno(), 0o600)
        f.write('[client.admin]\n'
                '\tkey = ' + admin_key + '\n')
    logger.info('Wrote keyring to %s' % args.output_keyring)

    with open(args.output_config, 'w') as f:
        f.write(config)
    logger.info('Wrote config to %s' % args.output_config)

    logger.info('Waiting for mgr to start...')
    def is_mgr_available():
        out = cli(['status', '-f', 'json-pretty'], timeout=30)
        j = json.loads(out)
        return j.get('mgrmap', {}).get('available', False)
    is_available('mgr', is_mgr_available)()

    # ssh
    if not args.skip_ssh:
        logger.info('Enabling cephadm module...')
        cli(['mgr', 'module', 'enable', 'cephadm'])
        logger.info('Setting orchestrator backend to cephadm...')
        cli(['orchestrator', 'set', 'backend', 'cephadm'])

        logger.info('Generating ssh key...')
        cli(['cephadm', 'generate-key'])
        ssh_pub = cli(['cephadm', 'get-pub-key'])

        with open(args.output_pub_ssh_key, 'w') as f:
            f.write(ssh_pub)
        logger.info('Wrote public SSH key to to %s' % args.output_pub_ssh_key)

        logger.info('Adding key to root@localhost\'s authorized_keys...')
        if not os.path.exists('/root/.ssh'):
            os.mkdir('/root/.ssh', 0o700)
        auth_keys_file = '/root/.ssh/authorized_keys'
        add_newline = False
        if os.path.exists(auth_keys_file):
            with open(auth_keys_file, 'r') as f:
                f.seek(0, os.SEEK_END)
                if f.tell() > 0:
                    f.seek(f.tell()-1, os.SEEK_SET) # go to last char
                    if f.read() != '\n':
                        add_newline = True
        with open(auth_keys_file, 'a') as f:
            os.fchmod(f.fileno(), 0o600)  # just in case we created it
            if add_newline:
                f.write('\n')
            f.write(ssh_pub.strip() + '\n')

        host = get_hostname()
        logger.info('Adding host %s...' % host)
        cli(['orchestrator', 'host', 'add', host])

    if not args.skip_dashboard:
        logger.info('Enabling the dashboard module...')
        cli(['mgr', 'module', 'enable', 'dashboard'])

        logger.info('Waiting for the dashboard to start...')
        def is_dashboard_available():
            out = cli(['-h'], timeout=30)
            return 'dashboard' in out
        is_available('Dashboard', is_dashboard_available)()

        logger.info('Generating a dashboard self-signed certificate...')
        cli(['dashboard', 'create-self-signed-cert'])

        logger.info('Creating initial admin user...')
        password = args.initial_dashboard_password or generate_password()
        cli(['dashboard', 'ac-user-create',
             args.initial_dashboard_user, password,
             'administrator',
             '--force-password'])
        logger.info('Fetching dashboard port number...')
        out = cli(['config', 'get', 'mgr', 'mgr/dashboard/ssl_server_port'])
        port = int(out)

        logger.info('Ceph Dashboard is now available at:\n\n'
                    '\t     URL: https://%s:%s/\n'
                    '\t    User: %s\n'
                    '\tPassword: %s\n' % (
                        get_fqdn(), port,
                        args.initial_dashboard_user,
                        password))

    logger.info('You can access the Ceph CLI with:\n\n'
                '\tsudo %s shell --fsid %s -c %s -k %s\n' % (
                    sys.argv[0],
                    fsid,
                    args.output_config,
                    args.output_keyring))

    logger.info('Bootstrap complete.')
    return 0

##################################

def command_deploy():
    # type: () -> None
    (daemon_type, daemon_id) = args.name.split('.', 1)

    l = FileLock(args.fsid)
    l.acquire()

    supported_daemons = list(Ceph.daemons)
    supported_daemons.extend(Monitoring.components)

    if daemon_type not in supported_daemons:
        raise Error('daemon type %s not recognized' % daemon_type)

    if daemon_type in Ceph.daemons:
        (config, keyring, crash_keyring) = get_config_and_both_keyrings()
        (uid, gid) = extract_uid_gid()
        make_var_run(args.fsid, uid, gid)
        c = get_container(args.fsid, daemon_type, daemon_id)
        deploy_daemon(args.fsid, daemon_type, daemon_id, c, uid, gid,
                      config, keyring,
                      osd_fsid=args.osd_fsid,
                      reconfig=args.reconfig)

        if crash_keyring and not args.reconfig:
            deploy_crash(args.fsid, uid, gid, config, crash_keyring)

    elif daemon_type in Monitoring.components:
        # monitoring daemon - prometheus, grafana, alertmanager, node-exporter
        monitoring_args = []  # type: List[str]

        # Default Checks
        daemon_port = Monitoring.port_map[daemon_type]
        if port_in_use(daemon_port):
            raise Error("TCP Port '{}' required for {} is already in use".format(daemon_port, daemon_type))
        elif args.image == DEFAULT_IMAGE:
            raise Error("--image parameter must be supplied for {}".format(daemon_type))

        if daemon_type in ['prometheus', 'grafana'] and not args.config_json:
                raise Error("config-json parameter is needed when deploying {} service".format(daemon_type))

        if daemon_type == 'prometheus':
            uid, gid = extract_uid_gid(file_path='/etc/prometheus')
        elif daemon_type == 'node-exporter':
            uid, gid = 65534, 65534
        elif daemon_type == 'grafana':
            uid, gid = extract_uid_gid(file_path='/var/lib/grafana')
        else:
            raise Error("{} not implemented yet".format(daemon_type))
        
        # Monitoring metadata is nested dicts, so asking mypy to ignore
        m = Monitoring.components[daemon_type]  # type: ignore
        metadata = m.get('image', dict())  # type: ignore
        monitoring_args = [
            '--user',
            str(uid),
            '--cpus',
            metadata.get('cpus', '2'),  # type: ignore
            '--memory',
            metadata.get('memory', '4GB')  # type: ignore
        ]

        c = get_container(args.fsid, daemon_type, daemon_id, container_args=monitoring_args)
        deploy_daemon(args.fsid, daemon_type, daemon_id, c, uid, gid,
                      reconfig=args.reconfig)
    else:
        raise Error("{} not implemented in command_deploy function".format(daemon_type))

##################################

def command_run():
    # type: () -> int
    (daemon_type, daemon_id) = args.name.split('.', 1)
    c = get_container(args.fsid, daemon_type, daemon_id)
    command = c.run_cmd()
    return call_timeout(command, args.timeout)

##################################

@infer_fsid
def command_shell():
    # type: () -> int
    if args.fsid:
        make_log_dir(args.fsid)
    if args.name:
        if '.' in args.name:
            (daemon_type, daemon_id) = args.name.split('.', 1)
        else:
            daemon_type = args.name
            daemon_id = None
    else:
        daemon_type = 'osd'  # get the most mounts
        daemon_id = None

    if daemon_id and not args.fsid:
        raise Error('must pass --fsid to specify cluster')

    mounts = get_container_mounts(args.fsid, daemon_type, daemon_id,
                                  no_config=True if args.config else False)
    if args.config:
        mounts[pathify(args.config)] = '/etc/ceph/ceph.conf:z'
    if args.keyring:
        mounts[pathify(args.keyring)] = '/etc/ceph/ceph.keyring:z'
    container_args = ['--privileged']
    if args.command:
        command = args.command
    else:
        command = ['bash']
        container_args += [
            '-it',
            '-e', 'LANG=C',
            '-e', "PS1=%s" % CUSTOM_PS1,
        ]

    c = CephContainer(
        image=args.image,
        entrypoint='doesnotmatter',
        args=[],
        container_args=container_args,
        volume_mounts=mounts)
    command = c.shell_cmd(command)

    return call_timeout(command, args.timeout)

##################################

@infer_fsid
def command_enter():
    # type: () -> int
    if not args.fsid:
        raise Error('must pass --fsid to specify cluster')
    (daemon_type, daemon_id) = args.name.split('.', 1)
    container_args = [] # type: List[str]
    if args.command:
        command = args.command
    else:
        command = ['bash']
        container_args += [
            '-it',
            '-e', 'LANG=C',
            '-e', "PS1=%s" % CUSTOM_PS1,
        ]
    c = get_container(args.fsid, daemon_type, daemon_id,
                      container_args=container_args)
    command = c.exec_cmd(command)
    return call_timeout(command, args.timeout)

##################################

@infer_fsid
def command_ceph_volume():
    # type: () -> None
    if args.fsid:
        make_log_dir(args.fsid)

    (uid, gid) = (0, 0) # ceph-volume runs as root
    mounts = get_container_mounts(args.fsid, 'osd', None)

    tmp_config = None
    tmp_keyring = None

    if args.config_and_keyring:
        # note: this will always pull from args.config_and_keyring (we
        # require it) and never args.config or args.keyring.
        (config, keyring) = get_config_and_keyring()

        # tmp keyring file
        tmp_keyring = write_tmp(keyring, uid, gid)

        # tmp config file
        tmp_config = write_tmp(config, uid, gid)

        mounts[tmp_config.name] = '/etc/ceph/ceph.conf:z'
        mounts[tmp_keyring.name] = '/var/lib/ceph/bootstrap-osd/ceph.keyring:z'

    c = CephContainer(
        image=args.image,
        entrypoint='/usr/sbin/ceph-volume',
        args=args.command,
        container_args=['--privileged'],
        volume_mounts=mounts,
    )
    out, err, code = call_throws(c.run_cmd(), verbose=True)
    if not code:
        print(out)

##################################

@infer_fsid
def command_unit():
    # type: () -> None
    if not args.fsid:
        raise Error('must pass --fsid to specify cluster')
    (daemon_type, daemon_id) = args.name.split('.', 1)
    unit_name = get_unit_name(args.fsid, daemon_type, daemon_id)
    call_throws([
        'systemctl',
        args.command,
        unit_name])

##################################

def command_ls():
    # type: () -> None
    ls = list_daemons(detail=not args.no_detail,
                      legacy_dir=args.legacy_dir)
    print(json.dumps(ls, indent=4))

def list_daemons(detail=True, legacy_dir=None):
    # type: (bool, Optional[str]) -> List[Dict[str, str]]
    host_version = None
    ls = []

    data_dir = args.data_dir
    if legacy_dir is not None:
        data_dir = os.path.abspath(legacy_dir + data_dir)

    # /var/lib/ceph
    if os.path.exists(data_dir):
        for i in os.listdir(data_dir):
            if i in ['mon', 'osd', 'mds', 'mgr']:
                daemon_type = i
                for j in os.listdir(os.path.join(data_dir, i)):
                    if '-' not in j:
                        continue
                    (cluster, daemon_id) = j.split('-', 1)
                    fsid = get_legacy_daemon_fsid(
                            cluster, daemon_type, daemon_id,
                            legacy_dir=legacy_dir)
                    i = {
                        'style': 'legacy',
                        'name': '%s.%s' % (daemon_type, daemon_id),
                        'fsid': fsid if fsid is not None else 'unknown',
                    }
                    if detail:
                        (i['enabled'], i['state']) = check_unit(
                            'ceph-%s@%s' % (daemon_type, daemon_id))
                        if not host_version:
                            try:
                                out, err, code = call(['ceph', '-v'])
                                if not code and out.startswith('ceph version '):
                                    host_version = out.split(' ')[2]
                            except Exception:
                                pass
                        i['host_version'] = host_version
                    ls.append(i)
            elif is_fsid(i):
                fsid = i
                for j in os.listdir(os.path.join(data_dir, i)):
                    if j == 'crash':
                        name = 'crash'
                        unit_name = 'ceph-%s-crash.service' % fsid
                    elif '.' in j:
                        name = j
                        (daemon_type, daemon_id) = j.split('.', 1)
                        unit_name = get_unit_name(fsid,
                                                  daemon_type,
                                                  daemon_id)
                    else:
                        continue
                    i = {
                        'style': 'cephadm:v1',
                        'name': name,
                        'fsid': fsid,
                    }
                    if detail:
                        # get container id
                        (i['enabled'], i['state']) = check_unit(unit_name)
                        container_id = None
                        image_name = None
                        image_id = None
                        version = None

                        if 'podman' in container_path and get_podman_version() < (1, 6, 2):
                            image_field = '.ImageID'
                        else:
                            image_field = '.Image'

                        out, err, code = call(
                            [
                                container_path, 'inspect',
                                '--format', '{{.Id}},{{.Config.Image}},{{%s}}' % image_field,
                                'ceph-%s-%s' % (fsid, j)
                            ],
                            verbose_on_failure=False)
                        if not code:
                            (container_id, image_name, image_id) = out.strip().split(',')
                            image_id = normalize_container_id(image_id)
                            out, err, code = call(
                                [container_path, 'exec', container_id,
                                 'ceph', '-v'])
                            if not code and out.startswith('ceph version '):
                                version = out.split(' ')[2]
                        i['container_id'] = container_id
                        i['container_image_name'] = image_name
                        i['container_image_id'] = image_id
                        i['version'] = version
                    ls.append(i)

    # /var/lib/rook
    # WRITE ME
    return ls


##################################

def command_adopt():
    # type: () -> None

    (daemon_type, daemon_id) = args.name.split('.', 1)
    (uid, gid) = extract_uid_gid()
    if args.style == 'legacy':
        fsid = get_legacy_daemon_fsid(args.cluster,
                                      daemon_type,
                                      daemon_id,
                                      legacy_dir=args.legacy_dir)
        if not fsid:
            raise Error('could not detect legacy fsid; set fsid in ceph.conf')

        l = FileLock(fsid)
        l.acquire()

        data_dir_src = ('/var/lib/ceph/%s/%s-%s' %
                        (daemon_type, args.cluster, daemon_id))
        data_dir_src = os.path.abspath(args.legacy_dir + data_dir_src)

        osd_fsid = None
        if daemon_type == 'osd':
            path = os.path.join(data_dir_src, 'fsid')
            try:
                with open(path, 'r') as f:
                    osd_fsid = f.read().strip()
            except IOError:
                raise Error('unable to read OSD fsid from %s' % path)

        # NOTE: implicit assumption here that the units correspond to the
        # cluster we are adopting based on the /etc/{defaults,sysconfig}/ceph
        # CLUSTER field.
        unit_name = 'ceph-%s@%s' % (daemon_type, daemon_id)
        (enabled, state) = check_unit(unit_name)

        if state == 'running':
            logger.info('Stopping old systemd unit %s...' % unit_name)
            call_throws(['systemctl', 'stop', unit_name])
        if enabled:
            logger.info('Disabling old systemd unit %s...' % unit_name)
            call_throws(['systemctl', 'disable', unit_name])

        # data
        logger.info('Moving data...')
        data_dir_dst = make_data_dir(fsid, daemon_type, daemon_id,
                                     uid=uid, gid=gid)
        move_files(glob(os.path.join(data_dir_src, '*')),
                   data_dir_dst,
                   uid=uid, gid=gid)
        logger.debug('Remove dir \'%s\'' % (data_dir_src))
        if os.path.ismount(data_dir_src):
            call_throws(['umount', data_dir_src])
        os.rmdir(data_dir_src)

        # config
        config_src = '/etc/ceph/%s.conf' % (args.cluster)
        config_src = os.path.abspath(args.legacy_dir + config_src)
        config_dst = os.path.join(data_dir_dst, 'config')
        copy_files([config_src], config_dst, uid=uid, gid=gid)

        # logs
        logger.info('Moving logs...')
        log_dir_src = ('/var/log/ceph/%s-%s.%s.log*' %
                        (args.cluster, daemon_type, daemon_id))
        log_dir_src = os.path.abspath(args.legacy_dir + log_dir_src)
        log_dir_dst = make_log_dir(fsid, uid=uid, gid=gid)
        move_files(glob(log_dir_src),
                   log_dir_dst,
                   uid=uid, gid=gid)

        logger.info('Creating new units...')
        make_var_run(fsid, uid, gid)
        c = get_container(fsid, daemon_type, daemon_id)
        deploy_daemon_units(fsid, uid, gid, daemon_type, daemon_id, c,
                            enable=True,  # unconditionally enable the new unit
                            start=(state == 'running'),
                            osd_fsid=osd_fsid)
        update_firewalld(daemon_type)

    else:
        raise Error('adoption of style %s not implemented' % args.style)

##################################

def command_rm_daemon():
    # type: () -> None

    l = FileLock(args.fsid)
    l.acquire()

    (daemon_type, daemon_id) = args.name.split('.', 1)
    if daemon_type in ['mon', 'osd'] and not args.force:
        raise Error('must pass --force to proceed: '
                      'this command may destroy precious data!')
    unit_name = get_unit_name(args.fsid, daemon_type, daemon_id)
    call(['systemctl', 'stop', unit_name],
         verbose_on_failure=False)
    call(['systemctl', 'reset-failed', unit_name],
         verbose_on_failure=False)
    call(['systemctl', 'disable', unit_name],
         verbose_on_failure=False)
    data_dir = get_data_dir(args.fsid, daemon_type, daemon_id)
    call_throws(['rm', '-rf', data_dir])

##################################

def command_rm_cluster():
    # type: () -> None
    if not args.force:
        raise Error('must pass --force to proceed: '
                      'this command may destroy precious data!')

    l = FileLock(args.fsid)
    l.acquire()

    # stop + disable individual daemon units
    for d in list_daemons(detail=False):
        if d['fsid'] != args.fsid:
            continue
        if d['style'] != 'cephadm:v1':
            continue
        unit_name = get_unit_name(args.fsid, d['name'])
        call(['systemctl', 'stop', unit_name],
             verbose_on_failure=False)
        call(['systemctl', 'reset-failed', unit_name],
             verbose_on_failure=False)
        call(['systemctl', 'disable', unit_name],
             verbose_on_failure=False)

    # cluster units
    for unit_name in ['ceph-%s.target' % args.fsid,
                      'ceph-%s-crash.service' % args.fsid]:
        call(['systemctl', 'stop', unit_name],
             verbose_on_failure=False)
        call(['systemctl', 'reset-failed', unit_name],
             verbose_on_failure=False)
        call(['systemctl', 'disable', unit_name],
             verbose_on_failure=False)

    slice_name = 'system-%s.slice' % (('ceph-%s' % args.fsid).replace('-',
                                                                      '\\x2d'))
    call(['systemctl', 'stop', slice_name],
         verbose_on_failure=False)

    # rm units
    call_throws(['rm', '-f', args.unit_dir +
                             '/ceph-%s@.service' % args.fsid])
    call_throws(['rm', '-f', args.unit_dir +
                             '/ceph-%s-crash.service' % args.fsid])
    call_throws(['rm', '-f', args.unit_dir +
                             '/ceph-%s.target' % args.fsid])
    call_throws(['rm', '-rf',
                  args.unit_dir + '/ceph-%s.target.wants' % args.fsid])
    # rm data
    call_throws(['rm', '-rf', args.data_dir + '/' + args.fsid])
    # rm logs
    call_throws(['rm', '-rf', args.log_dir + '/' + args.fsid])
    call_throws(['rm', '-rf', args.log_dir +
                             '/*.wants/ceph-%s@*' % args.fsid])
    # rm logrotate config
    call_throws(['rm', '-f', args.logrotate_dir + '/ceph-%s' % args.fsid])


##################################

def check_time_sync():
    units = [
        'chronyd.service', 'systemd-timesyncd.service',
        'ntpd.service', # el7 (at least)
        'ntp.service',  # 18.04 (at least)
    ]
    for u in units:
        (enabled, state) = check_unit(u)
        if enabled and state == 'running':
            logger.info('Time sync unit %s is enabled and running' % u)
            return True
    logger.warning('No time sync service is running; checked for %s' % units)
    return False

def command_check_host():
    # caller already checked for docker/podman
    logger.info('podman|docker (%s) is present' % container_path)

    if not find_program('systemctl'):
        raise RuntimeError('unable to location systemctl')
    logger.info('systemctl is present')

    if not find_program('lvcreate'):
        raise RuntimeError('LVM does not appear to be installed')
    logger.info('LVM2 is present')

    # check for configured+running chronyd or ntp
    if not check_time_sync():
        raise RuntimeError('No time synchronization is active')

    logger.info('Host looks OK')


##################################

class CustomValidation(argparse.Action):

    def _check_name(self, values):
        try:
            (daemon_type, daemon_id) = values.split('.', 1)
        except ValueError:
            raise argparse.ArgumentError(self,
                                         "must be of the format <type>.<id>. For example, osd.1 or prometheus.myhost.com")

        daemons = list(Ceph.daemons)
        daemons.extend(Monitoring.components.keys())

        if daemon_type not in daemons:
            raise argparse.ArgumentError(self,
                                         "name must declare the type of daemon e.g. "
                                         "{}".format(', '.join(daemons)))

    def __call__(self, parser, namespace, values, option_string=None):
        if self.dest == "name":
            self._check_name(values)
            setattr(namespace, self.dest, values)

##################################

def _get_parser():
    # type: () -> argparse.ArgumentParser
    parser = argparse.ArgumentParser(
        description='Bootstrap Ceph daemons with systemd and containers.',
        formatter_class=argparse.ArgumentDefaultsHelpFormatter)
    parser.add_argument(
        '--image',
        default=os.environ.get('CEPHADM_IMAGE', DEFAULT_IMAGE),
        help='container image. Can also be set via the "CEPHADM_IMAGE" '
        'env var')
    parser.add_argument(
        '--docker',
        action='store_true',
        help='use docker instead of podman')
    parser.add_argument(
        '--data-dir',
        default=DATA_DIR,
        help='base directory for daemon data')
    parser.add_argument(
        '--log-dir',
        default=LOG_DIR,
        help='base directory for daemon logs')
    parser.add_argument(
        '--logrotate-dir',
        default=LOGROTATE_DIR,
        help='location of logrotate configuration files')
    parser.add_argument(
        '--unit-dir',
        default=UNIT_DIR,
        help='base directory for systemd units')
    parser.add_argument(
        '--verbose', '-v',
        action='store_true',
        help='Show debug-level log messages')
    parser.add_argument(
        '--timeout',
        type=int,
        default=DEFAULT_TIMEOUT,
        help='timeout in seconds')

    subparsers = parser.add_subparsers(help='sub-command')

    parser_version = subparsers.add_parser(
        'version', help='get ceph version from container')
    parser_version.set_defaults(func=command_version)

    parser_pull = subparsers.add_parser(
        'pull', help='pull latest image version')
    parser_pull.set_defaults(func=command_pull)

    parser_ls = subparsers.add_parser(
        'ls', help='list daemon instances on this host')
    parser_ls.set_defaults(func=command_ls)
    parser_ls.add_argument(
        '--no-detail',
        action='store_true',
        help='Do not include daemon status')
    parser_ls.add_argument(
        '--legacy-dir',
        default='/',
        help='base directory for legacy daemon data')

    parser_adopt = subparsers.add_parser(
        'adopt', help='adopt daemon deployed with a different tool')
    parser_adopt.set_defaults(func=command_adopt)
    parser_adopt.add_argument(
        '--name', '-n',
        required=True,
        help='daemon name (type.id)')
    parser_adopt.add_argument(
        '--style',
        required=True,
        help='deployment style (legacy, ...)')
    parser_adopt.add_argument(
        '--cluster',
        default='ceph',
        help='cluster name')
    parser_adopt.add_argument(
        '--legacy-dir',
        default='/',
        help='base directory for legacy daemon data')
    parser_adopt.add_argument(
        '--skip-firewalld',
        action='store_true',
        help='Do not configure firewalld')

    parser_rm_daemon = subparsers.add_parser(
        'rm-daemon', help='remove daemon instance')
    parser_rm_daemon.set_defaults(func=command_rm_daemon)
    parser_rm_daemon.add_argument(
        '--name', '-n',
        required=True,
        action=CustomValidation,
        help='daemon name (type.id)')
    parser_rm_daemon.add_argument(
        '--fsid',
        required=True,
        help='cluster FSID')
    parser_rm_daemon.add_argument(
        '--force',
        action='store_true',
        help='proceed, even though this may destroy valuable data')

    parser_rm_cluster = subparsers.add_parser(
        'rm-cluster', help='remove all daemons for a cluster')
    parser_rm_cluster.set_defaults(func=command_rm_cluster)
    parser_rm_cluster.add_argument(
        '--fsid',
        required=True,
        help='cluster FSID')
    parser_rm_cluster.add_argument(
        '--force',
        action='store_true',
        help='proceed, even though this may destroy valuable data')

    parser_run = subparsers.add_parser(
        'run', help='run a ceph daemon, in a container, in the foreground')
    parser_run.set_defaults(func=command_run)
    parser_run.add_argument(
        '--name', '-n',
        required=True,
        help='daemon name (type.id)')
    parser_run.add_argument(
        '--fsid',
        required=True,
        help='cluster FSID')

    parser_shell = subparsers.add_parser(
        'shell', help='run an interactive shell inside a daemon container')
    parser_shell.set_defaults(func=command_shell)
    parser_shell.add_argument(
        '--fsid',
        help='cluster FSID')
    parser_shell.add_argument(
        '--name', '-n',
        help='daemon name (type.id)')
    parser_shell.add_argument(
        '--config', '-c',
        help='ceph.conf to pass through to the container')
    parser_shell.add_argument(
        '--keyring', '-k',
        help='ceph.keyring to pass through to the container')
    parser_shell.add_argument(
        'command', nargs='*',
        help='command (optional)')

    parser_enter = subparsers.add_parser(
        'enter', help='run an interactive shell inside a running daemon container')
    parser_enter.set_defaults(func=command_enter)
    parser_enter.add_argument(
        '--fsid',
        help='cluster FSID')
    parser_enter.add_argument(
        '--name', '-n',
        required=True,
        help='daemon name (type.id)')
    parser_enter.add_argument(
        'command', nargs='*',
        help='command')

    parser_ceph_volume = subparsers.add_parser(
        'ceph-volume', help='run ceph-volume inside a container')
    parser_ceph_volume.set_defaults(func=command_ceph_volume)
    parser_ceph_volume.add_argument(
        '--fsid',
        help='cluster FSID')
    parser_ceph_volume.add_argument(
        '--config-and-keyring',
        help='JSON file with config and (client.bootrap-osd) key')
    parser_ceph_volume.add_argument(
        'command', nargs='+',
        help='command')

    parser_unit = subparsers.add_parser(
        'unit', help='operate on the daemon\'s systemd unit')
    parser_unit.set_defaults(func=command_unit)
    parser_unit.add_argument(
        'command',
        help='systemd command (start, stop, restart, enable, disable, ...)')
    parser_unit.add_argument(
        '--fsid',
        help='cluster FSID')
    parser_unit.add_argument(
        '--name', '-n',
        required=True,
        help='daemon name (type.id)')

    parser_bootstrap = subparsers.add_parser(
        'bootstrap', help='bootstrap a cluster (mon + mgr daemons)')
    parser_bootstrap.set_defaults(func=command_bootstrap)
    parser_bootstrap.add_argument(
        '--config', '-c',
        help='ceph conf file to incorporate')
    parser_bootstrap.add_argument(
        '--mon-id',
        required=False,
        help='mon id (default: local hostname)')
    parser_bootstrap.add_argument(
        '--mon-addrv',
        help='mon IPs (e.g., [v2:localipaddr:3300,v1:localipaddr:6789])')
    parser_bootstrap.add_argument(
        '--mon-ip',
        help='mon IP')
    parser_bootstrap.add_argument(
        '--mgr-id',
        required=False,
        help='mgr id (default: randomly generated)')
    parser_bootstrap.add_argument(
        '--fsid',
        help='cluster FSID')
    parser_bootstrap.add_argument(
        '--output-keyring',
        default='ceph.client.admin.keyring',
        help='location to write keyring file with new cluster admin and mon keys')
    parser_bootstrap.add_argument(
        '--output-config',
        default='ceph.conf',
        help='location to write conf file to connect to new cluster')
    parser_bootstrap.add_argument(
        '--output-pub-ssh-key',
        default='ceph.pub',
        help='location to write the cluster\'s public SSH key')
    parser_bootstrap.add_argument(
        '--skip-ssh',
        action='store_true',
        help='skip setup of ssh key on local host')
    parser_bootstrap.add_argument(
        '--initial-dashboard-user',
        default='admin',
        help='Initial user for the dashboard')
    parser_bootstrap.add_argument(
        '--initial-dashboard-password',
        help='Initial password for the initial dashboard user')
    parser_bootstrap.add_argument(
        '--skip-dashboard',
        action='store_true',
        help='do not enable the Ceph Dashboard')
    parser_bootstrap.add_argument(
        '--no-minimize-config',
        action='store_true',
        help='do not assimilate and minimize the config file')
    parser_bootstrap.add_argument(
        '--skip-ping-check',
        action='store_true',
        help='do not verify that mon IP is pingable')
    parser_bootstrap.add_argument(
        '--skip-pull',
        action='store_true',
        help='do not pull the latest image before bootstrapping')
    parser_bootstrap.add_argument(
        '--skip-firewalld',
        action='store_true',
        help='Do not configure firewalld')
    parser_bootstrap.add_argument(
        '--allow-overwrite',
        action='store_true',
        help='allow overwrite of existing --output-* config/keyring/ssh files')

    parser_deploy = subparsers.add_parser(
        'deploy', help='deploy a daemon')
    parser_deploy.set_defaults(func=command_deploy)
    parser_deploy.add_argument(
        '--name',
        required=True,
        action=CustomValidation,
        help='daemon name (type.id)')
    parser_deploy.add_argument(
        '--fsid',
        required=True,
        help='cluster FSID')
    parser_deploy.add_argument(
        '--config', '-c',
        help='config file for new daemon')
    parser_deploy.add_argument(
        '--config-json',
        help='Additional configuration information in JSON format')
    parser_deploy.add_argument(
        '--keyring',
        help='keyring for new daemon')
    parser_deploy.add_argument(
        '--crash-keyring',
        help='crash keyring for crash agent daemon')
    parser_deploy.add_argument(
        '--key',
        help='key for new daemon')
    parser_deploy.add_argument(
        '--config-and-keyrings',
        help='JSON file with config and keyrings for the daemon and crash agent')
    parser_deploy.add_argument(
        '--osd-fsid',
        help='OSD uuid, if creating an OSD container')
    parser_deploy.add_argument(
        '--skip-firewalld',
        action='store_true',
        help='Do not configure firewalld')
    parser_deploy.add_argument(
        '--reconfig',
        action='store_true',
        help='Reconfigure a previously deployed daemon')

    parser_check_host = subparsers.add_parser(
        'check-host', help='check host configuration')
    parser_check_host.set_defaults(func=command_check_host)

    return parser


if __name__ == "__main__":
    # allow argv to be injected
    try:
        av = injected_argv # type: ignore
    except NameError:
        av = sys.argv[1:]
    parser = _get_parser()
    args = parser.parse_args(av)

    if args.verbose:
        logging.basicConfig(level=logging.DEBUG)
    else:
        logging.basicConfig(level=logging.INFO)
    logger = logging.getLogger('cephadm')

    # root?
    if os.geteuid() != 0:
        sys.stderr.write('ERROR: cephadm should be run as root\n')
        sys.exit(1)

    # podman or docker?
    if args.docker:
        container_path = find_program('docker')
    else:
        for i in CONTAINER_PREFERENCE:
            try:
                container_path = find_program(i)
                break
            except Exception as e:
                logger.debug('Could not locate %s: %s' % (i, e))
        if not container_path:
            sys.stderr.write('Unable to locate any of %s\n' % CONTAINER_PREFERENCE)
            sys.exit(1)

    if 'func' not in args:
        sys.stderr.write('No command specified; pass -h or --help for usage\n')
        sys.exit(1)
    try:
        r = args.func()
    except Error as e:
        if args.verbose:
            raise
        sys.stderr.write('ERROR: %s\n' % e)
        sys.exit(1)
    if not r:
        r = 0
    sys.exit(r)
