cli.py

#
#

# Copyright (C) 2006, 2007 Google Inc.
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
# 02110-1301, USA.


"""Module dealing with command line parsing"""


import sys
import textwrap
import os.path
import copy
import time
import logging
from cStringIO import StringIO

from ganeti import utils
from ganeti import errors
from ganeti import constants
from ganeti import opcodes
from ganeti import luxi
from ganeti import ssconf
from ganeti import rpc

from optparse import (OptionParser, TitledHelpFormatter,
                      Option, OptionValueError)


__all__ = ["DEBUG_OPT", "NOHDR_OPT", "SEP_OPT", "GenericMain",
           "SubmitOpCode", "GetClient",
           "cli_option",
           "GenerateTable", "AskUser",
           "USEUNITS_OPT", "FIELDS_OPT", "FORCE_OPT", "SUBMIT_OPT",
           "ListTags", "AddTags", "RemoveTags", "TAG_SRC_OPT",
           "FormatError", "SplitNodeOption", "SubmitOrSend",
           "JobSubmittedException", "FormatTimestamp", "ParseTimespec",
           "ToStderr", "ToStdout", "UsesRPC",
           "GetOnlineNodes", "JobExecutor", "SYNC_OPT", "CONFIRM_OPT",
           "ArgJobId", "ArgSuggest", "ArgUnknown", "ArgFile", "ArgCommand",
           "ArgInstance", "ArgNode", "ArgChoice",
           "ARGS_NONE", "ARGS_ONE_INSTANCE", "ARGS_ONE_NODE",
           "ARGS_MANY_INSTANCES", "ARGS_MANY_NODES",
           ]

NO_PREFIX = "no_"
UN_PREFIX = "-"


class _Argument:
  def __init__(self, min=0, max=None, suggest=None):
    self.min = min
    self.max = max

  def __repr__(self):
    return ("<%s min=%s max=%s>" %
            (self.__class__.__name__, self.min, self.max))


class ArgSuggest(_Argument):
  """Suggesting argument.

  Value can be any of the ones passed to the constructor.

  """
  def __init__(self, min=0, max=None, choices=None):
    _Argument.__init__(self, min=min, max=max)
    self.choices = choices

  def __repr__(self):
    return ("<%s min=%s max=%s choices=%r>" %
            (self.__class__.__name__, self.min, self.max, self.choices))


class ArgChoice(ArgSuggest):
  """Choice argument.

  Value can be any of the ones passed to the constructor. Like L{ArgSuggest},
  but value must be one of the choices.

  """


class ArgUnknown(_Argument):
  """Unknown argument to program (e.g. determined at runtime).

  """


class ArgInstance(_Argument):
  """Instances argument.

  """


class ArgNode(_Argument):
  """Node argument.

  """

class ArgJobId(_Argument):
  """Job ID argument.

  """


class ArgFile(_Argument):
  """File path argument.

  """


class ArgCommand(_Argument):
  """Command argument.

  """


ARGS_NONE = []
ARGS_MANY_INSTANCES = [ArgInstance()]
ARGS_MANY_NODES = [ArgNode()]
ARGS_ONE_INSTANCE = [ArgInstance(min=1, max=1)]
ARGS_ONE_NODE = [ArgNode(min=1, max=1)]


def _ExtractTagsObject(opts, args):
  """Extract the tag type object.

  Note that this function will modify its args parameter.

  """
  if not hasattr(opts, "tag_type"):
    raise errors.ProgrammerError("tag_type not passed to _ExtractTagsObject")
  kind = opts.tag_type
  if kind == constants.TAG_CLUSTER:
    retval = kind, kind
  elif kind == constants.TAG_NODE or kind == constants.TAG_INSTANCE:
    if not args:
      raise errors.OpPrereqError("no arguments passed to the command")
    name = args.pop(0)
    retval = kind, name
  else:
    raise errors.ProgrammerError("Unhandled tag type '%s'" % kind)
  return retval


def _ExtendTags(opts, args):
  """Extend the args if a source file has been given.

  This function will extend the tags with the contents of the file
  passed in the 'tags_source' attribute of the opts parameter. A file
  named '-' will be replaced by stdin.

  """
  fname = opts.tags_source
  if fname is None:
    return
  if fname == "-":
    new_fh = sys.stdin
  else:
    new_fh = open(fname, "r")
  new_data = []
  try:
    # we don't use the nice 'new_data = [line.strip() for line in fh]'
    # because of python bug 1633941
    while True:
      line = new_fh.readline()
      if not line:
        break
      new_data.append(line.strip())
  finally:
    new_fh.close()
  args.extend(new_data)


def ListTags(opts, args):
  """List the tags on a given object.

  This is a generic implementation that knows how to deal with all
  three cases of tag objects (cluster, node, instance). The opts
  argument is expected to contain a tag_type field denoting what
  object type we work on.

  """
  kind, name = _ExtractTagsObject(opts, args)
  op = opcodes.OpGetTags(kind=kind, name=name)
  result = SubmitOpCode(op)
  result = list(result)
  result.sort()
  for tag in result:
    ToStdout(tag)


def AddTags(opts, args):
  """Add tags on a given object.

  This is a generic implementation that knows how to deal with all
  three cases of tag objects (cluster, node, instance). The opts
  argument is expected to contain a tag_type field denoting what
  object type we work on.

  """
  kind, name = _ExtractTagsObject(opts, args)
  _ExtendTags(opts, args)
  if not args:
    raise errors.OpPrereqError("No tags to be added")
  op = opcodes.OpAddTags(kind=kind, name=name, tags=args)
  SubmitOpCode(op)


def RemoveTags(opts, args):
  """Remove tags from a given object.

  This is a generic implementation that knows how to deal with all
  three cases of tag objects (cluster, node, instance). The opts
  argument is expected to contain a tag_type field denoting what
  object type we work on.

  """
  kind, name = _ExtractTagsObject(opts, args)
  _ExtendTags(opts, args)
  if not args:
    raise errors.OpPrereqError("No tags to be removed")
  op = opcodes.OpDelTags(kind=kind, name=name, tags=args)
  SubmitOpCode(op)


def check_unit(option, opt, value):
  """OptParsers custom converter for units.

  """
  try:
    return utils.ParseUnit(value)
  except errors.UnitParseError, err:
    raise OptionValueError("option %s: %s" % (opt, err))


def _SplitKeyVal(opt, data):
  """Convert a KeyVal string into a dict.

  This function will convert a key=val[,...] string into a dict. Empty
  values will be converted specially: keys which have the prefix 'no_'
  will have the value=False and the prefix stripped, the others will
  have value=True.

  @type opt: string
  @param opt: a string holding the option name for which we process the
      data, used in building error messages
  @type data: string
  @param data: a string of the format key=val,key=val,...
  @rtype: dict
  @return: {key=val, key=val}
  @raises errors.ParameterError: if there are duplicate keys

  """
  kv_dict = {}
  if data:
    for elem in data.split(","):
      if "=" in elem:
        key, val = elem.split("=", 1)
      else:
        if elem.startswith(NO_PREFIX):
          key, val = elem[len(NO_PREFIX):], False
        elif elem.startswith(UN_PREFIX):
          key, val = elem[len(UN_PREFIX):], None
        else:
          key, val = elem, True
      if key in kv_dict:
        raise errors.ParameterError("Duplicate key '%s' in option %s" %
                                    (key, opt))
      kv_dict[key] = val
  return kv_dict


def check_ident_key_val(option, opt, value):
  """Custom parser for ident:key=val,key=val options.

  This will store the parsed values as a tuple (ident, {key: val}). As such,
  multiple uses of this option via action=append is possible.

  """
  if ":" not in value:
    ident, rest = value, ''
  else:
    ident, rest = value.split(":", 1)

  if ident.startswith(NO_PREFIX):
    if rest:
      msg = "Cannot pass options when removing parameter groups: %s" % value
      raise errors.ParameterError(msg)
    retval = (ident[len(NO_PREFIX):], False)
  elif ident.startswith(UN_PREFIX):
    if rest:
      msg = "Cannot pass options when removing parameter groups: %s" % value
      raise errors.ParameterError(msg)
    retval = (ident[len(UN_PREFIX):], None)
  else:
    kv_dict = _SplitKeyVal(opt, rest)
    retval = (ident, kv_dict)
  return retval


def check_key_val(option, opt, value):
  """Custom parser class for key=val,key=val options.

  This will store the parsed values as a dict {key: val}.

  """
  return _SplitKeyVal(opt, value)


class CliOption(Option):
  """Custom option class for optparse.

  """
  ATTRS = Option.ATTRS + [
    "completion_suggest",
    ]
  TYPES = Option.TYPES + (
    "identkeyval",
    "keyval",
    "unit",
    )
  TYPE_CHECKER = Option.TYPE_CHECKER.copy()
  TYPE_CHECKER["identkeyval"] = check_ident_key_val
  TYPE_CHECKER["keyval"] = check_key_val
  TYPE_CHECKER["unit"] = check_unit


# optparse.py sets make_option, so we do it for our own option class, too
cli_option = CliOption


DEBUG_OPT = cli_option("-d", "--debug", default=False,
                       action="store_true",
                       help="Turn debugging on")

NOHDR_OPT = cli_option("--no-headers", default=False,
                       action="store_true", dest="no_headers",
                       help="Don't display column headers")

SEP_OPT = cli_option("--separator", default=None,
                     action="store", dest="separator",
                     help=("Separator between output fields"
                           " (defaults to one space)"))

USEUNITS_OPT = cli_option("--units", default=None,
                          dest="units", choices=('h', 'm', 'g', 't'),
                          help="Specify units for output (one of hmgt)")

FIELDS_OPT = cli_option("-o", "--output", dest="output", action="store",
                        type="string", metavar="FIELDS",
                        help="Comma separated list of output fields")

FORCE_OPT = cli_option("-f", "--force", dest="force", action="store_true",
                       default=False, help="Force the operation")

CONFIRM_OPT = cli_option("--yes", dest="confirm", action="store_true",
                         default=False, help="Do not require confirmation")

TAG_SRC_OPT = cli_option("--from", dest="tags_source",
                         default=None, help="File with tag names")

SUBMIT_OPT = cli_option("--submit", dest="submit_only",
                        default=False, action="store_true",
                        help=("Submit the job and return the job ID, but"
                              " don't wait for the job to finish"))

SYNC_OPT = cli_option("--sync", dest="do_locking",
                      default=False, action="store_true",
                      help=("Grab locks while doing the queries"
                            " in order to ensure more consistent results"))

_DRY_RUN_OPT = cli_option("--dry-run", default=False,
                          action="store_true",
                          help=("Do not execute the operation, just run the"
                                " check steps and verify it it could be"
                                " executed"))


def _ParseArgs(argv, commands, aliases):
  """Parser for the command line arguments.

  This function parses the arguments and returns the function which
  must be executed together with its (modified) arguments.

  @param argv: the command line
  @param commands: dictionary with special contents, see the design
      doc for cmdline handling
  @param aliases: dictionary with command aliases {'alias': 'target, ...}

  """
  if len(argv) == 0:
    binary = "<command>"
  else:
    binary = argv[0].split("/")[-1]

  if len(argv) > 1 and argv[1] == "--version":
    ToStdout("%s (ganeti) %s", binary, constants.RELEASE_VERSION)
    # Quit right away. That way we don't have to care about this special
    # argument. optparse.py does it the same.
    sys.exit(0)

  if len(argv) < 2 or not (argv[1] in commands or
                           argv[1] in aliases):
    # let's do a nice thing
    sortedcmds = commands.keys()
    sortedcmds.sort()

    ToStdout("Usage: %s {command} [options...] [argument...]", binary)
    ToStdout("%s <command> --help to see details, or man %s", binary, binary)
    ToStdout("")

    # compute the max line length for cmd + usage
    mlen = max([len(" %s" % cmd) for cmd in commands])
    mlen = min(60, mlen) # should not get here...

    # and format a nice command list
    ToStdout("Commands:")
    for cmd in sortedcmds:
      cmdstr = " %s" % (cmd,)
      help_text = commands[cmd][4]
      help_lines = textwrap.wrap(help_text, 79 - 3 - mlen)
      ToStdout("%-*s - %s", mlen, cmdstr, help_lines.pop(0))
      for line in help_lines:
        ToStdout("%-*s   %s", mlen, "", line)

    ToStdout("")

    return None, None, None

  # get command, unalias it, and look it up in commands
  cmd = argv.pop(1)
  if cmd in aliases:
    if cmd in commands:
      raise errors.ProgrammerError("Alias '%s' overrides an existing"
                                   " command" % cmd)

    if aliases[cmd] not in commands:
      raise errors.ProgrammerError("Alias '%s' maps to non-existing"
                                   " command '%s'" % (cmd, aliases[cmd]))

    cmd = aliases[cmd]

  func, args_def, parser_opts, usage, description = commands[cmd]
  parser = OptionParser(option_list=parser_opts + [_DRY_RUN_OPT],
                        description=description,
                        formatter=TitledHelpFormatter(),
                        usage="%%prog %s %s" % (cmd, usage))
  parser.disable_interspersed_args()
  options, args = parser.parse_args()

  if not _CheckArguments(cmd, args_def, args):
    return None, None, None

  return func, options, args


def _CheckArguments(cmd, args_def, args):
  """Verifies the arguments using the argument definition.

  Algorithm:

    1. Abort with error if values specified by user but none expected.

    1. For each argument in definition

      1. Keep running count of minimum number of values (min_count)
      1. Keep running count of maximum number of values (max_count)
      1. If it has an unlimited number of values

        1. Abort with error if it's not the last argument in the definition

    1. If last argument has limited number of values

      1. Abort with error if number of values doesn't match or is too large

    1. Abort with error if user didn't pass enough values (min_count)

  """
  if args and not args_def:
    ToStderr("Error: Command %s expects no arguments", cmd)
    return False

  min_count = None
  max_count = None
  check_max = None

  last_idx = len(args_def) - 1

  for idx, arg in enumerate(args_def):
    if min_count is None:
      min_count = arg.min
    elif arg.min is not None:
      min_count += arg.min

    if max_count is None:
      max_count = arg.max
    elif arg.max is not None:
      max_count += arg.max

    if idx == last_idx:
      check_max = (arg.max is not None)

    elif arg.max is None:
      raise errors.ProgrammerError("Only the last argument can have max=None")

  if check_max:
    # Command with exact number of arguments
    if (min_count is not None and max_count is not None and
        min_count == max_count and len(args) != min_count):
      ToStderr("Error: Command %s expects %d argument(s)", cmd, min_count)
      return False

    # Command with limited number of arguments
    if max_count is not None and len(args) > max_count:
      ToStderr("Error: Command %s expects only %d argument(s)",
               cmd, max_count)
      return False

  # Command with some required arguments
  if min_count is not None and len(args) < min_count:
    ToStderr("Error: Command %s expects at least %d argument(s)",
             cmd, min_count)
    return False

  return True


def SplitNodeOption(value):
  """Splits the value of a --node option.

  """
  if value and ':' in value:
    return value.split(':', 1)
  else:
    return (value, None)


def UsesRPC(fn):
  def wrapper(*args, **kwargs):
    rpc.Init()
    try:
      return fn(*args, **kwargs)
    finally:
      rpc.Shutdown()
  return wrapper


def AskUser(text, choices=None):
  """Ask the user a question.

  @param text: the question to ask

  @param choices: list with elements tuples (input_char, return_value,
      description); if not given, it will default to: [('y', True,
      'Perform the operation'), ('n', False, 'Do no do the operation')];
      note that the '?' char is reserved for help

  @return: one of the return values from the choices list; if input is
      not possible (i.e. not running with a tty, we return the last
      entry from the list

  """
  if choices is None:
    choices = [('y', True, 'Perform the operation'),
               ('n', False, 'Do not perform the operation')]
  if not choices or not isinstance(choices, list):
    raise errors.ProgrammerError("Invalid choices argument to AskUser")
  for entry in choices:
    if not isinstance(entry, tuple) or len(entry) < 3 or entry[0] == '?':
      raise errors.ProgrammerError("Invalid choices element to AskUser")

  answer = choices[-1][1]
  new_text = []
  for line in text.splitlines():
    new_text.append(textwrap.fill(line, 70, replace_whitespace=False))
  text = "\n".join(new_text)
  try:
    f = file("/dev/tty", "a+")
  except IOError:
    return answer
  try:
    chars = [entry[0] for entry in choices]
    chars[-1] = "[%s]" % chars[-1]
    chars.append('?')
    maps = dict([(entry[0], entry[1]) for entry in choices])
    while True:
      f.write(text)
      f.write('\n')
      f.write("/".join(chars))
      f.write(": ")
      line = f.readline(2).strip().lower()
      if line in maps:
        answer = maps[line]
        break
      elif line == '?':
        for entry in choices:
          f.write(" %s - %s\n" % (entry[0], entry[2]))
        f.write("\n")
        continue
  finally:
    f.close()
  return answer


class JobSubmittedException(Exception):
  """Job was submitted, client should exit.

  This exception has one argument, the ID of the job that was
  submitted. The handler should print this ID.

  This is not an error, just a structured way to exit from clients.

  """


def SendJob(ops, cl=None):
  """Function to submit an opcode without waiting for the results.

  @type ops: list
  @param ops: list of opcodes
  @type cl: luxi.Client
  @param cl: the luxi client to use for communicating with the master;
             if None, a new client will be created

  """
  if cl is None:
    cl = GetClient()

  job_id = cl.SubmitJob(ops)

  return job_id


def PollJob(job_id, cl=None, feedback_fn=None):
  """Function to poll for the result of a job.

  @type job_id: job identified
  @param job_id: the job to poll for results
  @type cl: luxi.Client
  @param cl: the luxi client to use for communicating with the master;
             if None, a new client will be created

  """
  if cl is None:
    cl = GetClient()

  prev_job_info = None
  prev_logmsg_serial = None

  while True:
    result = cl.WaitForJobChange(job_id, ["status"], prev_job_info,
                                 prev_logmsg_serial)
    if not result:
      # job not found, go away!
      raise errors.JobLost("Job with id %s lost" % job_id)

    # Split result, a tuple of (field values, log entries)
    (job_info, log_entries) = result
    (status, ) = job_info

    if log_entries:
      for log_entry in log_entries:
        (serial, timestamp, _, message) = log_entry
        if callable(feedback_fn):
          feedback_fn(log_entry[1:])
        else:
          encoded = utils.SafeEncode(message)
          ToStdout("%s %s", time.ctime(utils.MergeTime(timestamp)), encoded)
        prev_logmsg_serial = max(prev_logmsg_serial, serial)

    # TODO: Handle canceled and archived jobs
    elif status in (constants.JOB_STATUS_SUCCESS,
                    constants.JOB_STATUS_ERROR,
                    constants.JOB_STATUS_CANCELING,
                    constants.JOB_STATUS_CANCELED):
      break

    prev_job_info = job_info

  jobs = cl.QueryJobs([job_id], ["status", "opstatus", "opresult"])
  if not jobs:
    raise errors.JobLost("Job with id %s lost" % job_id)

  status, opstatus, result = jobs[0]
  if status == constants.JOB_STATUS_SUCCESS:
    return result
  elif status in (constants.JOB_STATUS_CANCELING,
                  constants.JOB_STATUS_CANCELED):
    raise errors.OpExecError("Job was canceled")
  else:
    has_ok = False
    for idx, (status, msg) in enumerate(zip(opstatus, result)):
      if status == constants.OP_STATUS_SUCCESS:
        has_ok = True
      elif status == constants.OP_STATUS_ERROR:
        if has_ok:
          raise errors.OpExecError("partial failure (opcode %d): %s" %
                                   (idx, msg))
        else:
          raise errors.OpExecError(str(msg))
    # default failure mode
    raise errors.OpExecError(result)


def SubmitOpCode(op, cl=None, feedback_fn=None):
  """Legacy function to submit an opcode.

  This is just a simple wrapper over the construction of the processor
  instance. It should be extended to better handle feedback and
  interaction functions.

  """
  if cl is None:
    cl = GetClient()

  job_id = SendJob([op], cl)

  op_results = PollJob(job_id, cl=cl, feedback_fn=feedback_fn)

  return op_results[0]


def SubmitOrSend(op, opts, cl=None, feedback_fn=None):
  """Wrapper around SubmitOpCode or SendJob.

  This function will decide, based on the 'opts' parameter, whether to
  submit and wait for the result of the opcode (and return it), or
  whether to just send the job and print its identifier. It is used in
  order to simplify the implementation of the '--submit' option.

  It will also add the dry-run parameter from the options passed, if true.

  """
  if opts and opts.dry_run:
    op.dry_run = opts.dry_run
  if opts and opts.submit_only:
    job_id = SendJob([op], cl=cl)
    raise JobSubmittedException(job_id)
  else:
    return SubmitOpCode(op, cl=cl, feedback_fn=feedback_fn)


def GetClient():
  # TODO: Cache object?
  try:
    client = luxi.Client()
  except luxi.NoMasterError:
    master, myself = ssconf.GetMasterAndMyself()
    if master != myself:
      raise errors.OpPrereqError("This is not the master node, please connect"
                                 " to node '%s' and rerun the command" %
                                 master)
    else:
      raise
  return client


def FormatError(err):
  """Return a formatted error message for a given error.

  This function takes an exception instance and returns a tuple
  consisting of two values: first, the recommended exit code, and
  second, a string describing the error message (not
  newline-terminated).

  """
  retcode = 1
  obuf = StringIO()
  msg = str(err)
  if isinstance(err, errors.ConfigurationError):
    txt = "Corrupt configuration file: %s" % msg
    logging.error(txt)
    obuf.write(txt + "\n")
    obuf.write("Aborting.")
    retcode = 2
  elif isinstance(err, errors.HooksAbort):
    obuf.write("Failure: hooks execution failed:\n")
    for node, script, out in err.args[0]:
      if out:
        obuf.write("  node: %s, script: %s, output: %s\n" %
                   (node, script, out))
      else:
        obuf.write("  node: %s, script: %s (no output)\n" %
                   (node, script))
  elif isinstance(err, errors.HooksFailure):
    obuf.write("Failure: hooks general failure: %s" % msg)
  elif isinstance(err, errors.ResolverError):
    this_host = utils.HostInfo.SysName()
    if err.args[0] == this_host:
      msg = "Failure: can't resolve my own hostname ('%s')"
    else:
      msg = "Failure: can't resolve hostname '%s'"
    obuf.write(msg % err.args[0])
  elif isinstance(err, errors.OpPrereqError):
    obuf.write("Failure: prerequisites not met for this"
               " operation:\n%s" % msg)
  elif isinstance(err, errors.OpExecError):
    obuf.write("Failure: command execution error:\n%s" % msg)
  elif isinstance(err, errors.TagError):
    obuf.write("Failure: invalid tag(s) given:\n%s" % msg)
  elif isinstance(err, errors.JobQueueDrainError):
    obuf.write("Failure: the job queue is marked for drain and doesn't"
               " accept new requests\n")
  elif isinstance(err, errors.JobQueueFull):
    obuf.write("Failure: the job queue is full and doesn't accept new"
               " job submissions until old jobs are archived\n")
  elif isinstance(err, errors.TypeEnforcementError):
    obuf.write("Parameter Error: %s" % msg)
  elif isinstance(err, errors.ParameterError):
    obuf.write("Failure: unknown/wrong parameter name '%s'" % msg)
  elif isinstance(err, errors.GenericError):
    obuf.write("Unhandled Ganeti error: %s" % msg)
  elif isinstance(err, luxi.NoMasterError):
    obuf.write("Cannot communicate with the master daemon.\nIs it running"
               " and listening for connections?")
  elif isinstance(err, luxi.TimeoutError):
    obuf.write("Timeout while talking to the master daemon. Error:\n"
               "%s" % msg)
  elif isinstance(err, luxi.ProtocolError):
    obuf.write("Unhandled protocol error while talking to the master daemon:\n"
               "%s" % msg)
  elif isinstance(err, JobSubmittedException):
    obuf.write("JobID: %s\n" % err.args[0])
    retcode = 0
  else:
    obuf.write("Unhandled exception: %s" % msg)
  return retcode, obuf.getvalue().rstrip('\n')


def GenericMain(commands, override=None, aliases=None):
  """Generic main function for all the gnt-* commands.

  Arguments:
    - commands: a dictionary with a special structure, see the design doc
                for command line handling.
    - override: if not None, we expect a dictionary with keys that will
                override command line options; this can be used to pass
                options from the scripts to generic functions
    - aliases: dictionary with command aliases {'alias': 'target, ...}

  """
  # save the program name and the entire command line for later logging
  if sys.argv:
    binary = os.path.basename(sys.argv[0]) or sys.argv[0]
    if len(sys.argv) >= 2:
      binary += " " + sys.argv[1]
      old_cmdline = " ".join(sys.argv[2:])
    else:
      old_cmdline = ""
  else:
    binary = "<unknown program>"
    old_cmdline = ""

  if aliases is None:
    aliases = {}

  try:
    func, options, args = _ParseArgs(sys.argv, commands, aliases)
  except errors.ParameterError, err:
    result, err_msg = FormatError(err)
    ToStderr(err_msg)
    return 1

  if func is None: # parse error
    return 1

  if override is not None:
    for key, val in override.iteritems():
      setattr(options, key, val)

  utils.SetupLogging(constants.LOG_COMMANDS, debug=options.debug,
                     stderr_logging=True, program=binary)

  if old_cmdline:
    logging.info("run with arguments '%s'", old_cmdline)
  else:
    logging.info("run with no arguments")

  try:
    result = func(options, args)
  except (errors.GenericError, luxi.ProtocolError,
          JobSubmittedException), err:
    result, err_msg = FormatError(err)
    logging.exception("Error during command processing")
    ToStderr(err_msg)

  return result


def GenerateTable(headers, fields, separator, data,
                  numfields=None, unitfields=None,
                  units=None):
  """Prints a table with headers and different fields.

  @type headers: dict
  @param headers: dictionary mapping field names to headers for
      the table
  @type fields: list
  @param fields: the field names corresponding to each row in
      the data field
  @param separator: the separator to be used; if this is None,
      the default 'smart' algorithm is used which computes optimal
      field width, otherwise just the separator is used between
      each field
  @type data: list
  @param data: a list of lists, each sublist being one row to be output
  @type numfields: list
  @param numfields: a list with the fields that hold numeric
      values and thus should be right-aligned
  @type unitfields: list
  @param unitfields: a list with the fields that hold numeric
      values that should be formatted with the units field
  @type units: string or None
  @param units: the units we should use for formatting, or None for
      automatic choice (human-readable for non-separator usage, otherwise
      megabytes); this is a one-letter string

  """
  if units is None:
    if separator:
      units = "m"
    else:
      units = "h"

  if numfields is None:
    numfields = []
  if unitfields is None:
    unitfields = []

  numfields = utils.FieldSet(*numfields)
  unitfields = utils.FieldSet(*unitfields)

  format_fields = []
  for field in fields:
    if headers and field not in headers:
      # TODO: handle better unknown fields (either revert to old
      # style of raising exception, or deal more intelligently with
      # variable fields)
      headers[field] = field
    if separator is not None:
      format_fields.append("%s")
    elif numfields.Matches(field):
      format_fields.append("%*s")
    else:
      format_fields.append("%-*s")

  if separator is None:
    mlens = [0 for name in fields]
    format = ' '.join(format_fields)
  else:
    format = separator.replace("%", "%%").join(format_fields)

  for row in data:
    if row is None:
      continue
    for idx, val in enumerate(row):
      if unitfields.Matches(fields[idx]):
        try:
          val = int(val)
        except ValueError:
          pass
        else:
          val = row[idx] = utils.FormatUnit(val, units)
      val = row[idx] = str(val)
      if separator is None:
        mlens[idx] = max(mlens[idx], len(val))

  result = []
  if headers:
    args = []
    for idx, name in enumerate(fields):
      hdr = headers[name]
      if separator is None:
        mlens[idx] = max(mlens[idx], len(hdr))
        args.append(mlens[idx])
      args.append(hdr)
    result.append(format % tuple(args))

  for line in data:
    args = []
    if line is None:
      line = ['-' for _ in fields]
    for idx in xrange(len(fields)):
      if separator is None:
        args.append(mlens[idx])
      args.append(line[idx])
    result.append(format % tuple(args))

  return result


def FormatTimestamp(ts):
  """Formats a given timestamp.

  @type ts: timestamp
  @param ts: a timeval-type timestamp, a tuple of seconds and microseconds

  @rtype: string
  @return: a string with the formatted timestamp

  """
  if not isinstance (ts, (tuple, list)) or len(ts) != 2:
    return '?'
  sec, usec = ts
  return time.strftime("%F %T", time.localtime(sec)) + ".%06d" % usec


def ParseTimespec(value):
  """Parse a time specification.

  The following suffixed will be recognized:

    - s: seconds
    - m: minutes
    - h: hours
    - d: day
    - w: weeks

  Without any suffix, the value will be taken to be in seconds.

  """
  value = str(value)
  if not value:
    raise errors.OpPrereqError("Empty time specification passed")
  suffix_map = {
    's': 1,
    'm': 60,
    'h': 3600,
    'd': 86400,
    'w': 604800,
    }
  if value[-1] not in suffix_map:
    try:
      value = int(value)
    except ValueError:
      raise errors.OpPrereqError("Invalid time specification '%s'" % value)
  else:
    multiplier = suffix_map[value[-1]]
    value = value[:-1]
    if not value: # no data left after stripping the suffix
      raise errors.OpPrereqError("Invalid time specification (only"
                                 " suffix passed)")
    try:
      value = int(value) * multiplier
    except ValueError:
      raise errors.OpPrereqError("Invalid time specification '%s'" % value)
  return value


def GetOnlineNodes(nodes, cl=None, nowarn=False):
  """Returns the names of online nodes.

  This function will also log a warning on stderr with the names of
  the online nodes.

  @param nodes: if not empty, use only this subset of nodes (minus the
      offline ones)
  @param cl: if not None, luxi client to use
  @type nowarn: boolean
  @param nowarn: by default, this function will output a note with the
      offline nodes that are skipped; if this parameter is True the
      note is not displayed

  """
  if cl is None:
    cl = GetClient()

  result = cl.QueryNodes(names=nodes, fields=["name", "offline"],
                         use_locking=False)
  offline = [row[0] for row in result if row[1]]
  if offline and not nowarn:
    ToStderr("Note: skipping offline node(s): %s" % ", ".join(offline))
  return [row[0] for row in result if not row[1]]


def _ToStream(stream, txt, *args):
  """Write a message to a stream, bypassing the logging system

  @type stream: file object
  @param stream: the file to which we should write
  @type txt: str
  @param txt: the message

  """
  if args:
    args = tuple(args)
    stream.write(txt % args)
  else:
    stream.write(txt)
  stream.write('\n')
  stream.flush()


def ToStdout(txt, *args):
  """Write a message to stdout only, bypassing the logging system

  This is just a wrapper over _ToStream.

  @type txt: str
  @param txt: the message

  """
  _ToStream(sys.stdout, txt, *args)


def ToStderr(txt, *args):
  """Write a message to stderr only, bypassing the logging system

  This is just a wrapper over _ToStream.

  @type txt: str
  @param txt: the message

  """
  _ToStream(sys.stderr, txt, *args)


class JobExecutor(object):
  """Class which manages the submission and execution of multiple jobs.

  Note that instances of this class should not be reused between
  GetResults() calls.

  """
  def __init__(self, cl=None, verbose=True):
    self.queue = []
    if cl is None:
      cl = GetClient()
    self.cl = cl
    self.verbose = verbose
    self.jobs = []

  def QueueJob(self, name, *ops):
    """Record a job for later submit.

    @type name: string
    @param name: a description of the job, will be used in WaitJobSet
    """
    self.queue.append((name, ops))

  def SubmitPending(self):
    """Submit all pending jobs.

    """
    results = self.cl.SubmitManyJobs([row[1] for row in self.queue])
    for ((status, data), (name, _)) in zip(results, self.queue):
      self.jobs.append((status, data, name))

  def GetResults(self):
    """Wait for and return the results of all jobs.

    @rtype: list
    @return: list of tuples (success, job results), in the same order
        as the submitted jobs; if a job has failed, instead of the result
        there will be the error message

    """
    if not self.jobs:
      self.SubmitPending()
    results = []
    if self.verbose:
      ok_jobs = [row[1] for row in self.jobs if row[0]]
      if ok_jobs:
        ToStdout("Submitted jobs %s", ", ".join(ok_jobs))
    for submit_status, jid, name in self.jobs:
      if not submit_status:
        ToStderr("Failed to submit job for %s: %s", name, jid)
        results.append((False, jid))
        continue
      if self.verbose:
        ToStdout("Waiting for job %s for %s...", jid, name)
      try:
        job_result = PollJob(jid, cl=self.cl)
        success = True
      except (errors.GenericError, luxi.ProtocolError), err:
        _, job_result = FormatError(err)
        success = False
        # the error message will always be shown, verbose or not
        ToStderr("Job %s for %s has failed: %s", jid, name, job_result)

      results.append((success, job_result))
    return results

  def WaitOrShow(self, wait):
    """Wait for job results or only print the job IDs.

    @type wait: boolean
    @param wait: whether to wait or not

    """
    if wait:
      return self.GetResults()
    else:
      if not self.jobs:
        self.SubmitPending()
      for status, result, name in self.jobs:
        if status:
          ToStdout("%s: %s", result, name)
        else:
          ToStderr("Failure for %s: %s", name, result)