4 # Copyright (C) 2006, 2007, 2010 Google Inc.
6 # This program is free software; you can redistribute it and/or modify
7 # it under the terms of the GNU General Public License as published by
8 # the Free Software Foundation; either version 2 of the License, or
9 # (at your option) any later version.
11 # This program is distributed in the hope that it will be useful, but
12 # WITHOUT ANY WARRANTY; without even the implied warranty of
13 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 # General Public License for more details.
16 # You should have received a copy of the GNU General Public License
17 # along with this program; if not, write to the Free Software
18 # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
22 """Ganeti utility module.
24 This module holds functions that can be used in both daemons (all) and
25 the command line scripts.
45 import logging.handlers
53 from cStringIO import StringIO
56 # pylint: disable-msg=F0401
61 from ganeti import errors
62 from ganeti import constants
63 from ganeti import compat
64 from ganeti import netutils
68 _re_shell_unquoted = re.compile('^[-.,=:/_+@A-Za-z0-9]+$')
72 #: when set to True, L{RunCmd} is disabled
75 _RANDOM_UUID_FILE = "/proc/sys/kernel/random/uuid"
77 HEX_CHAR_RE = r"[a-zA-Z0-9]"
78 VALID_X509_SIGNATURE_SALT = re.compile("^%s+$" % HEX_CHAR_RE, re.S)
79 X509_SIGNATURE = re.compile(r"^%s:\s*(?P<salt>%s+)/(?P<sign>%s+)$" %
80 (re.escape(constants.X509_CERT_SIGNATURE_HEADER),
81 HEX_CHAR_RE, HEX_CHAR_RE),
84 _VALID_SERVICE_NAME_RE = re.compile("^[-_.a-zA-Z0-9]{1,128}$")
86 # Certificate verification results
88 CERT_ERROR) = range(1, 3)
90 # Flags for mlockall() (from bits/mman.h)
95 class RunResult(object):
96 """Holds the result of running external programs.
99 @ivar exit_code: the exit code of the program, or None (if the program
101 @type signal: int or None
102 @ivar signal: the signal that caused the program to finish, or None
103 (if the program wasn't terminated by a signal)
105 @ivar stdout: the standard output of the program
107 @ivar stderr: the standard error of the program
108 @type failed: boolean
109 @ivar failed: True in case the program was
110 terminated by a signal or exited with a non-zero exit code
111 @ivar fail_reason: a string detailing the termination reason
114 __slots__ = ["exit_code", "signal", "stdout", "stderr",
115 "failed", "fail_reason", "cmd"]
118 def __init__(self, exit_code, signal_, stdout, stderr, cmd):
120 self.exit_code = exit_code
121 self.signal = signal_
124 self.failed = (signal_ is not None or exit_code != 0)
126 if self.signal is not None:
127 self.fail_reason = "terminated by signal %s" % self.signal
128 elif self.exit_code is not None:
129 self.fail_reason = "exited with exit code %s" % self.exit_code
131 self.fail_reason = "unable to determine termination reason"
134 logging.debug("Command '%s' failed (%s); output: %s",
135 self.cmd, self.fail_reason, self.output)
137 def _GetOutput(self):
138 """Returns the combined stdout and stderr for easier usage.
141 return self.stdout + self.stderr
143 output = property(_GetOutput, None, None, "Return full output")
146 def _BuildCmdEnvironment(env, reset):
147 """Builds the environment for an external program.
153 cmd_env = os.environ.copy()
154 cmd_env["LC_ALL"] = "C"
162 def RunCmd(cmd, env=None, output=None, cwd="/", reset_env=False):
163 """Execute a (shell) command.
165 The command should not read from its standard input, as it will be
168 @type cmd: string or list
169 @param cmd: Command to run
171 @param env: Additional environment variables
173 @param output: if desired, the output of the command can be
174 saved in a file instead of the RunResult instance; this
175 parameter denotes the file name (if not None)
177 @param cwd: if specified, will be used as the working
178 directory for the command; the default will be /
179 @type reset_env: boolean
180 @param reset_env: whether to reset or keep the default os environment
182 @return: RunResult instance
183 @raise errors.ProgrammerError: if we call this when forks are disabled
187 raise errors.ProgrammerError("utils.RunCmd() called with fork() disabled")
189 if isinstance(cmd, basestring):
193 cmd = [str(val) for val in cmd]
194 strcmd = ShellQuoteArgs(cmd)
198 logging.debug("RunCmd %s, output file '%s'", strcmd, output)
200 logging.debug("RunCmd %s", strcmd)
202 cmd_env = _BuildCmdEnvironment(env, reset_env)
206 out, err, status = _RunCmdPipe(cmd, cmd_env, shell, cwd)
208 status = _RunCmdFile(cmd, cmd_env, shell, output, cwd)
211 if err.errno == errno.ENOENT:
212 raise errors.OpExecError("Can't execute '%s': not found (%s)" %
224 return RunResult(exitcode, signal_, out, err, strcmd)
227 def StartDaemon(cmd, env=None, cwd="/", output=None, output_fd=None,
229 """Start a daemon process after forking twice.
231 @type cmd: string or list
232 @param cmd: Command to run
234 @param env: Additional environment variables
236 @param cwd: Working directory for the program
238 @param output: Path to file in which to save the output
240 @param output_fd: File descriptor for output
241 @type pidfile: string
242 @param pidfile: Process ID file
244 @return: Daemon process ID
245 @raise errors.ProgrammerError: if we call this when forks are disabled
249 raise errors.ProgrammerError("utils.StartDaemon() called with fork()"
252 if output and not (bool(output) ^ (output_fd is not None)):
253 raise errors.ProgrammerError("Only one of 'output' and 'output_fd' can be"
256 if isinstance(cmd, basestring):
257 cmd = ["/bin/sh", "-c", cmd]
259 strcmd = ShellQuoteArgs(cmd)
262 logging.debug("StartDaemon %s, output file '%s'", strcmd, output)
264 logging.debug("StartDaemon %s", strcmd)
266 cmd_env = _BuildCmdEnvironment(env, False)
268 # Create pipe for sending PID back
269 (pidpipe_read, pidpipe_write) = os.pipe()
272 # Create pipe for sending error messages
273 (errpipe_read, errpipe_write) = os.pipe()
280 # Child process, won't return
281 _StartDaemonChild(errpipe_read, errpipe_write,
282 pidpipe_read, pidpipe_write,
284 output, output_fd, pidfile)
286 # Well, maybe child process failed
287 os._exit(1) # pylint: disable-msg=W0212
289 _CloseFDNoErr(errpipe_write)
291 # Wait for daemon to be started (or an error message to arrive) and read
292 # up to 100 KB as an error message
293 errormsg = RetryOnSignal(os.read, errpipe_read, 100 * 1024)
295 _CloseFDNoErr(errpipe_read)
297 _CloseFDNoErr(pidpipe_write)
299 # Read up to 128 bytes for PID
300 pidtext = RetryOnSignal(os.read, pidpipe_read, 128)
302 _CloseFDNoErr(pidpipe_read)
304 # Try to avoid zombies by waiting for child process
311 raise errors.OpExecError("Error when starting daemon process: %r" %
316 except (ValueError, TypeError), err:
317 raise errors.OpExecError("Error while trying to parse PID %r: %s" %
321 def _StartDaemonChild(errpipe_read, errpipe_write,
322 pidpipe_read, pidpipe_write,
324 output, fd_output, pidfile):
325 """Child process for starting daemon.
329 # Close parent's side
330 _CloseFDNoErr(errpipe_read)
331 _CloseFDNoErr(pidpipe_read)
333 # First child process
338 # And fork for the second time
341 # Exit first child process
342 os._exit(0) # pylint: disable-msg=W0212
344 # Make sure pipe is closed on execv* (and thereby notifies original process)
345 SetCloseOnExecFlag(errpipe_write, True)
347 # List of file descriptors to be left open
348 noclose_fds = [errpipe_write]
353 # TODO: Atomic replace with another locked file instead of writing into
355 fd_pidfile = os.open(pidfile, os.O_WRONLY | os.O_CREAT, 0600)
357 # Lock the PID file (and fail if not possible to do so). Any code
358 # wanting to send a signal to the daemon should try to lock the PID
359 # file before reading it. If acquiring the lock succeeds, the daemon is
360 # no longer running and the signal should not be sent.
363 os.write(fd_pidfile, "%d\n" % os.getpid())
364 except Exception, err:
365 raise Exception("Creating and locking PID file failed: %s" % err)
367 # Keeping the file open to hold the lock
368 noclose_fds.append(fd_pidfile)
370 SetCloseOnExecFlag(fd_pidfile, False)
375 fd_devnull = os.open(os.devnull, os.O_RDWR)
377 assert not output or (bool(output) ^ (fd_output is not None))
379 if fd_output is not None:
384 # TODO: Implement flag to set append=yes/no
385 fd_output = os.open(output, os.O_WRONLY | os.O_CREAT, 0600)
386 except EnvironmentError, err:
387 raise Exception("Opening output file failed: %s" % err)
389 fd_output = fd_devnull
391 # Redirect standard I/O
392 os.dup2(fd_devnull, 0)
393 os.dup2(fd_output, 1)
394 os.dup2(fd_output, 2)
396 # Send daemon PID to parent
397 RetryOnSignal(os.write, pidpipe_write, str(os.getpid()))
399 # Close all file descriptors except stdio and error message pipe
400 CloseFDs(noclose_fds=noclose_fds)
402 # Change working directory
406 os.execvp(args[0], args)
408 os.execvpe(args[0], args, env)
409 except: # pylint: disable-msg=W0702
411 # Report errors to original process
412 buf = str(sys.exc_info()[1])
414 RetryOnSignal(os.write, errpipe_write, buf)
415 except: # pylint: disable-msg=W0702
416 # Ignore errors in error handling
419 os._exit(1) # pylint: disable-msg=W0212
422 def _RunCmdPipe(cmd, env, via_shell, cwd):
423 """Run a command and return its output.
425 @type cmd: string or list
426 @param cmd: Command to run
428 @param env: The environment to use
429 @type via_shell: bool
430 @param via_shell: if we should run via the shell
432 @param cwd: the working directory for the program
434 @return: (out, err, status)
437 poller = select.poll()
438 child = subprocess.Popen(cmd, shell=via_shell,
439 stderr=subprocess.PIPE,
440 stdout=subprocess.PIPE,
441 stdin=subprocess.PIPE,
442 close_fds=True, env=env,
446 poller.register(child.stdout, select.POLLIN)
447 poller.register(child.stderr, select.POLLIN)
451 child.stdout.fileno(): (out, child.stdout),
452 child.stderr.fileno(): (err, child.stderr),
455 SetNonblockFlag(fd, True)
458 pollresult = RetryOnSignal(poller.poll)
460 for fd, event in pollresult:
461 if event & select.POLLIN or event & select.POLLPRI:
462 data = fdmap[fd][1].read()
463 # no data from read signifies EOF (the same as POLLHUP)
465 poller.unregister(fd)
468 fdmap[fd][0].write(data)
469 if (event & select.POLLNVAL or event & select.POLLHUP or
470 event & select.POLLERR):
471 poller.unregister(fd)
477 status = child.wait()
478 return out, err, status
481 def _RunCmdFile(cmd, env, via_shell, output, cwd):
482 """Run a command and save its output to a file.
484 @type cmd: string or list
485 @param cmd: Command to run
487 @param env: The environment to use
488 @type via_shell: bool
489 @param via_shell: if we should run via the shell
491 @param output: the filename in which to save the output
493 @param cwd: the working directory for the program
495 @return: the exit status
498 fh = open(output, "a")
500 child = subprocess.Popen(cmd, shell=via_shell,
501 stderr=subprocess.STDOUT,
503 stdin=subprocess.PIPE,
504 close_fds=True, env=env,
508 status = child.wait()
514 def SetCloseOnExecFlag(fd, enable):
515 """Sets or unsets the close-on-exec flag on a file descriptor.
518 @param fd: File descriptor
520 @param enable: Whether to set or unset it.
523 flags = fcntl.fcntl(fd, fcntl.F_GETFD)
526 flags |= fcntl.FD_CLOEXEC
528 flags &= ~fcntl.FD_CLOEXEC
530 fcntl.fcntl(fd, fcntl.F_SETFD, flags)
533 def SetNonblockFlag(fd, enable):
534 """Sets or unsets the O_NONBLOCK flag on on a file descriptor.
537 @param fd: File descriptor
539 @param enable: Whether to set or unset it
542 flags = fcntl.fcntl(fd, fcntl.F_GETFL)
545 flags |= os.O_NONBLOCK
547 flags &= ~os.O_NONBLOCK
549 fcntl.fcntl(fd, fcntl.F_SETFL, flags)
552 def RetryOnSignal(fn, *args, **kwargs):
553 """Calls a function again if it failed due to EINTR.
558 return fn(*args, **kwargs)
559 except EnvironmentError, err:
560 if err.errno != errno.EINTR:
562 except (socket.error, select.error), err:
563 # In python 2.6 and above select.error is an IOError, so it's handled
564 # above, in 2.5 and below it's not, and it's handled here.
565 if not (err.args and err.args[0] == errno.EINTR):
569 def RunParts(dir_name, env=None, reset_env=False):
570 """Run Scripts or programs in a directory
572 @type dir_name: string
573 @param dir_name: absolute path to a directory
575 @param env: The environment to use
576 @type reset_env: boolean
577 @param reset_env: whether to reset or keep the default os environment
578 @rtype: list of tuples
579 @return: list of (name, (one of RUNDIR_STATUS), RunResult)
585 dir_contents = ListVisibleFiles(dir_name)
587 logging.warning("RunParts: skipping %s (cannot list: %s)", dir_name, err)
590 for relname in sorted(dir_contents):
591 fname = PathJoin(dir_name, relname)
592 if not (os.path.isfile(fname) and os.access(fname, os.X_OK) and
593 constants.EXT_PLUGIN_MASK.match(relname) is not None):
594 rr.append((relname, constants.RUNPARTS_SKIP, None))
597 result = RunCmd([fname], env=env, reset_env=reset_env)
598 except Exception, err: # pylint: disable-msg=W0703
599 rr.append((relname, constants.RUNPARTS_ERR, str(err)))
601 rr.append((relname, constants.RUNPARTS_RUN, result))
606 def RemoveFile(filename):
607 """Remove a file ignoring some errors.
609 Remove a file, ignoring non-existing ones or directories. Other
613 @param filename: the file to be removed
619 if err.errno not in (errno.ENOENT, errno.EISDIR):
623 def RemoveDir(dirname):
624 """Remove an empty directory.
626 Remove a directory, ignoring non-existing ones.
627 Other errors are passed. This includes the case,
628 where the directory is not empty, so it can't be removed.
631 @param dirname: the empty directory to be removed
637 if err.errno != errno.ENOENT:
641 def RenameFile(old, new, mkdir=False, mkdir_mode=0750):
645 @param old: Original path
649 @param mkdir: Whether to create target directory if it doesn't exist
650 @type mkdir_mode: int
651 @param mkdir_mode: Mode for newly created directories
655 return os.rename(old, new)
657 # In at least one use case of this function, the job queue, directory
658 # creation is very rare. Checking for the directory before renaming is not
660 if mkdir and err.errno == errno.ENOENT:
661 # Create directory and try again
662 Makedirs(os.path.dirname(new), mode=mkdir_mode)
664 return os.rename(old, new)
669 def Makedirs(path, mode=0750):
670 """Super-mkdir; create a leaf directory and all intermediate ones.
672 This is a wrapper around C{os.makedirs} adding error handling not implemented
677 os.makedirs(path, mode)
679 # Ignore EEXIST. This is only handled in os.makedirs as included in
680 # Python 2.5 and above.
681 if err.errno != errno.EEXIST or not os.path.exists(path):
685 def ResetTempfileModule():
686 """Resets the random name generator of the tempfile module.
688 This function should be called after C{os.fork} in the child process to
689 ensure it creates a newly seeded random generator. Otherwise it would
690 generate the same random parts as the parent process. If several processes
691 race for the creation of a temporary file, this could lead to one not getting
695 # pylint: disable-msg=W0212
696 if hasattr(tempfile, "_once_lock") and hasattr(tempfile, "_name_sequence"):
697 tempfile._once_lock.acquire()
699 # Reset random name generator
700 tempfile._name_sequence = None
702 tempfile._once_lock.release()
704 logging.critical("The tempfile module misses at least one of the"
705 " '_once_lock' and '_name_sequence' attributes")
708 def _FingerprintFile(filename):
709 """Compute the fingerprint of a file.
711 If the file does not exist, a None will be returned
715 @param filename: the filename to checksum
717 @return: the hex digest of the sha checksum of the contents
721 if not (os.path.exists(filename) and os.path.isfile(filename)):
726 fp = compat.sha1_hash()
734 return fp.hexdigest()
737 def FingerprintFiles(files):
738 """Compute fingerprints for a list of files.
741 @param files: the list of filename to fingerprint
743 @return: a dictionary filename: fingerprint, holding only
749 for filename in files:
750 cksum = _FingerprintFile(filename)
752 ret[filename] = cksum
757 def ForceDictType(target, key_types, allowed_values=None):
758 """Force the values of a dict to have certain types.
761 @param target: the dict to update
762 @type key_types: dict
763 @param key_types: dict mapping target dict keys to types
764 in constants.ENFORCEABLE_TYPES
765 @type allowed_values: list
766 @keyword allowed_values: list of specially allowed values
769 if allowed_values is None:
772 if not isinstance(target, dict):
773 msg = "Expected dictionary, got '%s'" % target
774 raise errors.TypeEnforcementError(msg)
777 if key not in key_types:
778 msg = "Unknown key '%s'" % key
779 raise errors.TypeEnforcementError(msg)
781 if target[key] in allowed_values:
784 ktype = key_types[key]
785 if ktype not in constants.ENFORCEABLE_TYPES:
786 msg = "'%s' has non-enforceable type %s" % (key, ktype)
787 raise errors.ProgrammerError(msg)
789 if ktype in (constants.VTYPE_STRING, constants.VTYPE_MAYBE_STRING):
790 if target[key] is None and ktype == constants.VTYPE_MAYBE_STRING:
792 elif not isinstance(target[key], basestring):
793 if isinstance(target[key], bool) and not target[key]:
796 msg = "'%s' (value %s) is not a valid string" % (key, target[key])
797 raise errors.TypeEnforcementError(msg)
798 elif ktype == constants.VTYPE_BOOL:
799 if isinstance(target[key], basestring) and target[key]:
800 if target[key].lower() == constants.VALUE_FALSE:
802 elif target[key].lower() == constants.VALUE_TRUE:
805 msg = "'%s' (value %s) is not a valid boolean" % (key, target[key])
806 raise errors.TypeEnforcementError(msg)
811 elif ktype == constants.VTYPE_SIZE:
813 target[key] = ParseUnit(target[key])
814 except errors.UnitParseError, err:
815 msg = "'%s' (value %s) is not a valid size. error: %s" % \
816 (key, target[key], err)
817 raise errors.TypeEnforcementError(msg)
818 elif ktype == constants.VTYPE_INT:
820 target[key] = int(target[key])
821 except (ValueError, TypeError):
822 msg = "'%s' (value %s) is not a valid integer" % (key, target[key])
823 raise errors.TypeEnforcementError(msg)
826 def _GetProcStatusPath(pid):
827 """Returns the path for a PID's proc status file.
830 @param pid: Process ID
834 return "/proc/%d/status" % pid
837 def IsProcessAlive(pid):
838 """Check if a given pid exists on the system.
840 @note: zombie status is not handled, so zombie processes
841 will be returned as alive
843 @param pid: the process ID to check
845 @return: True if the process exists
852 except EnvironmentError, err:
853 if err.errno in (errno.ENOENT, errno.ENOTDIR):
855 elif err.errno == errno.EINVAL:
856 raise RetryAgain(err)
859 assert isinstance(pid, int), "pid must be an integer"
863 # /proc in a multiprocessor environment can have strange behaviors.
864 # Retry the os.stat a few times until we get a good result.
866 return Retry(_TryStat, (0.01, 1.5, 0.1), 0.5,
867 args=[_GetProcStatusPath(pid)])
868 except RetryTimeout, err:
872 def _ParseSigsetT(sigset):
873 """Parse a rendered sigset_t value.
875 This is the opposite of the Linux kernel's fs/proc/array.c:render_sigset_t
879 @param sigset: Rendered signal set from /proc/$pid/status
881 @return: Set of all enabled signal numbers
887 for ch in reversed(sigset):
890 # The following could be done in a loop, but it's easier to read and
891 # understand in the unrolled form
893 result.add(signum + 1)
895 result.add(signum + 2)
897 result.add(signum + 3)
899 result.add(signum + 4)
906 def _GetProcStatusField(pstatus, field):
907 """Retrieves a field from the contents of a proc status file.
909 @type pstatus: string
910 @param pstatus: Contents of /proc/$pid/status
912 @param field: Name of field whose value should be returned
916 for line in pstatus.splitlines():
917 parts = line.split(":", 1)
919 if len(parts) < 2 or parts[0] != field:
922 return parts[1].strip()
927 def IsProcessHandlingSignal(pid, signum, status_path=None):
928 """Checks whether a process is handling a signal.
931 @param pid: Process ID
933 @param signum: Signal number
937 if status_path is None:
938 status_path = _GetProcStatusPath(pid)
941 proc_status = ReadFile(status_path)
942 except EnvironmentError, err:
943 # In at least one case, reading /proc/$pid/status failed with ESRCH.
944 if err.errno in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL, errno.ESRCH):
948 sigcgt = _GetProcStatusField(proc_status, "SigCgt")
950 raise RuntimeError("%s is missing 'SigCgt' field" % status_path)
952 # Now check whether signal is handled
953 return signum in _ParseSigsetT(sigcgt)
956 def ReadPidFile(pidfile):
957 """Read a pid from a file.
959 @type pidfile: string
960 @param pidfile: path to the file containing the pid
962 @return: The process id, if the file exists and contains a valid PID,
967 raw_data = ReadOneLineFile(pidfile)
968 except EnvironmentError, err:
969 if err.errno != errno.ENOENT:
970 logging.exception("Can't read pid file")
975 except (TypeError, ValueError), err:
976 logging.info("Can't parse pid file contents", exc_info=True)
982 def ReadLockedPidFile(path):
983 """Reads a locked PID file.
985 This can be used together with L{StartDaemon}.
988 @param path: Path to PID file
989 @return: PID as integer or, if file was unlocked or couldn't be opened, None
993 fd = os.open(path, os.O_RDONLY)
994 except EnvironmentError, err:
995 if err.errno == errno.ENOENT:
996 # PID file doesn't exist
1002 # Try to acquire lock
1004 except errors.LockError:
1005 # Couldn't lock, daemon is running
1006 return int(os.read(fd, 100))
1013 def MatchNameComponent(key, name_list, case_sensitive=True):
1014 """Try to match a name against a list.
1016 This function will try to match a name like test1 against a list
1017 like C{['test1.example.com', 'test2.example.com', ...]}. Against
1018 this list, I{'test1'} as well as I{'test1.example'} will match, but
1019 not I{'test1.ex'}. A multiple match will be considered as no match
1020 at all (e.g. I{'test1'} against C{['test1.example.com',
1021 'test1.example.org']}), except when the key fully matches an entry
1022 (e.g. I{'test1'} against C{['test1', 'test1.example.com']}).
1025 @param key: the name to be searched
1026 @type name_list: list
1027 @param name_list: the list of strings against which to search the key
1028 @type case_sensitive: boolean
1029 @param case_sensitive: whether to provide a case-sensitive match
1032 @return: None if there is no match I{or} if there are multiple matches,
1033 otherwise the element from the list which matches
1036 if key in name_list:
1040 if not case_sensitive:
1041 re_flags |= re.IGNORECASE
1043 mo = re.compile("^%s(\..*)?$" % re.escape(key), re_flags)
1046 for name in name_list:
1047 if mo.match(name) is not None:
1048 names_filtered.append(name)
1049 if not case_sensitive and key == name.upper():
1050 string_matches.append(name)
1052 if len(string_matches) == 1:
1053 return string_matches[0]
1054 if len(names_filtered) == 1:
1055 return names_filtered[0]
1059 def ValidateServiceName(name):
1060 """Validate the given service name.
1062 @type name: number or string
1063 @param name: Service name or port specification
1068 except (ValueError, TypeError):
1069 # Non-numeric service name
1070 valid = _VALID_SERVICE_NAME_RE.match(name)
1072 # Numeric port (protocols other than TCP or UDP might need adjustments
1074 valid = (numport >= 0 and numport < (1 << 16))
1077 raise errors.OpPrereqError("Invalid service name '%s'" % name,
1083 def ListVolumeGroups():
1084 """List volume groups and their size
1088 Dictionary with keys volume name and values
1089 the size of the volume
1092 command = "vgs --noheadings --units m --nosuffix -o name,size"
1093 result = RunCmd(command)
1098 for line in result.stdout.splitlines():
1100 name, size = line.split()
1101 size = int(float(size))
1102 except (IndexError, ValueError), err:
1103 logging.error("Invalid output from vgs (%s): %s", err, line)
1111 def BridgeExists(bridge):
1112 """Check whether the given bridge exists in the system
1115 @param bridge: the bridge name to check
1117 @return: True if it does
1120 return os.path.isdir("/sys/class/net/%s/bridge" % bridge)
1123 def NiceSort(name_list):
1124 """Sort a list of strings based on digit and non-digit groupings.
1126 Given a list of names C{['a1', 'a10', 'a11', 'a2']} this function
1127 will sort the list in the logical order C{['a1', 'a2', 'a10',
1130 The sort algorithm breaks each name in groups of either only-digits
1131 or no-digits. Only the first eight such groups are considered, and
1132 after that we just use what's left of the string.
1134 @type name_list: list
1135 @param name_list: the names to be sorted
1137 @return: a copy of the name list sorted with our algorithm
1140 _SORTER_BASE = "(\D+|\d+)"
1141 _SORTER_FULL = "^%s%s?%s?%s?%s?%s?%s?%s?.*$" % (_SORTER_BASE, _SORTER_BASE,
1142 _SORTER_BASE, _SORTER_BASE,
1143 _SORTER_BASE, _SORTER_BASE,
1144 _SORTER_BASE, _SORTER_BASE)
1145 _SORTER_RE = re.compile(_SORTER_FULL)
1146 _SORTER_NODIGIT = re.compile("^\D*$")
1148 """Attempts to convert a variable to integer."""
1149 if val is None or _SORTER_NODIGIT.match(val):
1154 to_sort = [([_TryInt(grp) for grp in _SORTER_RE.match(name).groups()], name)
1155 for name in name_list]
1157 return [tup[1] for tup in to_sort]
1160 def TryConvert(fn, val):
1161 """Try to convert a value ignoring errors.
1163 This function tries to apply function I{fn} to I{val}. If no
1164 C{ValueError} or C{TypeError} exceptions are raised, it will return
1165 the result, else it will return the original value. Any other
1166 exceptions are propagated to the caller.
1169 @param fn: function to apply to the value
1170 @param val: the value to be converted
1171 @return: The converted value if the conversion was successful,
1172 otherwise the original value.
1177 except (ValueError, TypeError):
1182 def IsValidShellParam(word):
1183 """Verifies is the given word is safe from the shell's p.o.v.
1185 This means that we can pass this to a command via the shell and be
1186 sure that it doesn't alter the command line and is passed as such to
1189 Note that we are overly restrictive here, in order to be on the safe
1193 @param word: the word to check
1195 @return: True if the word is 'safe'
1198 return bool(re.match("^[-a-zA-Z0-9._+/:%@]+$", word))
1201 def BuildShellCmd(template, *args):
1202 """Build a safe shell command line from the given arguments.
1204 This function will check all arguments in the args list so that they
1205 are valid shell parameters (i.e. they don't contain shell
1206 metacharacters). If everything is ok, it will return the result of
1210 @param template: the string holding the template for the
1213 @return: the expanded command line
1217 if not IsValidShellParam(word):
1218 raise errors.ProgrammerError("Shell argument '%s' contains"
1219 " invalid characters" % word)
1220 return template % args
1223 def FormatUnit(value, units):
1224 """Formats an incoming number of MiB with the appropriate unit.
1227 @param value: integer representing the value in MiB (1048576)
1229 @param units: the type of formatting we should do:
1230 - 'h' for automatic scaling
1235 @return: the formatted value (with suffix)
1238 if units not in ('m', 'g', 't', 'h'):
1239 raise errors.ProgrammerError("Invalid unit specified '%s'" % str(units))
1243 if units == 'm' or (units == 'h' and value < 1024):
1246 return "%d%s" % (round(value, 0), suffix)
1248 elif units == 'g' or (units == 'h' and value < (1024 * 1024)):
1251 return "%0.1f%s" % (round(float(value) / 1024, 1), suffix)
1256 return "%0.1f%s" % (round(float(value) / 1024 / 1024, 1), suffix)
1259 def ParseUnit(input_string):
1260 """Tries to extract number and scale from the given string.
1262 Input must be in the format C{NUMBER+ [DOT NUMBER+] SPACE*
1263 [UNIT]}. If no unit is specified, it defaults to MiB. Return value
1264 is always an int in MiB.
1267 m = re.match('^([.\d]+)\s*([a-zA-Z]+)?$', str(input_string))
1269 raise errors.UnitParseError("Invalid format")
1271 value = float(m.groups()[0])
1273 unit = m.groups()[1]
1275 lcunit = unit.lower()
1279 if lcunit in ('m', 'mb', 'mib'):
1280 # Value already in MiB
1283 elif lcunit in ('g', 'gb', 'gib'):
1286 elif lcunit in ('t', 'tb', 'tib'):
1287 value *= 1024 * 1024
1290 raise errors.UnitParseError("Unknown unit: %s" % unit)
1292 # Make sure we round up
1293 if int(value) < value:
1296 # Round up to the next multiple of 4
1299 value += 4 - value % 4
1304 def ParseCpuMask(cpu_mask):
1305 """Parse a CPU mask definition and return the list of CPU IDs.
1307 CPU mask format: comma-separated list of CPU IDs
1308 or dash-separated ID ranges
1309 Example: "0-2,5" -> "0,1,2,5"
1312 @param cpu_mask: CPU mask definition
1314 @return: list of CPU IDs
1320 for range_def in cpu_mask.split(","):
1321 boundaries = range_def.split("-")
1322 n_elements = len(boundaries)
1324 raise errors.ParseError("Invalid CPU ID range definition"
1325 " (only one hyphen allowed): %s" % range_def)
1327 lower = int(boundaries[0])
1328 except (ValueError, TypeError), err:
1329 raise errors.ParseError("Invalid CPU ID value for lower boundary of"
1330 " CPU ID range: %s" % str(err))
1332 higher = int(boundaries[-1])
1333 except (ValueError, TypeError), err:
1334 raise errors.ParseError("Invalid CPU ID value for higher boundary of"
1335 " CPU ID range: %s" % str(err))
1337 raise errors.ParseError("Invalid CPU ID range definition"
1338 " (%d > %d): %s" % (lower, higher, range_def))
1339 cpu_list.extend(range(lower, higher + 1))
1343 def AddAuthorizedKey(file_obj, key):
1344 """Adds an SSH public key to an authorized_keys file.
1346 @type file_obj: str or file handle
1347 @param file_obj: path to authorized_keys file
1349 @param key: string containing key
1352 key_fields = key.split()
1354 if isinstance(file_obj, basestring):
1355 f = open(file_obj, 'a+')
1362 # Ignore whitespace changes
1363 if line.split() == key_fields:
1365 nl = line.endswith('\n')
1369 f.write(key.rstrip('\r\n'))
1376 def RemoveAuthorizedKey(file_name, key):
1377 """Removes an SSH public key from an authorized_keys file.
1379 @type file_name: str
1380 @param file_name: path to authorized_keys file
1382 @param key: string containing key
1385 key_fields = key.split()
1387 fd, tmpname = tempfile.mkstemp(dir=os.path.dirname(file_name))
1389 out = os.fdopen(fd, 'w')
1391 f = open(file_name, 'r')
1394 # Ignore whitespace changes while comparing lines
1395 if line.split() != key_fields:
1399 os.rename(tmpname, file_name)
1409 def SetEtcHostsEntry(file_name, ip, hostname, aliases):
1410 """Sets the name of an IP address and hostname in /etc/hosts.
1412 @type file_name: str
1413 @param file_name: path to the file to modify (usually C{/etc/hosts})
1415 @param ip: the IP address
1417 @param hostname: the hostname to be added
1419 @param aliases: the list of aliases to add for the hostname
1422 # FIXME: use WriteFile + fn rather than duplicating its efforts
1423 # Ensure aliases are unique
1424 aliases = UniqueSequence([hostname] + aliases)[1:]
1426 fd, tmpname = tempfile.mkstemp(dir=os.path.dirname(file_name))
1428 out = os.fdopen(fd, 'w')
1430 f = open(file_name, 'r')
1433 fields = line.split()
1434 if fields and not fields[0].startswith('#') and ip == fields[0]:
1438 out.write("%s\t%s" % (ip, hostname))
1440 out.write(" %s" % ' '.join(aliases))
1445 os.chmod(tmpname, 0644)
1446 os.rename(tmpname, file_name)
1456 def AddHostToEtcHosts(hostname):
1457 """Wrapper around SetEtcHostsEntry.
1460 @param hostname: a hostname that will be resolved and added to
1461 L{constants.ETC_HOSTS}
1464 SetEtcHostsEntry(constants.ETC_HOSTS, netutils.Hostname.GetIP(hostname),
1465 hostname, [hostname.split(".")[0]])
1468 def RemoveEtcHostsEntry(file_name, hostname):
1469 """Removes a hostname from /etc/hosts.
1471 IP addresses without names are removed from the file.
1473 @type file_name: str
1474 @param file_name: path to the file to modify (usually C{/etc/hosts})
1476 @param hostname: the hostname to be removed
1479 # FIXME: use WriteFile + fn rather than duplicating its efforts
1480 fd, tmpname = tempfile.mkstemp(dir=os.path.dirname(file_name))
1482 out = os.fdopen(fd, 'w')
1484 f = open(file_name, 'r')
1487 fields = line.split()
1488 if len(fields) > 1 and not fields[0].startswith('#'):
1490 if hostname in names:
1491 while hostname in names:
1492 names.remove(hostname)
1494 out.write("%s %s\n" % (fields[0], ' '.join(names)))
1501 os.chmod(tmpname, 0644)
1502 os.rename(tmpname, file_name)
1512 def RemoveHostFromEtcHosts(hostname):
1513 """Wrapper around RemoveEtcHostsEntry.
1516 @param hostname: hostname that will be resolved and its
1517 full and shot name will be removed from
1518 L{constants.ETC_HOSTS}
1521 RemoveEtcHostsEntry(constants.ETC_HOSTS, hostname)
1522 RemoveEtcHostsEntry(constants.ETC_HOSTS, hostname.split(".")[0])
1525 def TimestampForFilename():
1526 """Returns the current time formatted for filenames.
1528 The format doesn't contain colons as some shells and applications them as
1532 return time.strftime("%Y-%m-%d_%H_%M_%S")
1535 def CreateBackup(file_name):
1536 """Creates a backup of a file.
1538 @type file_name: str
1539 @param file_name: file to be backed up
1541 @return: the path to the newly created backup
1542 @raise errors.ProgrammerError: for invalid file names
1545 if not os.path.isfile(file_name):
1546 raise errors.ProgrammerError("Can't make a backup of a non-file '%s'" %
1549 prefix = ("%s.backup-%s." %
1550 (os.path.basename(file_name), TimestampForFilename()))
1551 dir_name = os.path.dirname(file_name)
1553 fsrc = open(file_name, 'rb')
1555 (fd, backup_name) = tempfile.mkstemp(prefix=prefix, dir=dir_name)
1556 fdst = os.fdopen(fd, 'wb')
1558 logging.debug("Backing up %s at %s", file_name, backup_name)
1559 shutil.copyfileobj(fsrc, fdst)
1568 def ShellQuote(value):
1569 """Quotes shell argument according to POSIX.
1572 @param value: the argument to be quoted
1574 @return: the quoted value
1577 if _re_shell_unquoted.match(value):
1580 return "'%s'" % value.replace("'", "'\\''")
1583 def ShellQuoteArgs(args):
1584 """Quotes a list of shell arguments.
1587 @param args: list of arguments to be quoted
1589 @return: the quoted arguments concatenated with spaces
1592 return ' '.join([ShellQuote(i) for i in args])
1596 """Helper class to write scripts with indentation.
1601 def __init__(self, fh):
1602 """Initializes this class.
1608 def IncIndent(self):
1609 """Increase indentation level by 1.
1614 def DecIndent(self):
1615 """Decrease indentation level by 1.
1618 assert self._indent > 0
1621 def Write(self, txt, *args):
1622 """Write line to output file.
1625 assert self._indent >= 0
1627 self._fh.write(self._indent * self.INDENT_STR)
1630 self._fh.write(txt % args)
1634 self._fh.write("\n")
1637 def ListVisibleFiles(path):
1638 """Returns a list of visible files in a directory.
1641 @param path: the directory to enumerate
1643 @return: the list of all files not starting with a dot
1644 @raise ProgrammerError: if L{path} is not an absolue and normalized path
1647 if not IsNormAbsPath(path):
1648 raise errors.ProgrammerError("Path passed to ListVisibleFiles is not"
1649 " absolute/normalized: '%s'" % path)
1650 files = [i for i in os.listdir(path) if not i.startswith(".")]
1654 def GetHomeDir(user, default=None):
1655 """Try to get the homedir of the given user.
1657 The user can be passed either as a string (denoting the name) or as
1658 an integer (denoting the user id). If the user is not found, the
1659 'default' argument is returned, which defaults to None.
1663 if isinstance(user, basestring):
1664 result = pwd.getpwnam(user)
1665 elif isinstance(user, (int, long)):
1666 result = pwd.getpwuid(user)
1668 raise errors.ProgrammerError("Invalid type passed to GetHomeDir (%s)" %
1672 return result.pw_dir
1676 """Returns a random UUID.
1678 @note: This is a Linux-specific method as it uses the /proc
1683 return ReadFile(_RANDOM_UUID_FILE, size=128).rstrip("\n")
1686 def GenerateSecret(numbytes=20):
1687 """Generates a random secret.
1689 This will generate a pseudo-random secret returning an hex string
1690 (so that it can be used where an ASCII string is needed).
1692 @param numbytes: the number of bytes which will be represented by the returned
1693 string (defaulting to 20, the length of a SHA1 hash)
1695 @return: an hex representation of the pseudo-random sequence
1698 return os.urandom(numbytes).encode('hex')
1701 def EnsureDirs(dirs):
1702 """Make required directories, if they don't exist.
1704 @param dirs: list of tuples (dir_name, dir_mode)
1705 @type dirs: list of (string, integer)
1708 for dir_name, dir_mode in dirs:
1710 os.mkdir(dir_name, dir_mode)
1711 except EnvironmentError, err:
1712 if err.errno != errno.EEXIST:
1713 raise errors.GenericError("Cannot create needed directory"
1714 " '%s': %s" % (dir_name, err))
1716 os.chmod(dir_name, dir_mode)
1717 except EnvironmentError, err:
1718 raise errors.GenericError("Cannot change directory permissions on"
1719 " '%s': %s" % (dir_name, err))
1720 if not os.path.isdir(dir_name):
1721 raise errors.GenericError("%s is not a directory" % dir_name)
1724 def ReadFile(file_name, size=-1):
1728 @param size: Read at most size bytes (if negative, entire file)
1730 @return: the (possibly partial) content of the file
1733 f = open(file_name, "r")
1740 def WriteFile(file_name, fn=None, data=None,
1741 mode=None, uid=-1, gid=-1,
1742 atime=None, mtime=None, close=True,
1743 dry_run=False, backup=False,
1744 prewrite=None, postwrite=None):
1745 """(Over)write a file atomically.
1747 The file_name and either fn (a function taking one argument, the
1748 file descriptor, and which should write the data to it) or data (the
1749 contents of the file) must be passed. The other arguments are
1750 optional and allow setting the file mode, owner and group, and the
1751 mtime/atime of the file.
1753 If the function doesn't raise an exception, it has succeeded and the
1754 target file has the new contents. If the function has raised an
1755 exception, an existing target file should be unmodified and the
1756 temporary file should be removed.
1758 @type file_name: str
1759 @param file_name: the target filename
1761 @param fn: content writing function, called with
1762 file descriptor as parameter
1764 @param data: contents of the file
1766 @param mode: file mode
1768 @param uid: the owner of the file
1770 @param gid: the group of the file
1772 @param atime: a custom access time to be set on the file
1774 @param mtime: a custom modification time to be set on the file
1775 @type close: boolean
1776 @param close: whether to close file after writing it
1777 @type prewrite: callable
1778 @param prewrite: function to be called before writing content
1779 @type postwrite: callable
1780 @param postwrite: function to be called after writing content
1783 @return: None if the 'close' parameter evaluates to True,
1784 otherwise the file descriptor
1786 @raise errors.ProgrammerError: if any of the arguments are not valid
1789 if not os.path.isabs(file_name):
1790 raise errors.ProgrammerError("Path passed to WriteFile is not"
1791 " absolute: '%s'" % file_name)
1793 if [fn, data].count(None) != 1:
1794 raise errors.ProgrammerError("fn or data required")
1796 if [atime, mtime].count(None) == 1:
1797 raise errors.ProgrammerError("Both atime and mtime must be either"
1800 if backup and not dry_run and os.path.isfile(file_name):
1801 CreateBackup(file_name)
1803 dir_name, base_name = os.path.split(file_name)
1804 fd, new_name = tempfile.mkstemp('.new', base_name, dir_name)
1806 # here we need to make sure we remove the temp file, if any error
1807 # leaves it in place
1809 if uid != -1 or gid != -1:
1810 os.chown(new_name, uid, gid)
1812 os.chmod(new_name, mode)
1813 if callable(prewrite):
1815 if data is not None:
1819 if callable(postwrite):
1822 if atime is not None and mtime is not None:
1823 os.utime(new_name, (atime, mtime))
1825 os.rename(new_name, file_name)
1834 RemoveFile(new_name)
1839 def ReadOneLineFile(file_name, strict=False):
1840 """Return the first non-empty line from a file.
1842 @type strict: boolean
1843 @param strict: if True, abort if the file has more than one
1847 file_lines = ReadFile(file_name).splitlines()
1848 full_lines = filter(bool, file_lines)
1849 if not file_lines or not full_lines:
1850 raise errors.GenericError("No data in one-liner file %s" % file_name)
1851 elif strict and len(full_lines) > 1:
1852 raise errors.GenericError("Too many lines in one-liner file %s" %
1854 return full_lines[0]
1857 def FirstFree(seq, base=0):
1858 """Returns the first non-existing integer from seq.
1860 The seq argument should be a sorted list of positive integers. The
1861 first time the index of an element is smaller than the element
1862 value, the index will be returned.
1864 The base argument is used to start at a different offset,
1865 i.e. C{[3, 4, 6]} with I{offset=3} will return 5.
1867 Example: C{[0, 1, 3]} will return I{2}.
1870 @param seq: the sequence to be analyzed.
1872 @param base: use this value as the base index of the sequence
1874 @return: the first non-used index in the sequence
1877 for idx, elem in enumerate(seq):
1878 assert elem >= base, "Passed element is higher than base offset"
1879 if elem > idx + base:
1885 def SingleWaitForFdCondition(fdobj, event, timeout):
1886 """Waits for a condition to occur on the socket.
1888 Immediately returns at the first interruption.
1890 @type fdobj: integer or object supporting a fileno() method
1891 @param fdobj: entity to wait for events on
1892 @type event: integer
1893 @param event: ORed condition (see select module)
1894 @type timeout: float or None
1895 @param timeout: Timeout in seconds
1897 @return: None for timeout, otherwise occured conditions
1900 check = (event | select.POLLPRI |
1901 select.POLLNVAL | select.POLLHUP | select.POLLERR)
1903 if timeout is not None:
1904 # Poller object expects milliseconds
1907 poller = select.poll()
1908 poller.register(fdobj, event)
1910 # TODO: If the main thread receives a signal and we have no timeout, we
1911 # could wait forever. This should check a global "quit" flag or something
1913 io_events = poller.poll(timeout)
1914 except select.error, err:
1915 if err[0] != errno.EINTR:
1918 if io_events and io_events[0][1] & check:
1919 return io_events[0][1]
1924 class FdConditionWaiterHelper(object):
1925 """Retry helper for WaitForFdCondition.
1927 This class contains the retried and wait functions that make sure
1928 WaitForFdCondition can continue waiting until the timeout is actually
1933 def __init__(self, timeout):
1934 self.timeout = timeout
1936 def Poll(self, fdobj, event):
1937 result = SingleWaitForFdCondition(fdobj, event, self.timeout)
1943 def UpdateTimeout(self, timeout):
1944 self.timeout = timeout
1947 def WaitForFdCondition(fdobj, event, timeout):
1948 """Waits for a condition to occur on the socket.
1950 Retries until the timeout is expired, even if interrupted.
1952 @type fdobj: integer or object supporting a fileno() method
1953 @param fdobj: entity to wait for events on
1954 @type event: integer
1955 @param event: ORed condition (see select module)
1956 @type timeout: float or None
1957 @param timeout: Timeout in seconds
1959 @return: None for timeout, otherwise occured conditions
1962 if timeout is not None:
1963 retrywaiter = FdConditionWaiterHelper(timeout)
1965 result = Retry(retrywaiter.Poll, RETRY_REMAINING_TIME, timeout,
1966 args=(fdobj, event), wait_fn=retrywaiter.UpdateTimeout)
1967 except RetryTimeout:
1971 while result is None:
1972 result = SingleWaitForFdCondition(fdobj, event, timeout)
1976 def UniqueSequence(seq):
1977 """Returns a list with unique elements.
1979 Element order is preserved.
1982 @param seq: the sequence with the source elements
1984 @return: list of unique elements from seq
1988 return [i for i in seq if i not in seen and not seen.add(i)]
1991 def NormalizeAndValidateMac(mac):
1992 """Normalizes and check if a MAC address is valid.
1994 Checks whether the supplied MAC address is formally correct, only
1995 accepts colon separated format. Normalize it to all lower.
1998 @param mac: the MAC to be validated
2000 @return: returns the normalized and validated MAC.
2002 @raise errors.OpPrereqError: If the MAC isn't valid
2005 mac_check = re.compile("^([0-9a-f]{2}(:|$)){6}$", re.I)
2006 if not mac_check.match(mac):
2007 raise errors.OpPrereqError("Invalid MAC address specified: %s" %
2008 mac, errors.ECODE_INVAL)
2013 def TestDelay(duration):
2014 """Sleep for a fixed amount of time.
2016 @type duration: float
2017 @param duration: the sleep duration
2019 @return: False for negative value, True otherwise
2023 return False, "Invalid sleep duration"
2024 time.sleep(duration)
2028 def _CloseFDNoErr(fd, retries=5):
2029 """Close a file descriptor ignoring errors.
2032 @param fd: the file descriptor
2034 @param retries: how many retries to make, in case we get any
2035 other error than EBADF
2040 except OSError, err:
2041 if err.errno != errno.EBADF:
2043 _CloseFDNoErr(fd, retries - 1)
2044 # else either it's closed already or we're out of retries, so we
2045 # ignore this and go on
2048 def CloseFDs(noclose_fds=None):
2049 """Close file descriptors.
2051 This closes all file descriptors above 2 (i.e. except
2054 @type noclose_fds: list or None
2055 @param noclose_fds: if given, it denotes a list of file descriptor
2056 that should not be closed
2059 # Default maximum for the number of available file descriptors.
2060 if 'SC_OPEN_MAX' in os.sysconf_names:
2062 MAXFD = os.sysconf('SC_OPEN_MAX')
2069 maxfd = resource.getrlimit(resource.RLIMIT_NOFILE)[1]
2070 if (maxfd == resource.RLIM_INFINITY):
2073 # Iterate through and close all file descriptors (except the standard ones)
2074 for fd in range(3, maxfd):
2075 if noclose_fds and fd in noclose_fds:
2080 def Mlockall(_ctypes=ctypes):
2081 """Lock current process' virtual address space into RAM.
2083 This is equivalent to the C call mlockall(MCL_CURRENT|MCL_FUTURE),
2084 see mlock(2) for more details. This function requires ctypes module.
2086 @raises errors.NoCtypesError: if ctypes module is not found
2090 raise errors.NoCtypesError()
2092 libc = _ctypes.cdll.LoadLibrary("libc.so.6")
2094 logging.error("Cannot set memory lock, ctypes cannot load libc")
2097 # Some older version of the ctypes module don't have built-in functionality
2098 # to access the errno global variable, where function error codes are stored.
2099 # By declaring this variable as a pointer to an integer we can then access
2100 # its value correctly, should the mlockall call fail, in order to see what
2101 # the actual error code was.
2102 # pylint: disable-msg=W0212
2103 libc.__errno_location.restype = _ctypes.POINTER(_ctypes.c_int)
2105 if libc.mlockall(_MCL_CURRENT | _MCL_FUTURE):
2106 # pylint: disable-msg=W0212
2107 logging.error("Cannot set memory lock: %s",
2108 os.strerror(libc.__errno_location().contents.value))
2111 logging.debug("Memory lock set")
2114 def Daemonize(logfile, run_uid, run_gid):
2115 """Daemonize the current process.
2117 This detaches the current process from the controlling terminal and
2118 runs it in the background as a daemon.
2121 @param logfile: the logfile to which we should redirect stdout/stderr
2123 @param run_uid: Run the child under this uid
2125 @param run_gid: Run the child under this gid
2127 @return: the value zero
2130 # pylint: disable-msg=W0212
2131 # yes, we really want os._exit
2137 if (pid == 0): # The first child.
2139 # FIXME: When removing again and moving to start-stop-daemon privilege drop
2140 # make sure to check for config permission and bail out when invoked
2145 pid = os.fork() # Fork a second child.
2146 if (pid == 0): # The second child.
2150 # exit() or _exit()? See below.
2151 os._exit(0) # Exit parent (the first child) of the second child.
2153 os._exit(0) # Exit parent of the first child.
2157 i = os.open("/dev/null", os.O_RDONLY) # stdin
2158 assert i == 0, "Can't close/reopen stdin"
2159 i = os.open(logfile, os.O_WRONLY|os.O_CREAT|os.O_APPEND, 0600) # stdout
2160 assert i == 1, "Can't close/reopen stdout"
2161 # Duplicate standard output to standard error.
2166 def DaemonPidFileName(name):
2167 """Compute a ganeti pid file absolute path
2170 @param name: the daemon name
2172 @return: the full path to the pidfile corresponding to the given
2176 return PathJoin(constants.RUN_GANETI_DIR, "%s.pid" % name)
2179 def EnsureDaemon(name):
2180 """Check for and start daemon if not alive.
2183 result = RunCmd([constants.DAEMON_UTIL, "check-and-start", name])
2185 logging.error("Can't start daemon '%s', failure %s, output: %s",
2186 name, result.fail_reason, result.output)
2192 def StopDaemon(name):
2196 result = RunCmd([constants.DAEMON_UTIL, "stop", name])
2198 logging.error("Can't stop daemon '%s', failure %s, output: %s",
2199 name, result.fail_reason, result.output)
2205 def WritePidFile(name):
2206 """Write the current process pidfile.
2208 The file will be written to L{constants.RUN_GANETI_DIR}I{/name.pid}
2211 @param name: the daemon name to use
2212 @raise errors.GenericError: if the pid file already exists and
2213 points to a live process
2217 pidfilename = DaemonPidFileName(name)
2218 if IsProcessAlive(ReadPidFile(pidfilename)):
2219 raise errors.GenericError("%s contains a live process" % pidfilename)
2221 WriteFile(pidfilename, data="%d\n" % pid)
2224 def RemovePidFile(name):
2225 """Remove the current process pidfile.
2227 Any errors are ignored.
2230 @param name: the daemon name used to derive the pidfile name
2233 pidfilename = DaemonPidFileName(name)
2234 # TODO: we could check here that the file contains our pid
2236 RemoveFile(pidfilename)
2237 except: # pylint: disable-msg=W0702
2241 def KillProcess(pid, signal_=signal.SIGTERM, timeout=30,
2243 """Kill a process given by its pid.
2246 @param pid: The PID to terminate.
2248 @param signal_: The signal to send, by default SIGTERM
2250 @param timeout: The timeout after which, if the process is still alive,
2251 a SIGKILL will be sent. If not positive, no such checking
2253 @type waitpid: boolean
2254 @param waitpid: If true, we should waitpid on this process after
2255 sending signals, since it's our own child and otherwise it
2256 would remain as zombie
2259 def _helper(pid, signal_, wait):
2260 """Simple helper to encapsulate the kill/waitpid sequence"""
2261 if IgnoreProcessNotFound(os.kill, pid, signal_) and wait:
2263 os.waitpid(pid, os.WNOHANG)
2268 # kill with pid=0 == suicide
2269 raise errors.ProgrammerError("Invalid pid given '%s'" % pid)
2271 if not IsProcessAlive(pid):
2274 _helper(pid, signal_, waitpid)
2279 def _CheckProcess():
2280 if not IsProcessAlive(pid):
2284 (result_pid, _) = os.waitpid(pid, os.WNOHANG)
2294 # Wait up to $timeout seconds
2295 Retry(_CheckProcess, (0.01, 1.5, 0.1), timeout)
2296 except RetryTimeout:
2299 if IsProcessAlive(pid):
2300 # Kill process if it's still alive
2301 _helper(pid, signal.SIGKILL, waitpid)
2304 def FindFile(name, search_path, test=os.path.exists):
2305 """Look for a filesystem object in a given path.
2307 This is an abstract method to search for filesystem object (files,
2308 dirs) under a given search path.
2311 @param name: the name to look for
2312 @type search_path: str
2313 @param search_path: location to start at
2314 @type test: callable
2315 @param test: a function taking one argument that should return True
2316 if the a given object is valid; the default value is
2317 os.path.exists, causing only existing files to be returned
2319 @return: full path to the object if found, None otherwise
2322 # validate the filename mask
2323 if constants.EXT_PLUGIN_MASK.match(name) is None:
2324 logging.critical("Invalid value passed for external script name: '%s'",
2328 for dir_name in search_path:
2329 # FIXME: investigate switch to PathJoin
2330 item_name = os.path.sep.join([dir_name, name])
2331 # check the user test and that we're indeed resolving to the given
2333 if test(item_name) and os.path.basename(item_name) == name:
2338 def CheckVolumeGroupSize(vglist, vgname, minsize):
2339 """Checks if the volume group list is valid.
2341 The function will check if a given volume group is in the list of
2342 volume groups and has a minimum size.
2345 @param vglist: dictionary of volume group names and their size
2347 @param vgname: the volume group we should check
2349 @param minsize: the minimum size we accept
2351 @return: None for success, otherwise the error message
2354 vgsize = vglist.get(vgname, None)
2356 return "volume group '%s' missing" % vgname
2357 elif vgsize < minsize:
2358 return ("volume group '%s' too small (%s MiB required, %d MiB found)" %
2359 (vgname, minsize, vgsize))
2363 def SplitTime(value):
2364 """Splits time as floating point number into a tuple.
2366 @param value: Time in seconds
2367 @type value: int or float
2368 @return: Tuple containing (seconds, microseconds)
2371 (seconds, microseconds) = divmod(int(value * 1000000), 1000000)
2373 assert 0 <= seconds, \
2374 "Seconds must be larger than or equal to 0, but are %s" % seconds
2375 assert 0 <= microseconds <= 999999, \
2376 "Microseconds must be 0-999999, but are %s" % microseconds
2378 return (int(seconds), int(microseconds))
2381 def MergeTime(timetuple):
2382 """Merges a tuple into time as a floating point number.
2384 @param timetuple: Time as tuple, (seconds, microseconds)
2385 @type timetuple: tuple
2386 @return: Time as a floating point number expressed in seconds
2389 (seconds, microseconds) = timetuple
2391 assert 0 <= seconds, \
2392 "Seconds must be larger than or equal to 0, but are %s" % seconds
2393 assert 0 <= microseconds <= 999999, \
2394 "Microseconds must be 0-999999, but are %s" % microseconds
2396 return float(seconds) + (float(microseconds) * 0.000001)
2399 class LogFileHandler(logging.FileHandler):
2400 """Log handler that doesn't fallback to stderr.
2402 When an error occurs while writing on the logfile, logging.FileHandler tries
2403 to log on stderr. This doesn't work in ganeti since stderr is redirected to
2404 the logfile. This class avoids failures reporting errors to /dev/console.
2407 def __init__(self, filename, mode="a", encoding=None):
2408 """Open the specified file and use it as the stream for logging.
2410 Also open /dev/console to report errors while logging.
2413 logging.FileHandler.__init__(self, filename, mode, encoding)
2414 self.console = open(constants.DEV_CONSOLE, "a")
2416 def handleError(self, record): # pylint: disable-msg=C0103
2417 """Handle errors which occur during an emit() call.
2419 Try to handle errors with FileHandler method, if it fails write to
2424 logging.FileHandler.handleError(self, record)
2425 except Exception: # pylint: disable-msg=W0703
2427 self.console.write("Cannot log message:\n%s\n" % self.format(record))
2428 except Exception: # pylint: disable-msg=W0703
2429 # Log handler tried everything it could, now just give up
2433 def SetupLogging(logfile, debug=0, stderr_logging=False, program="",
2434 multithreaded=False, syslog=constants.SYSLOG_USAGE,
2435 console_logging=False):
2436 """Configures the logging module.
2439 @param logfile: the filename to which we should log
2440 @type debug: integer
2441 @param debug: if greater than zero, enable debug messages, otherwise
2442 only those at C{INFO} and above level
2443 @type stderr_logging: boolean
2444 @param stderr_logging: whether we should also log to the standard error
2446 @param program: the name under which we should log messages
2447 @type multithreaded: boolean
2448 @param multithreaded: if True, will add the thread name to the log file
2449 @type syslog: string
2450 @param syslog: one of 'no', 'yes', 'only':
2451 - if no, syslog is not used
2452 - if yes, syslog is used (in addition to file-logging)
2453 - if only, only syslog is used
2454 @type console_logging: boolean
2455 @param console_logging: if True, will use a FileHandler which falls back to
2456 the system console if logging fails
2457 @raise EnvironmentError: if we can't open the log file and
2458 syslog/stderr logging is disabled
2461 fmt = "%(asctime)s: " + program + " pid=%(process)d"
2462 sft = program + "[%(process)d]:"
2464 fmt += "/%(threadName)s"
2465 sft += " (%(threadName)s)"
2467 fmt += " %(module)s:%(lineno)s"
2468 # no debug info for syslog loggers
2469 fmt += " %(levelname)s %(message)s"
2470 # yes, we do want the textual level, as remote syslog will probably
2471 # lose the error level, and it's easier to grep for it
2472 sft += " %(levelname)s %(message)s"
2473 formatter = logging.Formatter(fmt)
2474 sys_fmt = logging.Formatter(sft)
2476 root_logger = logging.getLogger("")
2477 root_logger.setLevel(logging.NOTSET)
2479 # Remove all previously setup handlers
2480 for handler in root_logger.handlers:
2482 root_logger.removeHandler(handler)
2485 stderr_handler = logging.StreamHandler()
2486 stderr_handler.setFormatter(formatter)
2488 stderr_handler.setLevel(logging.NOTSET)
2490 stderr_handler.setLevel(logging.CRITICAL)
2491 root_logger.addHandler(stderr_handler)
2493 if syslog in (constants.SYSLOG_YES, constants.SYSLOG_ONLY):
2494 facility = logging.handlers.SysLogHandler.LOG_DAEMON
2495 syslog_handler = logging.handlers.SysLogHandler(constants.SYSLOG_SOCKET,
2497 syslog_handler.setFormatter(sys_fmt)
2498 # Never enable debug over syslog
2499 syslog_handler.setLevel(logging.INFO)
2500 root_logger.addHandler(syslog_handler)
2502 if syslog != constants.SYSLOG_ONLY:
2503 # this can fail, if the logging directories are not setup or we have
2504 # a permisssion problem; in this case, it's best to log but ignore
2505 # the error if stderr_logging is True, and if false we re-raise the
2506 # exception since otherwise we could run but without any logs at all
2509 logfile_handler = LogFileHandler(logfile)
2511 logfile_handler = logging.FileHandler(logfile)
2512 logfile_handler.setFormatter(formatter)
2514 logfile_handler.setLevel(logging.DEBUG)
2516 logfile_handler.setLevel(logging.INFO)
2517 root_logger.addHandler(logfile_handler)
2518 except EnvironmentError:
2519 if stderr_logging or syslog == constants.SYSLOG_YES:
2520 logging.exception("Failed to enable logging to file '%s'", logfile)
2522 # we need to re-raise the exception
2526 def IsNormAbsPath(path):
2527 """Check whether a path is absolute and also normalized
2529 This avoids things like /dir/../../other/path to be valid.
2532 return os.path.normpath(path) == path and os.path.isabs(path)
2535 def PathJoin(*args):
2536 """Safe-join a list of path components.
2539 - the first argument must be an absolute path
2540 - no component in the path must have backtracking (e.g. /../),
2541 since we check for normalization at the end
2543 @param args: the path components to be joined
2544 @raise ValueError: for invalid paths
2547 # ensure we're having at least one path passed in
2549 # ensure the first component is an absolute and normalized path name
2551 if not IsNormAbsPath(root):
2552 raise ValueError("Invalid parameter to PathJoin: '%s'" % str(args[0]))
2553 result = os.path.join(*args)
2554 # ensure that the whole path is normalized
2555 if not IsNormAbsPath(result):
2556 raise ValueError("Invalid parameters to PathJoin: '%s'" % str(args))
2557 # check that we're still under the original prefix
2558 prefix = os.path.commonprefix([root, result])
2560 raise ValueError("Error: path joining resulted in different prefix"
2561 " (%s != %s)" % (prefix, root))
2565 def TailFile(fname, lines=20):
2566 """Return the last lines from a file.
2568 @note: this function will only read and parse the last 4KB of
2569 the file; if the lines are very long, it could be that less
2570 than the requested number of lines are returned
2572 @param fname: the file name
2574 @param lines: the (maximum) number of lines to return
2577 fd = open(fname, "r")
2581 pos = max(0, pos-4096)
2583 raw_data = fd.read()
2587 rows = raw_data.splitlines()
2588 return rows[-lines:]
2591 def FormatTimestampWithTZ(secs):
2592 """Formats a Unix timestamp with the local timezone.
2595 return time.strftime("%F %T %Z", time.gmtime(secs))
2598 def _ParseAsn1Generalizedtime(value):
2599 """Parses an ASN1 GENERALIZEDTIME timestamp as used by pyOpenSSL.
2602 @param value: ASN1 GENERALIZEDTIME timestamp
2605 m = re.match(r"^(\d+)([-+]\d\d)(\d\d)$", value)
2608 asn1time = m.group(1)
2609 hours = int(m.group(2))
2610 minutes = int(m.group(3))
2611 utcoffset = (60 * hours) + minutes
2613 if not value.endswith("Z"):
2614 raise ValueError("Missing timezone")
2615 asn1time = value[:-1]
2618 parsed = time.strptime(asn1time, "%Y%m%d%H%M%S")
2620 tt = datetime.datetime(*(parsed[:7])) - datetime.timedelta(minutes=utcoffset)
2622 return calendar.timegm(tt.utctimetuple())
2625 def GetX509CertValidity(cert):
2626 """Returns the validity period of the certificate.
2628 @type cert: OpenSSL.crypto.X509
2629 @param cert: X509 certificate object
2632 # The get_notBefore and get_notAfter functions are only supported in
2633 # pyOpenSSL 0.7 and above.
2635 get_notbefore_fn = cert.get_notBefore
2636 except AttributeError:
2639 not_before_asn1 = get_notbefore_fn()
2641 if not_before_asn1 is None:
2644 not_before = _ParseAsn1Generalizedtime(not_before_asn1)
2647 get_notafter_fn = cert.get_notAfter
2648 except AttributeError:
2651 not_after_asn1 = get_notafter_fn()
2653 if not_after_asn1 is None:
2656 not_after = _ParseAsn1Generalizedtime(not_after_asn1)
2658 return (not_before, not_after)
2661 def _VerifyCertificateInner(expired, not_before, not_after, now,
2662 warn_days, error_days):
2663 """Verifies certificate validity.
2666 @param expired: Whether pyOpenSSL considers the certificate as expired
2667 @type not_before: number or None
2668 @param not_before: Unix timestamp before which certificate is not valid
2669 @type not_after: number or None
2670 @param not_after: Unix timestamp after which certificate is invalid
2672 @param now: Current time as Unix timestamp
2673 @type warn_days: number or None
2674 @param warn_days: How many days before expiration a warning should be reported
2675 @type error_days: number or None
2676 @param error_days: How many days before expiration an error should be reported
2680 msg = "Certificate is expired"
2682 if not_before is not None and not_after is not None:
2683 msg += (" (valid from %s to %s)" %
2684 (FormatTimestampWithTZ(not_before),
2685 FormatTimestampWithTZ(not_after)))
2686 elif not_before is not None:
2687 msg += " (valid from %s)" % FormatTimestampWithTZ(not_before)
2688 elif not_after is not None:
2689 msg += " (valid until %s)" % FormatTimestampWithTZ(not_after)
2691 return (CERT_ERROR, msg)
2693 elif not_before is not None and not_before > now:
2694 return (CERT_WARNING,
2695 "Certificate not yet valid (valid from %s)" %
2696 FormatTimestampWithTZ(not_before))
2698 elif not_after is not None:
2699 remaining_days = int((not_after - now) / (24 * 3600))
2701 msg = "Certificate expires in about %d days" % remaining_days
2703 if error_days is not None and remaining_days <= error_days:
2704 return (CERT_ERROR, msg)
2706 if warn_days is not None and remaining_days <= warn_days:
2707 return (CERT_WARNING, msg)
2712 def VerifyX509Certificate(cert, warn_days, error_days):
2713 """Verifies a certificate for LUVerifyCluster.
2715 @type cert: OpenSSL.crypto.X509
2716 @param cert: X509 certificate object
2717 @type warn_days: number or None
2718 @param warn_days: How many days before expiration a warning should be reported
2719 @type error_days: number or None
2720 @param error_days: How many days before expiration an error should be reported
2723 # Depending on the pyOpenSSL version, this can just return (None, None)
2724 (not_before, not_after) = GetX509CertValidity(cert)
2726 return _VerifyCertificateInner(cert.has_expired(), not_before, not_after,
2727 time.time(), warn_days, error_days)
2730 def SignX509Certificate(cert, key, salt):
2731 """Sign a X509 certificate.
2733 An RFC822-like signature header is added in front of the certificate.
2735 @type cert: OpenSSL.crypto.X509
2736 @param cert: X509 certificate object
2738 @param key: Key for HMAC
2740 @param salt: Salt for HMAC
2742 @return: Serialized and signed certificate in PEM format
2745 if not VALID_X509_SIGNATURE_SALT.match(salt):
2746 raise errors.GenericError("Invalid salt: %r" % salt)
2748 # Dumping as PEM here ensures the certificate is in a sane format
2749 cert_pem = OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM, cert)
2751 return ("%s: %s/%s\n\n%s" %
2752 (constants.X509_CERT_SIGNATURE_HEADER, salt,
2753 Sha1Hmac(key, cert_pem, salt=salt),
2757 def _ExtractX509CertificateSignature(cert_pem):
2758 """Helper function to extract signature from X509 certificate.
2761 # Extract signature from original PEM data
2762 for line in cert_pem.splitlines():
2763 if line.startswith("---"):
2766 m = X509_SIGNATURE.match(line.strip())
2768 return (m.group("salt"), m.group("sign"))
2770 raise errors.GenericError("X509 certificate signature is missing")
2773 def LoadSignedX509Certificate(cert_pem, key):
2774 """Verifies a signed X509 certificate.
2776 @type cert_pem: string
2777 @param cert_pem: Certificate in PEM format and with signature header
2779 @param key: Key for HMAC
2780 @rtype: tuple; (OpenSSL.crypto.X509, string)
2781 @return: X509 certificate object and salt
2784 (salt, signature) = _ExtractX509CertificateSignature(cert_pem)
2787 cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM, cert_pem)
2789 # Dump again to ensure it's in a sane format
2790 sane_pem = OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM, cert)
2792 if not VerifySha1Hmac(key, sane_pem, signature, salt=salt):
2793 raise errors.GenericError("X509 certificate signature is invalid")
2798 def Sha1Hmac(key, text, salt=None):
2799 """Calculates the HMAC-SHA1 digest of a text.
2801 HMAC is defined in RFC2104.
2804 @param key: Secret key
2809 salted_text = salt + text
2813 return hmac.new(key, salted_text, compat.sha1).hexdigest()
2816 def VerifySha1Hmac(key, text, digest, salt=None):
2817 """Verifies the HMAC-SHA1 digest of a text.
2819 HMAC is defined in RFC2104.
2822 @param key: Secret key
2824 @type digest: string
2825 @param digest: Expected digest
2827 @return: Whether HMAC-SHA1 digest matches
2830 return digest.lower() == Sha1Hmac(key, text, salt=salt).lower()
2833 def SafeEncode(text):
2834 """Return a 'safe' version of a source string.
2836 This function mangles the input string and returns a version that
2837 should be safe to display/encode as ASCII. To this end, we first
2838 convert it to ASCII using the 'backslashreplace' encoding which
2839 should get rid of any non-ASCII chars, and then we process it
2840 through a loop copied from the string repr sources in the python; we
2841 don't use string_escape anymore since that escape single quotes and
2842 backslashes too, and that is too much; and that escaping is not
2843 stable, i.e. string_escape(string_escape(x)) != string_escape(x).
2845 @type text: str or unicode
2846 @param text: input data
2848 @return: a safe version of text
2851 if isinstance(text, unicode):
2852 # only if unicode; if str already, we handle it below
2853 text = text.encode('ascii', 'backslashreplace')
2863 elif c < 32 or c >= 127: # non-printable
2864 resu += "\\x%02x" % (c & 0xff)
2870 def UnescapeAndSplit(text, sep=","):
2871 """Split and unescape a string based on a given separator.
2873 This function splits a string based on a separator where the
2874 separator itself can be escape in order to be an element of the
2875 elements. The escaping rules are (assuming coma being the
2877 - a plain , separates the elements
2878 - a sequence \\\\, (double backslash plus comma) is handled as a
2879 backslash plus a separator comma
2880 - a sequence \, (backslash plus comma) is handled as a
2884 @param text: the string to split
2886 @param text: the separator
2888 @return: a list of strings
2891 # we split the list by sep (with no escaping at this stage)
2892 slist = text.split(sep)
2893 # next, we revisit the elements and if any of them ended with an odd
2894 # number of backslashes, then we join it with the next
2898 if e1.endswith("\\"):
2899 num_b = len(e1) - len(e1.rstrip("\\"))
2902 # here the backslashes remain (all), and will be reduced in
2904 rlist.append(e1 + sep + e2)
2907 # finally, replace backslash-something with something
2908 rlist = [re.sub(r"\\(.)", r"\1", v) for v in rlist]
2912 def CommaJoin(names):
2913 """Nicely join a set of identifiers.
2915 @param names: set, list or tuple
2916 @return: a string with the formatted results
2919 return ", ".join([str(val) for val in names])
2922 def BytesToMebibyte(value):
2923 """Converts bytes to mebibytes.
2926 @param value: Value in bytes
2928 @return: Value in mebibytes
2931 return int(round(value / (1024.0 * 1024.0), 0))
2934 def CalculateDirectorySize(path):
2935 """Calculates the size of a directory recursively.
2938 @param path: Path to directory
2940 @return: Size in mebibytes
2945 for (curpath, _, files) in os.walk(path):
2946 for filename in files:
2947 st = os.lstat(PathJoin(curpath, filename))
2950 return BytesToMebibyte(size)
2953 def GetMounts(filename=constants.PROC_MOUNTS):
2954 """Returns the list of mounted filesystems.
2956 This function is Linux-specific.
2958 @param filename: path of mounts file (/proc/mounts by default)
2959 @rtype: list of tuples
2960 @return: list of mount entries (device, mountpoint, fstype, options)
2963 # TODO(iustin): investigate non-Linux options (e.g. via mount output)
2965 mountlines = ReadFile(filename).splitlines()
2966 for line in mountlines:
2967 device, mountpoint, fstype, options, _ = line.split(None, 4)
2968 data.append((device, mountpoint, fstype, options))
2973 def GetFilesystemStats(path):
2974 """Returns the total and free space on a filesystem.
2977 @param path: Path on filesystem to be examined
2979 @return: tuple of (Total space, Free space) in mebibytes
2982 st = os.statvfs(path)
2984 fsize = BytesToMebibyte(st.f_bavail * st.f_frsize)
2985 tsize = BytesToMebibyte(st.f_blocks * st.f_frsize)
2986 return (tsize, fsize)
2989 def RunInSeparateProcess(fn, *args):
2990 """Runs a function in a separate process.
2992 Note: Only boolean return values are supported.
2995 @param fn: Function to be called
2997 @return: Function's result
3004 # In case the function uses temporary files
3005 ResetTempfileModule()
3008 result = int(bool(fn(*args)))
3009 assert result in (0, 1)
3010 except: # pylint: disable-msg=W0702
3011 logging.exception("Error while calling function in separate process")
3012 # 0 and 1 are reserved for the return value
3015 os._exit(result) # pylint: disable-msg=W0212
3019 # Avoid zombies and check exit code
3020 (_, status) = os.waitpid(pid, 0)
3022 if os.WIFSIGNALED(status):
3024 signum = os.WTERMSIG(status)
3026 exitcode = os.WEXITSTATUS(status)
3029 if not (exitcode in (0, 1) and signum is None):
3030 raise errors.GenericError("Child program failed (code=%s, signal=%s)" %
3033 return bool(exitcode)
3036 def IgnoreProcessNotFound(fn, *args, **kwargs):
3037 """Ignores ESRCH when calling a process-related function.
3039 ESRCH is raised when a process is not found.
3042 @return: Whether process was found
3047 except EnvironmentError, err:
3049 if err.errno == errno.ESRCH:
3056 def IgnoreSignals(fn, *args, **kwargs):
3057 """Tries to call a function ignoring failures due to EINTR.
3061 return fn(*args, **kwargs)
3062 except EnvironmentError, err:
3063 if err.errno == errno.EINTR:
3067 except (select.error, socket.error), err:
3068 # In python 2.6 and above select.error is an IOError, so it's handled
3069 # above, in 2.5 and below it's not, and it's handled here.
3070 if err.args and err.args[0] == errno.EINTR:
3077 """Locks a file using POSIX locks.
3080 @param fd: the file descriptor we need to lock
3084 fcntl.flock(fd, fcntl.LOCK_EX | fcntl.LOCK_NB)
3085 except IOError, err:
3086 if err.errno == errno.EAGAIN:
3087 raise errors.LockError("File already locked")
3091 def FormatTime(val):
3092 """Formats a time value.
3094 @type val: float or None
3095 @param val: the timestamp as returned by time.time()
3096 @return: a string value or N/A if we don't have a valid timestamp
3099 if val is None or not isinstance(val, (int, float)):
3101 # these two codes works on Linux, but they are not guaranteed on all
3103 return time.strftime("%F %T", time.localtime(val))
3106 def FormatSeconds(secs):
3107 """Formats seconds for easier reading.
3110 @param secs: Number of seconds
3112 @return: Formatted seconds (e.g. "2d 9h 19m 49s")
3117 secs = round(secs, 0)
3120 # Negative values would be a bit tricky
3121 for unit, one in [("d", 24 * 60 * 60), ("h", 60 * 60), ("m", 60)]:
3122 (complete, secs) = divmod(secs, one)
3123 if complete or parts:
3124 parts.append("%d%s" % (complete, unit))
3126 parts.append("%ds" % secs)
3128 return " ".join(parts)
3131 def ReadWatcherPauseFile(filename, now=None, remove_after=3600):
3132 """Reads the watcher pause file.
3134 @type filename: string
3135 @param filename: Path to watcher pause file
3136 @type now: None, float or int
3137 @param now: Current time as Unix timestamp
3138 @type remove_after: int
3139 @param remove_after: Remove watcher pause file after specified amount of
3140 seconds past the pause end time
3147 value = ReadFile(filename)
3148 except IOError, err:
3149 if err.errno != errno.ENOENT:
3153 if value is not None:
3157 logging.warning(("Watcher pause file (%s) contains invalid value,"
3158 " removing it"), filename)
3159 RemoveFile(filename)
3162 if value is not None:
3163 # Remove file if it's outdated
3164 if now > (value + remove_after):
3165 RemoveFile(filename)
3174 class RetryTimeout(Exception):
3175 """Retry loop timed out.
3177 Any arguments which was passed by the retried function to RetryAgain will be
3178 preserved in RetryTimeout, if it is raised. If such argument was an exception
3179 the RaiseInner helper method will reraise it.
3182 def RaiseInner(self):
3183 if self.args and isinstance(self.args[0], Exception):
3186 raise RetryTimeout(*self.args)
3189 class RetryAgain(Exception):
3192 Any arguments passed to RetryAgain will be preserved, if a timeout occurs, as
3193 arguments to RetryTimeout. If an exception is passed, the RaiseInner() method
3194 of the RetryTimeout() method can be used to reraise it.
3199 class _RetryDelayCalculator(object):
3200 """Calculator for increasing delays.
3210 def __init__(self, start, factor, limit):
3211 """Initializes this class.
3214 @param start: Initial delay
3216 @param factor: Factor for delay increase
3217 @type limit: float or None
3218 @param limit: Upper limit for delay or None for no limit
3222 assert factor >= 1.0
3223 assert limit is None or limit >= 0.0
3226 self._factor = factor
3232 """Returns current delay and calculates the next one.
3235 current = self._next
3237 # Update for next run
3238 if self._limit is None or self._next < self._limit:
3239 self._next = min(self._limit, self._next * self._factor)
3244 #: Special delay to specify whole remaining timeout
3245 RETRY_REMAINING_TIME = object()
3248 def Retry(fn, delay, timeout, args=None, wait_fn=time.sleep,
3249 _time_fn=time.time):
3250 """Call a function repeatedly until it succeeds.
3252 The function C{fn} is called repeatedly until it doesn't throw L{RetryAgain}
3253 anymore. Between calls a delay, specified by C{delay}, is inserted. After a
3254 total of C{timeout} seconds, this function throws L{RetryTimeout}.
3256 C{delay} can be one of the following:
3257 - callable returning the delay length as a float
3258 - Tuple of (start, factor, limit)
3259 - L{RETRY_REMAINING_TIME} to sleep until the timeout expires (this is
3260 useful when overriding L{wait_fn} to wait for an external event)
3261 - A static delay as a number (int or float)
3264 @param fn: Function to be called
3265 @param delay: Either a callable (returning the delay), a tuple of (start,
3266 factor, limit) (see L{_RetryDelayCalculator}),
3267 L{RETRY_REMAINING_TIME} or a number (int or float)
3268 @type timeout: float
3269 @param timeout: Total timeout
3270 @type wait_fn: callable
3271 @param wait_fn: Waiting function
3272 @return: Return value of function
3276 assert callable(wait_fn)
3277 assert callable(_time_fn)
3282 end_time = _time_fn() + timeout
3285 # External function to calculate delay
3288 elif isinstance(delay, (tuple, list)):
3289 # Increasing delay with optional upper boundary
3290 (start, factor, limit) = delay
3291 calc_delay = _RetryDelayCalculator(start, factor, limit)
3293 elif delay is RETRY_REMAINING_TIME:
3294 # Always use the remaining time
3299 calc_delay = lambda: delay
3301 assert calc_delay is None or callable(calc_delay)
3306 # pylint: disable-msg=W0142
3308 except RetryAgain, err:
3309 retry_args = err.args
3310 except RetryTimeout:
3311 raise errors.ProgrammerError("Nested retry loop detected that didn't"
3312 " handle RetryTimeout")
3314 remaining_time = end_time - _time_fn()
3316 if remaining_time < 0.0:
3317 # pylint: disable-msg=W0142
3318 raise RetryTimeout(*retry_args)
3320 assert remaining_time >= 0.0
3322 if calc_delay is None:
3323 wait_fn(remaining_time)
3325 current_delay = calc_delay()
3326 if current_delay > 0.0:
3327 wait_fn(current_delay)
3330 def GetClosedTempfile(*args, **kwargs):
3331 """Creates a temporary file and returns its path.
3334 (fd, path) = tempfile.mkstemp(*args, **kwargs)
3339 def GenerateSelfSignedX509Cert(common_name, validity):
3340 """Generates a self-signed X509 certificate.
3342 @type common_name: string
3343 @param common_name: commonName value
3345 @param validity: Validity for certificate in seconds
3348 # Create private and public key
3349 key = OpenSSL.crypto.PKey()
3350 key.generate_key(OpenSSL.crypto.TYPE_RSA, constants.RSA_KEY_BITS)
3352 # Create self-signed certificate
3353 cert = OpenSSL.crypto.X509()
3355 cert.get_subject().CN = common_name
3356 cert.set_serial_number(1)
3357 cert.gmtime_adj_notBefore(0)
3358 cert.gmtime_adj_notAfter(validity)
3359 cert.set_issuer(cert.get_subject())
3360 cert.set_pubkey(key)
3361 cert.sign(key, constants.X509_CERT_SIGN_DIGEST)
3363 key_pem = OpenSSL.crypto.dump_privatekey(OpenSSL.crypto.FILETYPE_PEM, key)
3364 cert_pem = OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM, cert)
3366 return (key_pem, cert_pem)
3369 def GenerateSelfSignedSslCert(filename, common_name=constants.X509_CERT_CN,
3370 validity=constants.X509_CERT_DEFAULT_VALIDITY):
3371 """Legacy function to generate self-signed X509 certificate.
3374 @param filename: path to write certificate to
3375 @type common_name: string
3376 @param common_name: commonName value
3378 @param validity: validity of certificate in number of days
3381 # TODO: Investigate using the cluster name instead of X505_CERT_CN for
3382 # common_name, as cluster-renames are very seldom, and it'd be nice if RAPI
3383 # and node daemon certificates have the proper Subject/Issuer.
3384 (key_pem, cert_pem) = GenerateSelfSignedX509Cert(common_name,
3385 validity * 24 * 60 * 60)
3387 WriteFile(filename, mode=0400, data=key_pem + cert_pem)
3390 class FileLock(object):
3391 """Utility class for file locks.
3394 def __init__(self, fd, filename):
3395 """Constructor for FileLock.
3398 @param fd: File object
3400 @param filename: Path of the file opened at I{fd}
3404 self.filename = filename
3407 def Open(cls, filename):
3408 """Creates and opens a file to be used as a file-based lock.
3410 @type filename: string
3411 @param filename: path to the file to be locked
3414 # Using "os.open" is necessary to allow both opening existing file
3415 # read/write and creating if not existing. Vanilla "open" will truncate an
3416 # existing file -or- allow creating if not existing.
3417 return cls(os.fdopen(os.open(filename, os.O_RDWR | os.O_CREAT), "w+"),
3424 """Close the file and release the lock.
3427 if hasattr(self, "fd") and self.fd:
3431 def _flock(self, flag, blocking, timeout, errmsg):
3432 """Wrapper for fcntl.flock.
3435 @param flag: operation flag
3436 @type blocking: bool
3437 @param blocking: whether the operation should be done in blocking mode.
3438 @type timeout: None or float
3439 @param timeout: for how long the operation should be retried (implies
3441 @type errmsg: string
3442 @param errmsg: error message in case operation fails.
3445 assert self.fd, "Lock was closed"
3446 assert timeout is None or timeout >= 0, \
3447 "If specified, timeout must be positive"
3448 assert not (flag & fcntl.LOCK_NB), "LOCK_NB must not be set"
3450 # When a timeout is used, LOCK_NB must always be set
3451 if not (timeout is None and blocking):
3452 flag |= fcntl.LOCK_NB
3455 self._Lock(self.fd, flag, timeout)
3458 Retry(self._Lock, (0.1, 1.2, 1.0), timeout,
3459 args=(self.fd, flag, timeout))
3460 except RetryTimeout:
3461 raise errors.LockError(errmsg)
3464 def _Lock(fd, flag, timeout):
3466 fcntl.flock(fd, flag)
3467 except IOError, err:
3468 if timeout is not None and err.errno == errno.EAGAIN:
3471 logging.exception("fcntl.flock failed")
3474 def Exclusive(self, blocking=False, timeout=None):
3475 """Locks the file in exclusive mode.
3477 @type blocking: boolean
3478 @param blocking: whether to block and wait until we
3479 can lock the file or return immediately
3480 @type timeout: int or None
3481 @param timeout: if not None, the duration to wait for the lock
3485 self._flock(fcntl.LOCK_EX, blocking, timeout,
3486 "Failed to lock %s in exclusive mode" % self.filename)
3488 def Shared(self, blocking=False, timeout=None):
3489 """Locks the file in shared mode.
3491 @type blocking: boolean
3492 @param blocking: whether to block and wait until we
3493 can lock the file or return immediately
3494 @type timeout: int or None
3495 @param timeout: if not None, the duration to wait for the lock
3499 self._flock(fcntl.LOCK_SH, blocking, timeout,
3500 "Failed to lock %s in shared mode" % self.filename)
3502 def Unlock(self, blocking=True, timeout=None):
3503 """Unlocks the file.
3505 According to C{flock(2)}, unlocking can also be a nonblocking
3508 To make a non-blocking request, include LOCK_NB with any of the above
3511 @type blocking: boolean
3512 @param blocking: whether to block and wait until we
3513 can lock the file or return immediately
3514 @type timeout: int or None
3515 @param timeout: if not None, the duration to wait for the lock
3519 self._flock(fcntl.LOCK_UN, blocking, timeout,
3520 "Failed to unlock %s" % self.filename)
3524 """Splits data chunks into lines separated by newline.
3526 Instances provide a file-like interface.
3529 def __init__(self, line_fn, *args):
3530 """Initializes this class.
3532 @type line_fn: callable
3533 @param line_fn: Function called for each line, first parameter is line
3534 @param args: Extra arguments for L{line_fn}
3537 assert callable(line_fn)
3540 # Python 2.4 doesn't have functools.partial yet
3542 lambda line: line_fn(line, *args) # pylint: disable-msg=W0142
3544 self._line_fn = line_fn
3546 self._lines = collections.deque()
3549 def write(self, data):
3550 parts = (self._buffer + data).split("\n")
3551 self._buffer = parts.pop()
3552 self._lines.extend(parts)
3556 self._line_fn(self._lines.popleft().rstrip("\r\n"))
3561 self._line_fn(self._buffer)
3564 def SignalHandled(signums):
3565 """Signal Handled decoration.
3567 This special decorator installs a signal handler and then calls the target
3568 function. The function must accept a 'signal_handlers' keyword argument,
3569 which will contain a dict indexed by signal number, with SignalHandler
3572 The decorator can be safely stacked with iself, to handle multiple signals
3573 with different handlers.
3576 @param signums: signals to intercept
3580 def sig_function(*args, **kwargs):
3581 assert 'signal_handlers' not in kwargs or \
3582 kwargs['signal_handlers'] is None or \
3583 isinstance(kwargs['signal_handlers'], dict), \
3584 "Wrong signal_handlers parameter in original function call"
3585 if 'signal_handlers' in kwargs and kwargs['signal_handlers'] is not None:
3586 signal_handlers = kwargs['signal_handlers']
3588 signal_handlers = {}
3589 kwargs['signal_handlers'] = signal_handlers
3590 sighandler = SignalHandler(signums)
3593 signal_handlers[sig] = sighandler
3594 return fn(*args, **kwargs)
3601 class SignalWakeupFd(object):
3603 # This is only supported in Python 2.5 and above (some distributions
3604 # backported it to Python 2.4)
3605 _set_wakeup_fd_fn = signal.set_wakeup_fd
3606 except AttributeError:
3608 def _SetWakeupFd(self, _): # pylint: disable-msg=R0201
3611 def _SetWakeupFd(self, fd):
3612 return self._set_wakeup_fd_fn(fd)
3615 """Initializes this class.
3618 (read_fd, write_fd) = os.pipe()
3620 # Once these succeeded, the file descriptors will be closed automatically.
3621 # Buffer size 0 is important, otherwise .read() with a specified length
3622 # might buffer data and the file descriptors won't be marked readable.
3623 self._read_fh = os.fdopen(read_fd, "r", 0)
3624 self._write_fh = os.fdopen(write_fd, "w", 0)
3626 self._previous = self._SetWakeupFd(self._write_fh.fileno())
3629 self.fileno = self._read_fh.fileno
3630 self.read = self._read_fh.read
3633 """Restores the previous wakeup file descriptor.
3636 if hasattr(self, "_previous") and self._previous is not None:
3637 self._SetWakeupFd(self._previous)
3638 self._previous = None
3641 """Notifies the wakeup file descriptor.
3644 self._write_fh.write("\0")
3647 """Called before object deletion.
3653 class SignalHandler(object):
3654 """Generic signal handler class.
3656 It automatically restores the original handler when deconstructed or
3657 when L{Reset} is called. You can either pass your own handler
3658 function in or query the L{called} attribute to detect whether the
3662 @ivar signum: the signals we handle
3663 @type called: boolean
3664 @ivar called: tracks whether any of the signals have been raised
3667 def __init__(self, signum, handler_fn=None, wakeup=None):
3668 """Constructs a new SignalHandler instance.
3670 @type signum: int or list of ints
3671 @param signum: Single signal number or set of signal numbers
3672 @type handler_fn: callable
3673 @param handler_fn: Signal handling function
3676 assert handler_fn is None or callable(handler_fn)
3678 self.signum = set(signum)
3681 self._handler_fn = handler_fn
3682 self._wakeup = wakeup
3686 for signum in self.signum:
3688 prev_handler = signal.signal(signum, self._HandleSignal)
3690 self._previous[signum] = prev_handler
3692 # Restore previous handler
3693 signal.signal(signum, prev_handler)
3696 # Reset all handlers
3698 # Here we have a race condition: a handler may have already been called,
3699 # but there's not much we can do about it at this point.
3706 """Restore previous handler.
3708 This will reset all the signals to their previous handlers.
3711 for signum, prev_handler in self._previous.items():
3712 signal.signal(signum, prev_handler)
3713 # If successful, remove from dict
3714 del self._previous[signum]
3717 """Unsets the L{called} flag.
3719 This function can be used in case a signal may arrive several times.
3724 def _HandleSignal(self, signum, frame):
3725 """Actual signal handling function.
3728 # This is not nice and not absolutely atomic, but it appears to be the only
3729 # solution in Python -- there are no atomic types.
3733 # Notify whoever is interested in signals
3734 self._wakeup.Notify()
3736 if self._handler_fn:
3737 self._handler_fn(signum, frame)
3740 class FieldSet(object):
3741 """A simple field set.
3743 Among the features are:
3744 - checking if a string is among a list of static string or regex objects
3745 - checking if a whole list of string matches
3746 - returning the matching groups from a regex match
3748 Internally, all fields are held as regular expression objects.
3751 def __init__(self, *items):
3752 self.items = [re.compile("^%s$" % value) for value in items]
3754 def Extend(self, other_set):
3755 """Extend the field set with the items from another one"""
3756 self.items.extend(other_set.items)
3758 def Matches(self, field):
3759 """Checks if a field matches the current set
3762 @param field: the string to match
3763 @return: either None or a regular expression match object
3766 for m in itertools.ifilter(None, (val.match(field) for val in self.items)):
3770 def NonMatching(self, items):
3771 """Returns the list of fields not matching the current set
3774 @param items: the list of fields to check
3776 @return: list of non-matching fields
3779 return [val for val in items if not self.Matches(val)]