2 # NEPI, a framework to manage network experiments
3 # Copyright (C) 2013 INRIA
5 # This program is free software: you can redistribute it and/or modify
6 # it under the terms of the GNU General Public License as published by
7 # the Free Software Foundation, either version 3 of the License, or
8 # (at your option) any later version.
10 # This program is distributed in the hope that it will be useful,
11 # but WITHOUT ANY WARRANTY; without even the implied warranty of
12 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 # GNU General Public License for more details.
15 # You should have received a copy of the GNU General Public License
16 # along with this program. If not, see <http://www.gnu.org/licenses/>.
18 # Author: Alina Quereilhac <alina.quereilhac@inria.fr>
19 # Claudio Freire <claudio-daniel.freire@inria.fr>
21 ## TODO: This code needs reviewing !!!
38 logger = logging.getLogger("sshfuncs")
40 def log(msg, level, out = None, err = None):
42 msg += " - OUT: %s " % out
45 msg += " - ERROR: %s " % err
47 logger.log(level, msg)
50 if hasattr(os, "devnull"):
53 DEV_NULL = "/dev/null"
55 SHELL_SAFE = re.compile('^[-a-zA-Z0-9_=+:.,/]*$')
59 Special value that when given to rspawn in stderr causes stderr to
60 redirect to whatever stdout was redirected to.
65 Codes for status of remote spawned process
67 # Process is still running
73 # Process hasn't started running yet (this should be very rare)
76 hostbyname_cache = dict()
77 hostbyname_cache_lock = threading.Lock()
79 def gethostbyname(host):
80 global hostbyname_cache
81 global hostbyname_cache_lock
83 hostbyname = hostbyname_cache.get(host)
85 with hostbyname_cache_lock:
86 hostbyname = socket.gethostbyname(host)
87 hostbyname_cache[host] = hostbyname
89 msg = " Added hostbyname %s - %s " % (host, hostbyname)
90 log(msg, logging.DEBUG)
94 OPENSSH_HAS_PERSIST = None
96 def openssh_has_persist():
97 """ The ssh_config options ControlMaster and ControlPersist allow to
98 reuse a same network connection for multiple ssh sessions. In this
99 way limitations on number of open ssh connections can be bypassed.
100 However, older versions of openSSH do not support this feature.
101 This function is used to determine if ssh connection persist features
104 global OPENSSH_HAS_PERSIST
105 if OPENSSH_HAS_PERSIST is None:
106 proc = subprocess.Popen(["ssh","-v"],
107 stdout = subprocess.PIPE,
108 stderr = subprocess.STDOUT,
109 stdin = open("/dev/null","r") )
110 out,err = proc.communicate()
113 vre = re.compile(r'OpenSSH_(?:[6-9]|5[.][8-9]|5[.][1-9][0-9]|[1-9][0-9]).*', re.I)
114 OPENSSH_HAS_PERSIST = bool(vre.match(out))
115 return OPENSSH_HAS_PERSIST
117 def make_server_key_args(server_key, host, port):
118 """ Returns a reference to a temporary known_hosts file, to which
119 the server key has been added.
121 Make sure to hold onto the temp file reference until the process is
124 :param server_key: the server public key
125 :type server_key: str
127 :param host: the hostname
130 :param port: the ssh port
135 host = '%s:%s' % (host, str(port))
137 # Create a temporary server key file
138 tmp_known_hosts = tempfile.NamedTemporaryFile()
140 hostbyname = gethostbyname(host)
142 # Add the intended host key
143 tmp_known_hosts.write('%s,%s %s\n' % (host, hostbyname, server_key))
145 # If we're not in strict mode, add user-configured keys
146 if os.environ.get('NEPI_STRICT_AUTH_MODE',"").lower() not in ('1','true','on'):
147 user_hosts_path = '%s/.ssh/known_hosts' % (os.environ.get('HOME',""),)
148 if os.access(user_hosts_path, os.R_OK):
149 f = open(user_hosts_path, "r")
150 tmp_known_hosts.write(f.read())
153 tmp_known_hosts.flush()
155 return tmp_known_hosts
157 def make_control_path(agent, forward_x11):
158 ctrl_path = "/tmp/nepi_ssh"
166 ctrl_path += "-%r@%h:%p"
171 """ Escapes strings so that they are safe to use as command-line
173 if SHELL_SAFE.match(s):
174 # safe string - no escaping needed
177 # unsafe string - escape
179 if (32 <= ord(c) < 127 or c in ('\r','\n','\t')) and c not in ("'",'"'):
182 return "'$'\\x%02x''" % (ord(c),)
183 s = ''.join(map(escp,s))
186 def eintr_retry(func):
187 """Retries a function invocation when a EINTR occurs"""
189 @functools.wraps(func)
191 retry = kw.pop("_retry", False)
192 for i in xrange(0 if retry else 4):
194 return func(*p, **kw)
195 except (select.error, socket.error), args:
196 if args[0] == errno.EINTR:
201 if e.errno == errno.EINTR:
206 return func(*p, **kw)
209 def rexec(command, host, user,
218 connect_timeout = 30,
223 strict_host_checking = True):
225 Executes a remote command, returns ((stdout,stderr),process)
228 tmp_known_hosts = None
229 hostip = gethostbyname(host)
232 # Don't bother with localhost. Makes test easier
233 '-o', 'NoHostAuthenticationForLocalhost=yes',
234 '-o', 'ConnectTimeout=%d' % (int(connect_timeout),),
235 '-o', 'ConnectionAttempts=3',
236 '-o', 'ServerAliveInterval=30',
237 '-o', 'TCPKeepAlive=yes',
238 '-l', user, hostip or host]
240 if persistent and openssh_has_persist():
242 '-o', 'ControlMaster=auto',
243 '-o', 'ControlPath=%s' % (make_control_path(agent, forward_x11),),
244 '-o', 'ControlPersist=60' ])
246 if not strict_host_checking:
247 # Do not check for Host key. Unsafe.
248 args.extend(['-o', 'StrictHostKeyChecking=no'])
254 args.append('-p%d' % port)
257 args.extend(('-i', identity))
267 # Create a temporary server key file
268 tmp_known_hosts = make_server_key_args(server_key, host, port)
269 args.extend(['-o', 'UserKnownHostsFile=%s' % (tmp_known_hosts.name,)])
272 command = "sudo " + command
276 log_msg = " rexec - host %s - command %s " % (host, " ".join(args))
278 return _retry_rexec(args, log_msg, env = env, retry = retry,
279 tmp_known_hosts = tmp_known_hosts,
282 def rcopy(source, dest,
289 strict_host_checking = True):
291 Copies from/to remote sites.
293 Source and destination should have the user and host encoded
296 Source can be a list of files to copy to a single destination,
297 in which case it is advised that the destination be a folder.
300 # Parse destination as <user>@<server>:<path>
301 if isinstance(dest, basestring) and ':' in dest:
302 remspec, path = dest.split(':',1)
303 elif isinstance(source, basestring) and ':' in source:
304 remspec, path = source.split(':',1)
306 raise ValueError, "Both endpoints cannot be local"
307 user,host = remspec.rsplit('@',1)
310 tmp_known_hosts = None
312 args = ['scp', '-q', '-p', '-C',
313 # Speed up transfer using blowfish cypher specification which is
314 # faster than the default one (3des)
316 # Don't bother with localhost. Makes test easier
317 '-o', 'NoHostAuthenticationForLocalhost=yes',
318 '-o', 'ConnectTimeout=60',
319 '-o', 'ConnectionAttempts=3',
320 '-o', 'ServerAliveInterval=30',
321 '-o', 'TCPKeepAlive=yes' ]
324 args.append('-P%d' % port)
330 args.extend(('-i', identity))
333 # Create a temporary server key file
334 tmp_known_hosts = make_server_key_args(server_key, host, port)
335 args.extend(['-o', 'UserKnownHostsFile=%s' % (tmp_known_hosts.name,)])
337 if not strict_host_checking:
338 # Do not check for Host key. Unsafe.
339 args.extend(['-o', 'StrictHostKeyChecking=no'])
341 if isinstance(source,list):
344 if openssh_has_persist():
346 '-o', 'ControlMaster=auto',
347 '-o', 'ControlPath=%s' % (make_control_path(agent, False),)
353 log_msg = " rcopy - host %s - command %s " % (host, " ".join(args))
355 return _retry_rexec(args, log_msg, env = None, retry = retry,
356 tmp_known_hosts = tmp_known_hosts,
359 def rspawn(command, pidfile,
360 stdout = '/dev/null',
374 Spawn a remote command such that it will continue working asynchronously in
377 :param command: The command to run, it should be a single line.
380 :param pidfile: Path to a file where to store the pid and ppid of the
384 :param stdout: Path to file to redirect standard output.
385 The default value is /dev/null
388 :param stderr: Path to file to redirect standard error.
389 If the special STDOUT value is used, stderr will
390 be redirected to the same file as stdout
393 :param stdin: Path to a file with input to be piped into the command's standard input
396 :param home: Path to working directory folder.
397 It is assumed to exist unless the create_home flag is set.
400 :param create_home: Flag to force creation of the home folder before
402 :type create_home: bool
404 :param sudo: Flag forcing execution with sudo user
409 (stdout, stderr), process
411 Of the spawning process, which only captures errors at spawning time.
412 Usually only useful for diagnostics.
414 # Start process in a "daemonized" way, using nohup and heavy
415 # stdin/out redirection to avoid connection issues
419 stderr = ' ' + stderr
421 daemon_command = '{ { %(command)s > %(stdout)s 2>%(stderr)s < %(stdin)s & } ; echo $! 1 > %(pidfile)s ; }' % {
423 'pidfile' : shell_escape(pidfile),
429 cmd = "%(create)s%(gohome)s rm -f %(pidfile)s ; %(sudo)s nohup bash -c %(command)s " % {
430 'command' : shell_escape(daemon_command),
431 'sudo' : 'sudo -S' if sudo else '',
432 'pidfile' : shell_escape(pidfile),
433 'gohome' : 'cd %s ; ' % (shell_escape(home),) if home else '',
434 'create' : 'mkdir -p %s ; ' % (shell_escape(home),) if create_home and home else '',
437 (out,err),proc = rexec(
444 server_key = server_key,
449 raise RuntimeError, "Failed to set up application on host %s: %s %s" % (host, out,err,)
451 return ((out, err), proc)
462 Returns the pid and ppid of a process from a remote file where the
463 information was stored.
465 :param home: Path to directory where the pidfile is located
468 :param pidfile: Name of file containing the pid information
473 A (pid, ppid) tuple useful for calling rstatus and rkill,
474 or None if the pidfile isn't valid yet (can happen when process is staring up)
477 (out,err),proc = rexec(
478 "cat %(pidfile)s" % {
486 server_key = server_key
494 return map(int,out.strip().split(' ',1))
496 # Ignore, many ways to fail that don't matter that much
500 def rstatus(pid, ppid,
508 Returns a code representing the the status of a remote process
510 :param pid: Process id of the process
513 :param ppid: Parent process id of process
516 :rtype: int (One of NOT_STARTED, RUNNING, FINISHED)
519 (out,err),proc = rexec(
520 # Check only by pid. pid+ppid does not always work (especially with sudo)
521 " (( ps --pid %(pid)d -o pid | grep -c %(pid)d && echo 'wait') || echo 'done' ) | tail -n 1" % {
530 server_key = server_key
534 return ProcStatus.NOT_STARTED
538 if err.strip().find("Error, do this: mount -t proc none /proc") >= 0:
541 status = (out.strip() == 'wait')
543 return ProcStatus.NOT_STARTED
544 return ProcStatus.RUNNING if status else ProcStatus.FINISHED
557 Sends a kill signal to a remote process.
559 First tries a SIGTERM, and if the process does not end in 10 seconds,
562 :param pid: Process id of process to be killed
565 :param ppid: Parent process id of process to be killed
568 :param sudo: Flag indicating if sudo should be used to kill the process
572 subkill = "$(ps --ppid %(pid)d -o pid h)" % { 'pid' : pid }
574 SUBKILL="%(subkill)s" ;
575 %(sudo)s kill -- -%(pid)d $SUBKILL || /bin/true
576 %(sudo)s kill %(pid)d $SUBKILL || /bin/true
577 for x in 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 ; do
579 if [ `ps --pid %(pid)d -o pid | grep -c %(pid)d` == '0' ]; then
582 %(sudo)s kill -- -%(pid)d $SUBKILL || /bin/true
583 %(sudo)s kill %(pid)d $SUBKILL || /bin/true
587 if [ `ps --pid %(pid)d -o pid | grep -c %(pid)d` != '0' ]; then
588 %(sudo)s kill -9 -- -%(pid)d $SUBKILL || /bin/true
589 %(sudo)s kill -9 %(pid)d $SUBKILL || /bin/true
593 cmd = "( %s ) >/dev/null 2>/dev/null </dev/null &" % (cmd,)
595 (out,err),proc = rexec(
599 'sudo' : 'sudo -S' if sudo else '',
607 server_key = server_key
610 # wait, don't leave zombies around
613 return (out, err), proc
615 def _retry_rexec(args,
619 tmp_known_hosts = None,
622 for x in xrange(retry):
623 # connects to the remote host and starts a remote connection
624 proc = subprocess.Popen(args,
626 stdout = subprocess.PIPE,
627 stdin = subprocess.PIPE,
628 stderr = subprocess.PIPE)
630 # attach tempfile object to the process, to make sure the file stays
631 # alive until the process is finished with it
632 proc._known_hosts = tmp_known_hosts
634 # The argument block == False forces to rexec to return immediately,
638 (out, err) = proc.communicate()
640 err = proc.stderr.read()
641 out = proc.stdout.read()
643 log(log_msg, logging.DEBUG, out, err)
648 if err.strip().startswith('ssh: ') or err.strip().startswith('mux_client_hello_exchange: '):
649 # SSH error, can safely retry
652 # Probably timed out or plain failed but can retry
657 msg = "SLEEPING %d ... ATEMPT %d - command %s " % (
658 t, x, " ".join(args))
659 log(msg, logging.DEBUG)
664 except RuntimeError, e:
665 msg = " rexec EXCEPTION - TIMEOUT -> %s \n %s" % ( e.args, log_msg )
666 log(msg, logging.DEBUG, out, err)
672 return ((out, err), proc)