src/nepi/resources/linux/udptunnel.py

   1 #
   2 #    NEPI, a framework to manage network experiments
   3 #    Copyright (C) 2013 INRIA
   4 #
   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.
   9 #
  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.
  14 #
  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/>.
  17 #
  18 # Author: Alina Quereilhac <alina.quereilhac@inria.fr>
  19
  20 from nepi.execution.attribute import Attribute, Flags, Types
  21 from nepi.execution.resource import ResourceManager, clsinit_copy, ResourceState, \
  22         reschedule_delay
  23 from nepi.resources.linux.application import LinuxApplication
  24 from nepi.util.sshfuncs import ProcStatus
  25 from nepi.util.timefuncs import tnow, tdiffsec
  26
  27 import os
  28 import socket
  29 import time
  30
  31 @clsinit_copy
  32 class UdpTunnel(LinuxApplication):
  33     _rtype = "UdpTunnel"
  34
  35     @classmethod
  36     def _register_attributes(cls):
  37         cipher = Attribute("cipher",
  38                "Cipher to encript communication. "
  39                 "One of PLAIN, AES, Blowfish, DES, DES3. ",
  40                 default = None,
  41                 allowed = ["PLAIN", "AES", "Blowfish", "DES", "DES3"],
  42                 type = Types.Enumerate,
  43                 flags = Flags.ExecReadOnly)
  44
  45         cipher_key = Attribute("cipherKey",
  46                 "Specify a symmetric encryption key with which to protect "
  47                 "packets across the tunnel. python-crypto must be installed "
  48                 "on the system." ,
  49                 flags = Flags.ExecReadOnly)
  50
  51         txqueuelen = Attribute("txQueueLen",
  52                 "Specifies the interface's transmission queue length. "
  53                 "Defaults to 1000. ",
  54                 type = Types.Integer,
  55                 flags = Flags.ExecReadOnly)
  56
  57         bwlimit = Attribute("bwLimit",
  58                 "Specifies the interface's emulated bandwidth in bytes "
  59                 "per second.",
  60                 type = Types.Integer,
  61                 flags = Flags.ExecReadOnly)
  62
  63         cls._register_attribute(cipher)
  64         cls._register_attribute(cipher_key)
  65         cls._register_attribute(txqueuelen)
  66         cls._register_attribute(bwlimit)
  67
  68     def __init__(self, ec, guid):
  69         super(UdpTunnel, self).__init__(ec, guid)
  70         self._home = "udp-tunnel-%s" % self.guid
  71         self._pid1 = None
  72         self._ppid1 = None
  73         self._pid2 = None
  74         self._ppid2 = None
  75
  76     def log_message(self, msg):
  77         return " guid %d - tunnel %s - %s - %s " % (self.guid,
  78                 self.endpoint1.node.get("hostname"),
  79                 self.endpoint2.node.get("hostname"),
  80                 msg)
  81
  82     def get_endpoints(self):
  83         """ Returns the list of RM that are endpoints to the tunnel
  84         """
  85         connected = []
  86         for guid in self.connections:
  87             rm = self.ec.get_resource(guid)
  88             if hasattr(rm, "udp_connect_command"):
  89                 connected.append(rm)
  90         return connected
  91
  92     @property
  93     def endpoint1(self):
  94         endpoints = self.get_endpoints()
  95         if endpoints: return endpoints[0]
  96         return None
  97
  98     @property
  99     def endpoint2(self):
 100         endpoints = self.get_endpoints()
 101         if endpoints and len(endpoints) > 1: return endpoints[1]
 102         return None
 103
 104     def app_home(self, endpoint):
 105         return os.path.join(endpoint.node.exp_home, self._home)
 106
 107     def run_home(self, endpoint):
 108         return os.path.join(self.app_home(endpoint), self.ec.run_id)
 109
 110     def udp_connect(self, endpoint, remote_ip):
 111         # Get udp connect command
 112         local_port_file = os.path.join(self.run_home(endpoint),
 113                 "local_port")
 114         remote_port_file = os.path.join(self.run_home(endpoint),
 115                 "remote_port")
 116         ret_file = os.path.join(self.run_home(endpoint),
 117                 "ret_file")
 118         cipher = self.get("cipher")
 119         cipher_key = self.get("cipherKey")
 120         bwlimit = self.get("bwLimit")
 121         txqueuelen = self.get("txQueueLen")
 122         udp_connect_command = endpoint.udp_connect_command(
 123                 remote_ip, local_port_file, remote_port_file,
 124                 ret_file, cipher, cipher_key, bwlimit, txqueuelen)
 125
 126         # upload command to connect.sh script
 127         shfile = os.path.join(self.app_home(endpoint), "udp-connect.sh")
 128         endpoint.node.upload(udp_connect_command,
 129                 shfile,
 130                 text = True,
 131                 overwrite = False)
 132
 133         # invoke connect script
 134         cmd = "bash %s" % shfile
 135         (out, err), proc = endpoint.node.run(cmd, self.run_home(endpoint))
 136
 137         # check if execution errors occurred
 138         msg = " Failed to connect endpoints "
 139
 140         if proc.poll():
 141             self.fail()
 142             self.error(msg, out, err)
 143             raise RuntimeError, msg
 144
 145         # Wait for pid file to be generated
 146         pid, ppid = endpoint.node.wait_pid(self.run_home(endpoint))
 147
 148         # If the process is not running, check for error information
 149         # on the remote machine
 150         if not pid or not ppid:
 151             (out, err), proc = endpoint.node.check_errors(self.run_home(endpoint))
 152             # Out is what was written in the stderr file
 153             if err:
 154                 self.fail()
 155                 msg = " Failed to start command '%s' " % command
 156                 self.error(msg, out, err)
 157                 raise RuntimeError, msg
 158
 159         # wait until port is written to file
 160         port = self.wait_local_port(endpoint)
 161         return (port, pid, ppid)
 162
 163     def provision(self):
 164         # create run dir for tunnel on each node
 165         self.endpoint1.node.mkdir(self.run_home(self.endpoint1))
 166         self.endpoint2.node.mkdir(self.run_home(self.endpoint2))
 167
 168         # Invoke connect script in endpoint 1
 169         remote_ip1 = socket.gethostbyname(self.endpoint2.node.get("hostname"))
 170         (port1, self._pid1, self._ppid1) = self.udp_connect(self.endpoint1,
 171                 remote_ip1)
 172
 173         # Invoke connect script in endpoint 2
 174         remote_ip2 = socket.gethostbyname(self.endpoint1.node.get("hostname"))
 175         (port2, self._pid2, self._ppid2) = self.udp_connect(self.endpoint2,
 176                 remote_ip2)
 177
 178         # upload file with port 2 to endpoint 1
 179         self.upload_remote_port(self.endpoint1, port2)
 180
 181         # upload file with port 1 to endpoint 2
 182         self.upload_remote_port(self.endpoint2, port1)
 183
 184         # check if connection was successful on both sides
 185         self.wait_result(self.endpoint1)
 186         self.wait_result(self.endpoint2)
 187
 188         self.info("Provisioning finished")
 189
 190         self.debug("----- READY ---- ")
 191         self._provision_time = tnow()
 192         self._state = ResourceState.PROVISIONED
 193
 194     def deploy(self):
 195         if (not self.endpoint1 or self.endpoint1.state < ResourceState.READY) or \
 196             (not self.endpoint2 or self.endpoint2.state < ResourceState.READY):
 197             self.ec.schedule(reschedule_delay, self.deploy)
 198         else:
 199             try:
 200                 self.discover()
 201                 self.provision()
 202             except:
 203                 self.fail()
 204                 raise
 205
 206             self.debug("----- READY ---- ")
 207             self._ready_time = tnow()
 208             self._state = ResourceState.READY
 209
 210     def start(self):
 211         if self._state == ResourceState.READY:
 212             command = self.get("command")
 213             self.info("Starting command '%s'" % command)
 214
 215             self._start_time = tnow()
 216             self._state = ResourceState.STARTED
 217         else:
 218             msg = " Failed to execute command '%s'" % command
 219             self.error(msg, out, err)
 220             self._state = ResourceState.FAILED
 221             raise RuntimeError, msg
 222
 223     # XXX: Leaves process unkilled!!
 224     #       Implement another mechanism to kill the tunnel!
 225     def stop(self):
 226         """ Stops application execution
 227         """
 228         if self.state == ResourceState.STARTED:
 229             stopped = True
 230             self.info("Stopping tunnel")
 231
 232             # Only try to kill the process if the pid and ppid
 233             # were retrieved
 234             if self._pid1 and self._ppid1 and self._pid2 and self._ppid2:
 235                 (out1, err1), proc1 = self.endpoint1.node.kill(self._pid1,
 236                         self._ppid1, sudo = True)
 237                 (out2, err2), proc2 = self.endpoint2.node.kill(self._pid2,
 238                         self._ppid2, sudo = True)
 239
 240                 if err1 or err2 or proc1.poll() or proc2.poll():
 241                     # check if execution errors occurred
 242                     msg = " Failed to STOP tunnel"
 243                     self.error(msg, err1, err2)
 244                     self.fail()
 245                     stopped = False
 246
 247             if stopped:
 248                 self._stop_time = tnow()
 249                 self._state = ResourceState.STOPPED
 250
 251     @property
 252     def state(self):
 253         """ Returns the state of the application
 254         """
 255         if self._state == ResourceState.STARTED:
 256             # In order to avoid overwhelming the remote host and
 257             # the local processor with too many ssh queries, the state is only
 258             # requested every 'state_check_delay' seconds.
 259             state_check_delay = 0.5
 260             if tdiffsec(tnow(), self._last_state_check) > state_check_delay:
 261                 if self._pid1 and self._ppid1 and self._pid2 and self._ppid2:
 262                     # Make sure the process is still running in background
 263                     # No execution errors occurred. Make sure the background
 264                     # process with the recorded pid is still running.
 265                     status1 = self.endpoint1.node.status(self._pid1, self._ppid1)
 266                     status2 = self.endpoint2.node.status(self._pid2, self._ppid2)
 267
 268                     if status1 == ProcStatus.FINISHED and \
 269                             status2 == ProcStatus.FINISHED:
 270
 271                         # check if execution errors occurred
 272                         (out1, err1), proc1 = self.endpoint1.node.check_errors(
 273                                 self.run_home(self.endpoint1))
 274
 275                         (out2, err2), proc2 = self.endpoint2.node.check_errors(
 276                                 self.run_home(self.endpoint2))
 277
 278                         if err1 or err2:
 279                             msg = "Error occurred in tunnel"
 280                             self.error(msg, err1, err2)
 281                             self.fail()
 282                         else:
 283                             self._state = ResourceState.FINISHED
 284
 285                 self._last_state_check = tnow()
 286
 287         return self._state
 288
 289     def wait_local_port(self, endpoint):
 290         """ Waits until the local_port file for the endpoint is generated,
 291             and returns the port number """
 292         return self.wait_file(endpoint, "local_port")
 293
 294     def wait_result(self, endpoint):
 295         """ Waits until the return code file for the endpoint is generated """
 296         return self.wait_file(endpoint, "ret_file")
 297
 298     def wait_file(self, endpoint, filename):
 299         """ Waits until file on endpoint is generated """
 300         result = None
 301         delay = 1.0
 302
 303         for i in xrange(4):
 304             (out, err), proc = endpoint.node.check_output(
 305                     self.run_home(endpoint), filename)
 306
 307             if out:
 308                 result = out.strip()
 309                 break
 310             else:
 311                 time.sleep(delay)
 312                 delay = delay * 1.5
 313         else:
 314             msg = "Couldn't retrieve %s" % filename
 315             self.error(msg, out, err)
 316             self.fail()
 317             raise RuntimeError, msg
 318
 319         return result
 320
 321     def upload_remote_port(self, endpoint, port):
 322         # upload remote port number to file
 323         port = "%s\n" % port
 324         endpoint.node.upload(port,
 325                 os.path.join(self.run_home(endpoint), "remote_port"),
 326                 text = True,
 327                 overwrite = False)
 328
 329     def valid_connection(self, guid):
 330         # TODO: Validate!
 331         return True
 332