src/nepi/resources/linux/gretunnel.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 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 LinuxGRETunnel(LinuxApplication):
  33     _rtype = "LinuxGRETunnel"
  34     _help = "Constructs a tunnel between two Linux endpoints using a UDP connection "
  35     _backend = "linux"
  36
  37     @classmethod
  38     def _register_attributes(cls):
  39         bwlimit = Attribute("bwLimit",
  40                 "Specifies the interface's emulated bandwidth in bytes "
  41                 "per second.",
  42                 type = Types.Integer,
  43                 flags = Flags.Design)
  44
  45         cls._register_attribute(bwlimit)
  46
  47     def __init__(self, ec, guid):
  48         super(LinuxGRETunnel, self).__init__(ec, guid)
  49
  50     def log_message(self, msg):
  51         return " guid %d - GRE tunnel %s - %s - %s " % (self.guid,
  52                 self.endpoint1.node.get("hostname"),
  53                 self.endpoint2.node.get("hostname"),
  54                 msg)
  55
  56     def get_endpoints(self):
  57         """ Returns the list of RM that are endpoints to the tunnel
  58         """
  59         connected = []
  60         for guid in self.connections:
  61             rm = self.ec.get_resource(guid)
  62             if hasattr(rm, "udp_connect_command"):
  63                 connected.append(rm)
  64         return connected
  65
  66     @property
  67     def endpoint1(self):
  68         endpoints = self.get_endpoints()
  69         if endpoints: return endpoints[0]
  70         return None
  71
  72     @property
  73     def endpoint2(self):
  74         endpoints = self.get_endpoints()
  75         if endpoints and len(endpoints) > 1: return endpoints[1]
  76         return None
  77
  78     def app_home(self, endpoint):
  79         return os.path.join(endpoint.node.exp_home, self._home)
  80
  81     def run_home(self, endpoint):
  82         return os.path.join(self.app_home(endpoint), self.ec.run_id)
  83
  84     def udp_connect(self, endpoint, remote_ip):
  85         # Get udp connect command
  86         local_port_file = os.path.join(self.run_home(endpoint),
  87                 "local_port")
  88         remote_port_file = os.path.join(self.run_home(endpoint),
  89                 "remote_port")
  90         ret_file = os.path.join(self.run_home(endpoint),
  91                 "ret_file")
  92         cipher = self.get("cipher")
  93         cipher_key = self.get("cipherKey")
  94         bwlimit = self.get("bwLimit")
  95         txqueuelen = self.get("txQueueLen")
  96         udp_connect_command = endpoint.udp_connect_command(
  97                 remote_ip, local_port_file, remote_port_file,
  98                 ret_file, cipher, cipher_key, bwlimit, txqueuelen)
  99
 100         # upload command to connect.sh script
 101         shfile = os.path.join(self.app_home(endpoint), "udp-connect.sh")
 102         endpoint.node.upload(udp_connect_command,
 103                 shfile,
 104                 text = True,
 105                 overwrite = False)
 106
 107         # invoke connect script
 108         cmd = "bash %s" % shfile
 109         (out, err), proc = endpoint.node.run(cmd, self.run_home(endpoint))
 110
 111         # check if execution errors occurred
 112         msg = " Failed to connect endpoints "
 113
 114         if proc.poll():
 115             self.error(msg, out, err)
 116             raise RuntimeError, msg
 117
 118         # Wait for pid file to be generated
 119         pid, ppid = endpoint.node.wait_pid(self.run_home(endpoint))
 120
 121         # If the process is not running, check for error information
 122         # on the remote machine
 123         if not pid or not ppid:
 124             (out, err), proc = endpoint.node.check_errors(self.run_home(endpoint))
 125             # Out is what was written in the stderr file
 126             if err:
 127                 msg = " Failed to start command '%s' " % command
 128                 self.error(msg, out, err)
 129                 raise RuntimeError, msg
 130
 131         # wait until port is written to file
 132         port = self.wait_local_port(endpoint)
 133         return (port, pid, ppid)
 134
 135     def do_provision(self):
 136         # create run dir for tunnel on each node
 137         self.endpoint1.node.mkdir(self.run_home(self.endpoint1))
 138         self.endpoint2.node.mkdir(self.run_home(self.endpoint2))
 139
 140         # Invoke connect script in endpoint 1
 141         remote_ip1 = socket.gethostbyname(self.endpoint2.node.get("hostname"))
 142         (port1, self._pid1, self._ppid1) = self.udp_connect(self.endpoint1,
 143                 remote_ip1)
 144
 145         # Invoke connect script in endpoint 2
 146         remote_ip2 = socket.gethostbyname(self.endpoint1.node.get("hostname"))
 147         (port2, self._pid2, self._ppid2) = self.udp_connect(self.endpoint2,
 148                 remote_ip2)
 149
 150         # upload file with port 2 to endpoint 1
 151         self.upload_remote_port(self.endpoint1, port2)
 152
 153         # upload file with port 1 to endpoint 2
 154         self.upload_remote_port(self.endpoint2, port1)
 155
 156         # check if connection was successful on both sides
 157         self.wait_result(self.endpoint1)
 158         self.wait_result(self.endpoint2)
 159
 160         self.info("Provisioning finished")
 161
 162         self.set_provisioned()
 163
 164     def do_deploy(self):
 165         if (not self.endpoint1 or self.endpoint1.state < ResourceState.READY) or \
 166             (not self.endpoint2 or self.endpoint2.state < ResourceState.READY):
 167             self.ec.schedule(reschedule_delay, self.deploy)
 168         else:
 169             self.do_discover()
 170             self.do_provision()
 171
 172             self.set_ready()
 173
 174     def do_start(self):
 175         if self.state == ResourceState.READY:
 176             command = self.get("command")
 177             self.info("Starting command '%s'" % command)
 178
 179             self.set_started()
 180         else:
 181             msg = " Failed to execute command '%s'" % command
 182             self.error(msg, out, err)
 183             raise RuntimeError, msg
 184
 185     def do_stop(self):
 186         """ Stops application execution
 187         """
 188         if self.state == ResourceState.STARTED:
 189             self.info("Stopping tunnel")
 190
 191             # Only try to kill the process if the pid and ppid
 192             # were retrieved
 193             if self._pid1 and self._ppid1 and self._pid2 and self._ppid2:
 194                 (out1, err1), proc1 = self.endpoint1.node.kill(self._pid1,
 195                         self._ppid1, sudo = True)
 196                 (out2, err2), proc2 = self.endpoint2.node.kill(self._pid2,
 197                         self._ppid2, sudo = True)
 198
 199                 if (proc1.poll() and err1) or (proc2.poll() and err2):
 200                     # check if execution errors occurred
 201                     msg = " Failed to STOP tunnel"
 202                     self.error(msg, err1, err2)
 203                     raise RuntimeError, msg
 204
 205             self.set_stopped()
 206
 207     @property
 208     def state(self):
 209         """ Returns the state of the application
 210         """
 211         if self._state == ResourceState.STARTED:
 212             # In order to avoid overwhelming the remote host and
 213             # the local processor with too many ssh queries, the state is only
 214             # requested every 'state_check_delay' seconds.
 215             state_check_delay = 0.5
 216             if tdiffsec(tnow(), self._last_state_check) > state_check_delay:
 217                 if self._pid1 and self._ppid1 and self._pid2 and self._ppid2:
 218                     # Make sure the process is still running in background
 219                     # No execution errors occurred. Make sure the background
 220                     # process with the recorded pid is still running.
 221                     status1 = self.endpoint1.node.status(self._pid1, self._ppid1)
 222                     status2 = self.endpoint2.node.status(self._pid2, self._ppid2)
 223
 224                     if status1 == ProcStatus.FINISHED and \
 225                             status2 == ProcStatus.FINISHED:
 226
 227                         # check if execution errors occurred
 228                         (out1, err1), proc1 = self.endpoint1.node.check_errors(
 229                                 self.run_home(self.endpoint1))
 230
 231                         (out2, err2), proc2 = self.endpoint2.node.check_errors(
 232                                 self.run_home(self.endpoint2))
 233
 234                         if err1 or err2:
 235                             msg = "Error occurred in tunnel"
 236                             self.error(msg, err1, err2)
 237                             self.fail()
 238                         else:
 239                             self.set_stopped()
 240
 241                 self._last_state_check = tnow()
 242
 243         return self._state
 244
 245     def wait_local_port(self, endpoint):
 246         """ Waits until the local_port file for the endpoint is generated,
 247         and returns the port number
 248
 249         """
 250         return self.wait_file(endpoint, "local_port")
 251
 252     def wait_result(self, endpoint):
 253         """ Waits until the return code file for the endpoint is generated
 254
 255         """
 256         return self.wait_file(endpoint, "ret_file")
 257
 258     def wait_file(self, endpoint, filename):
 259         """ Waits until file on endpoint is generated """
 260         result = None
 261         delay = 1.0
 262
 263         for i in xrange(20):
 264             (out, err), proc = endpoint.node.check_output(
 265                     self.run_home(endpoint), filename)
 266
 267             if out:
 268                 result = out.strip()
 269                 break
 270             else:
 271                 time.sleep(delay)
 272                 delay = delay * 1.5
 273         else:
 274             msg = "Couldn't retrieve %s" % filename
 275             self.error(msg, out, err)
 276             raise RuntimeError, msg
 277
 278         return result
 279
 280     def upload_remote_port(self, endpoint, port):
 281         # upload remote port number to file
 282         port = "%s\n" % port
 283         endpoint.node.upload(port,
 284                 os.path.join(self.run_home(endpoint), "remote_port"),
 285                 text = True,
 286                 overwrite = False)
 287
 288     def valid_connection(self, guid):
 289         # TODO: Validate!
 290         return True
 291