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 # Lucia Guevgeozian <lucia.guevgeozian_odizzio@inria.fr>
21 from nepi.execution.attribute import Attribute, Flags, Types
22 from nepi.execution.resource import ResourceManager, clsinit_copy, \
23 ResourceState, reschedule_delay
24 from nepi.resources.linux.node import LinuxNode
25 from nepi.resources.planetlab.plcapi import PLCAPIFactory
26 from nepi.util.execfuncs import lexec
27 from nepi.util import sshfuncs
29 from random import randint
36 class PlanetlabNode(LinuxNode):
37 _rtype = "PlanetlabNode"
38 _help = "Controls a PlanetLab host accessible using a SSH key " \
39 "associated to a PlanetLab user account"
40 _backend = "planetlab"
42 lock = threading.Lock()
45 def _register_attributes(cls):
46 ip = Attribute("ip", "PlanetLab host public IP address",
49 pl_url = Attribute("plcApiUrl", "URL of PlanetLab PLCAPI host \
50 (e.g. www.planet-lab.eu or www.planet-lab.org) ",
51 default = "www.planet-lab.eu",
52 flags = Flags.Credential)
54 pl_ptn = Attribute("plcApiPattern", "PLC API service regexp pattern \
55 (e.g. https://%(hostname)s:443/PLCAPI/ ) ",
56 default = "https://%(hostname)s:443/PLCAPI/",
59 pl_user = Attribute("pluser", "PlanetLab account user, as the one to \
60 authenticate in the website) ",
61 flags = Flags.Credential)
63 pl_password = Attribute("plpassword",
64 "PlanetLab account password, as \
65 the one to authenticate in the website) ",
66 flags = Flags.Credential)
68 city = Attribute("city", "Constrain location (city) during resource \
69 discovery. May use wildcards.",
72 country = Attribute("country", "Constrain location (country) during \
73 resource discovery. May use wildcards.",
76 region = Attribute("region", "Constrain location (region) during \
77 resource discovery. May use wildcards.",
80 architecture = Attribute("architecture", "Constrain architecture \
81 during resource discovery.",
82 type = Types.Enumerate,
87 operating_system = Attribute("operatingSystem", "Constrain operating \
88 system during resource discovery.",
89 type = Types.Enumerate,
97 #site = Attribute("site", "Constrain the PlanetLab site this node \
99 # type = Types.Enumerate,
103 # flags = Flags.Filter)
105 min_reliability = Attribute("minReliability", "Constrain reliability \
106 while picking PlanetLab nodes. Specifies a lower \
110 flags = Flags.Filter)
112 max_reliability = Attribute("maxReliability", "Constrain reliability \
113 while picking PlanetLab nodes. Specifies an upper \
117 flags = Flags.Filter)
119 min_bandwidth = Attribute("minBandwidth", "Constrain available \
120 bandwidth while picking PlanetLab nodes. \
121 Specifies a lower acceptable bound.",
124 flags = Flags.Filter)
126 max_bandwidth = Attribute("maxBandwidth", "Constrain available \
127 bandwidth while picking PlanetLab nodes. \
128 Specifies an upper acceptable bound.",
131 flags = Flags.Filter)
133 min_load = Attribute("minLoad", "Constrain node load average while \
134 picking PlanetLab nodes. Specifies a lower acceptable \
138 flags = Flags.Filter)
140 max_load = Attribute("maxLoad", "Constrain node load average while \
141 picking PlanetLab nodes. Specifies an upper acceptable \
145 flags = Flags.Filter)
147 min_cpu = Attribute("minCpu", "Constrain available cpu time while \
148 picking PlanetLab nodes. Specifies a lower acceptable \
152 flags = Flags.Filter)
154 max_cpu = Attribute("maxCpu", "Constrain available cpu time while \
155 picking PlanetLab nodes. Specifies an upper acceptable \
159 flags = Flags.Filter)
161 timeframe = Attribute("timeframe", "Past time period in which to check\
162 information about the node. Values are year,month, \
165 type = Types.Enumerate,
170 flags = Flags.Filter)
172 # plblacklist = Attribute("blacklist", "Take into account the file plblacklist \
173 # in the user's home directory under .nepi directory. This file \
174 # contains a list of PL nodes to blacklist, and at the end \
175 # of the experiment execution the new blacklisted nodes are added.",
178 # flags = Flags.ReadOnly)
181 cls._register_attribute(ip)
182 cls._register_attribute(pl_url)
183 cls._register_attribute(pl_ptn)
184 cls._register_attribute(pl_user)
185 cls._register_attribute(pl_password)
186 #cls._register_attribute(site)
187 cls._register_attribute(city)
188 cls._register_attribute(country)
189 cls._register_attribute(region)
190 cls._register_attribute(architecture)
191 cls._register_attribute(operating_system)
192 cls._register_attribute(min_reliability)
193 cls._register_attribute(max_reliability)
194 cls._register_attribute(min_bandwidth)
195 cls._register_attribute(max_bandwidth)
196 cls._register_attribute(min_load)
197 cls._register_attribute(max_load)
198 cls._register_attribute(min_cpu)
199 cls._register_attribute(max_cpu)
200 cls._register_attribute(timeframe)
202 def __init__(self, ec, guid):
203 super(PlanetlabNode, self).__init__(ec, guid)
206 self._node_to_provision = None
207 self._slicenode = False
208 self._hostname = False
210 if self.get("gateway") or self.get("gatewayUser"):
211 self.set("gateway", None)
212 self.set("gatewayUser", None)
214 def _skip_provision(self):
215 pl_user = self.get("pluser")
216 pl_pass = self.get("plpassword")
217 if not pl_user and not pl_pass:
224 pl_user = self.get("pluser")
225 pl_pass = self.get("plpassword")
226 pl_url = self.get("plcApiUrl")
227 pl_ptn = self.get("plcApiPattern")
229 self._plapi = PLCAPIFactory.get_api(pl_user, pl_pass, pl_url,
237 def do_discover(self):
239 Based on the attributes defined by the user, discover the suitable
242 if self._skip_provision():
243 super(PlanetlabNode, self).do_discover()
246 hostname = self._get_hostname()
248 # the user specified one particular node to be provisioned
249 self._hostname = True
250 node_id = self._get_nodes_id({'hostname':hostname})
251 node_id = node_id.pop()['node_id']
253 # check that the node is not blacklisted or being provisioned
255 with PlanetlabNode.lock:
256 plist = self.plapi.reserved()
257 blist = self.plapi.blacklisted()
258 if node_id not in blist and node_id not in plist:
260 # check that is really alive, by performing ping
261 ping_ok = self._do_ping(node_id)
263 self._blacklist_node(node_id)
264 self.fail_node_not_alive(hostname)
266 if self._check_if_in_slice([node_id]):
267 self._slicenode = True
268 self._put_node_in_provision(node_id)
269 self._node_to_provision = node_id
271 self.fail_node_not_available(hostname)
272 super(PlanetlabNode, self).do_discover()
275 # the user specifies constraints based on attributes, zero, one or
276 # more nodes can match these constraints
277 nodes = self._filter_based_on_attributes()
279 # nodes that are already part of user's slice have the priority to
281 nodes_inslice = self._check_if_in_slice(nodes)
282 nodes_not_inslice = list(set(nodes) - set(nodes_inslice))
286 node_id = self._choose_random_node(nodes_inslice)
287 self._slicenode = True
290 # Either there were no matching nodes in the user's slice, or
291 # the nodes in the slice were blacklisted or being provisioned
292 # by other RM. Note nodes_not_inslice is never empty
293 node_id = self._choose_random_node(nodes_not_inslice)
294 self._slicenode = False
297 self._node_to_provision = node_id
299 self._set_hostname_attr(node_id)
300 self.info(" Selected node to provision ")
301 super(PlanetlabNode, self).do_discover()
303 with PlanetlabNode.lock:
304 self._blacklist_node(node_id)
307 self.fail_not_enough_nodes()
309 def do_provision(self):
311 Add node to user's slice after verifing that the node is functioning
314 if self._skip_provision():
315 super(PlanetlabNode, self).do_provision()
323 while not provision_ok:
324 node = self._node_to_provision
325 if not self._slicenode:
326 self._add_node_to_slice(node)
328 # check ssh connection
330 while t < timeout and not ssh_ok:
332 cmd = 'echo \'GOOD NODE\''
333 ((out, err), proc) = self.execute(cmd)
334 if out.find("GOOD NODE") < 0:
342 cmd = 'echo \'GOOD NODE\''
343 ((out, err), proc) = self.execute(cmd)
344 if not out.find("GOOD NODE") < 0:
348 # the timeout was reach without establishing ssh connection
349 # the node is blacklisted, deleted from the slice, and a new
350 # node to provision is discovered
351 with PlanetlabNode.lock:
352 self.warning(" Could not SSH login ")
353 self._blacklist_node(node)
354 #self._delete_node_from_slice(node)
358 # check /proc directory is mounted (ssh_ok = True)
359 # and file system is not read only
361 cmd = 'mount |grep proc'
362 ((out1, err1), proc1) = self.execute(cmd)
363 cmd = 'touch /tmp/tmpfile; rm /tmp/tmpfile'
364 ((out2, err2), proc2) = self.execute(cmd)
365 if out1.find("/proc type proc") < 0 or \
366 "Read-only file system".lower() in err2.lower():
367 with PlanetlabNode.lock:
368 self.warning(" Corrupted file system ")
369 self._blacklist_node(node)
370 #self._delete_node_from_slice(node)
376 if not self.get('hostname'):
377 self._set_hostname_attr(node)
379 ip = self._get_ip(node)
381 self.info(" Node provisioned ")
383 super(PlanetlabNode, self).do_provision()
385 def _filter_based_on_attributes(self):
387 Retrive the list of nodes ids that match user's constraints
389 # Map user's defined attributes with tagnames of PlanetLab
390 timeframe = self.get("timeframe")[0]
393 'country' : 'country',
395 'architecture' : 'arch',
396 'operatingSystem' : 'fcdistro',
397 #'site' : 'pldistro',
398 'minReliability' : 'reliability%s' % timeframe,
399 'maxReliability' : 'reliability%s' % timeframe,
400 'minBandwidth' : 'bw%s' % timeframe,
401 'maxBandwidth' : 'bw%s' % timeframe,
402 'minLoad' : 'load%s' % timeframe,
403 'maxLoad' : 'load%s' % timeframe,
404 'minCpu' : 'cpu%s' % timeframe,
405 'maxCpu' : 'cpu%s' % timeframe,
411 for attr_name, attr_obj in self._attrs.iteritems():
412 attr_value = self.get(attr_name)
414 if attr_value is not None and attr_obj.has_flag(Flags.Filter) and \
415 attr_name != 'timeframe':
417 attr_tag = attr_to_tags[attr_name]
418 filters['tagname'] = attr_tag
420 # filter nodes by fixed constraints e.g. operating system
421 if not 'min' in attr_name and not 'max' in attr_name:
422 filters['value'] = attr_value
423 nodes_id = self._filter_by_fixed_attr(filters, nodes_id)
425 # filter nodes by range constraints e.g. max bandwidth
426 elif ('min' or 'max') in attr_name:
427 nodes_id = self._filter_by_range_attr(attr_name, attr_value, filters, nodes_id)
430 nodes = self._get_nodes_id()
432 nodes_id.append(node['node_id'])
435 def _filter_by_fixed_attr(self, filters, nodes_id):
437 Query PLCAPI for nodes ids matching fixed attributes defined by the
440 node_tags = self.plapi.get_node_tags(filters)
441 if node_tags is not None:
443 if len(nodes_id) == 0:
444 # first attribute being matched
445 for node_tag in node_tags:
446 nodes_id.append(node_tag['node_id'])
448 # remove the nodes ids that don't match the new attribute
449 # that is being match
452 for node_tag in node_tags:
453 if node_tag['node_id'] in nodes_id:
454 nodes_id_tmp.append(node_tag['node_id'])
456 if len(nodes_id_tmp):
457 nodes_id = set(nodes_id) & set(nodes_id_tmp)
459 # no node from before match the new constraint
460 self.fail_discovery()
462 # no nodes match the filter applied
463 self.fail_discovery()
467 def _filter_by_range_attr(self, attr_name, attr_value, filters, nodes_id):
469 Query PLCAPI for nodes ids matching attributes defined in a certain
472 node_tags = self.plapi.get_node_tags(filters)
475 if len(nodes_id) == 0:
476 # first attribute being matched
477 for node_tag in node_tags:
479 # check that matches the min or max restriction
480 if 'min' in attr_name and node_tag['value'] != 'n/a' and \
481 float(node_tag['value']) > attr_value:
482 nodes_id.append(node_tag['node_id'])
484 elif 'max' in attr_name and node_tag['value'] != 'n/a' and \
485 float(node_tag['value']) < attr_value:
486 nodes_id.append(node_tag['node_id'])
489 # remove the nodes ids that don't match the new attribute
490 # that is being match
492 for node_tag in node_tags:
494 # check that matches the min or max restriction and was a
495 # matching previous filters
496 if 'min' in attr_name and node_tag['value'] != 'n/a' and \
497 float(node_tag['value']) > attr_value and \
498 node_tag['node_id'] in nodes_id:
499 nodes_id_tmp.append(node_tag['node_id'])
501 elif 'max' in attr_name and node_tag['value'] != 'n/a' and \
502 float(node_tag['value']) < attr_value and \
503 node_tag['node_id'] in nodes_id:
504 nodes_id_tmp.append(node_tag['node_id'])
506 if len(nodes_id_tmp):
507 nodes_id = set(nodes_id) & set(nodes_id_tmp)
509 # no node from before match the new constraint
510 self.fail_discovery()
513 # no nodes match the filter applied
514 self.fail_discovery()
518 def _choose_random_node(self, nodes):
520 From the possible nodes for provision, choose randomly to decrese the
521 probability of different RMs choosing the same node for provision
526 index = randint(0, size)
527 node_id = nodes[index]
528 nodes[index] = nodes[size]
530 # check the node is not blacklisted or being provision by other RM
531 # and perform ping to check that is really alive
532 with PlanetlabNode.lock:
534 blist = self.plapi.blacklisted()
535 plist = self.plapi.reserved()
536 if node_id not in blist and node_id not in plist:
537 ping_ok = self._do_ping(node_id)
539 self._set_hostname_attr(node_id)
540 self.warning(" Node not responding PING ")
541 self._blacklist_node(node_id)
543 # discovered node for provision, added to provision list
544 self._put_node_in_provision(node_id)
547 def _get_nodes_id(self, filters=None):
548 return self.plapi.get_nodes(filters, fields=['node_id'])
550 def _add_node_to_slice(self, node_id):
551 self.info(" Adding node to slice ")
552 slicename = self.get("username")
553 with PlanetlabNode.lock:
554 slice_nodes = self.plapi.get_slice_nodes(slicename)
555 slice_nodes.append(node_id)
556 self.plapi.add_slice_nodes(slicename, slice_nodes)
558 def _delete_node_from_slice(self, node):
559 self.warning(" Deleting node from slice ")
560 slicename = self.get("username")
561 self.plapi.delete_slice_node(slicename, [node])
563 def _get_hostname(self):
564 hostname = self.get("hostname")
569 hostname = socket.gethostbyaddr(ip)[0]
570 self.set('hostname', hostname)
575 def _set_hostname_attr(self, node):
577 Query PLCAPI for the hostname of a certain node id and sets the
578 attribute hostname, it will over write the previous value
580 hostname = self.plapi.get_nodes(node, ['hostname'])
581 self.set("hostname", hostname[0]['hostname'])
583 def _check_if_in_slice(self, nodes_id):
585 Query PLCAPI to find out if any node id from nodes_id is in the user's
588 slicename = self.get("username")
589 slice_nodes = self.plapi.get_slice_nodes(slicename)
590 nodes_inslice = list(set(nodes_id) & set(slice_nodes))
593 def _do_ping(self, node_id):
595 Perform ping command on node's IP matching node id
598 ip = self._get_ip(node_id)
599 if not ip: return ping_ok
601 command = "ping -c4 %s" % ip
603 (out, err) = lexec(command)
604 if not str(out).find("2 received") < 0 or not str(out).find("3 received") < 0 or not \
605 str(out).find("4 received") < 0:
610 def _blacklist_node(self, node):
612 Add node mal functioning node to blacklist
614 self.warning(" Blacklisting malfunctioning node ")
615 self.plapi.blacklist_host(node)
616 if not self._hostname:
617 self.set('hostname', None)
619 def _put_node_in_provision(self, node):
621 Add node to the list of nodes being provisioned, in order for other RMs
622 to not try to provision the same one again
624 self.plapi.reserve_host(node)
626 def _get_ip(self, node_id):
628 Query PLCAPI for the IP of a node with certain node id
630 hostname = self.get("hostname") or \
631 self.plapi.get_nodes(node_id, ['hostname'])[0]['hostname']
633 ip = sshfuncs.gethostbyname(hostname)
635 # Fail while trying to find the IP
639 def fail_discovery(self):
640 msg = "Discovery failed. No candidates found for node"
642 raise RuntimeError, msg
644 def fail_node_not_alive(self, hostname=None):
645 msg = "Node %s not alive" % hostname
646 raise RuntimeError, msg
648 def fail_node_not_available(self, hostname):
649 msg = "Node %s not available for provisioning" % hostname
650 raise RuntimeError, msg
652 def fail_not_enough_nodes(self):
653 msg = "Not enough nodes available for provisioning"
654 raise RuntimeError, msg
656 def fail_plapi(self):
657 msg = "Failing while trying to instanciate the PLC API.\nSet the" + \
658 " attributes pluser and plpassword."
659 raise RuntimeError, msg
661 def valid_connection(self, guid):