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",
47 flags = Flags.ReadOnly)
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/",
57 flags = Flags.ExecReadOnly)
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 cls._register_attribute(ip)
173 cls._register_attribute(pl_url)
174 cls._register_attribute(pl_ptn)
175 cls._register_attribute(pl_user)
176 cls._register_attribute(pl_password)
177 #cls._register_attribute(site)
178 cls._register_attribute(city)
179 cls._register_attribute(country)
180 cls._register_attribute(region)
181 cls._register_attribute(architecture)
182 cls._register_attribute(operating_system)
183 cls._register_attribute(min_reliability)
184 cls._register_attribute(max_reliability)
185 cls._register_attribute(min_bandwidth)
186 cls._register_attribute(max_bandwidth)
187 cls._register_attribute(min_load)
188 cls._register_attribute(max_load)
189 cls._register_attribute(min_cpu)
190 cls._register_attribute(max_cpu)
191 cls._register_attribute(timeframe)
193 def __init__(self, ec, guid):
194 super(PlanetlabNode, self).__init__(ec, guid)
197 self._node_to_provision = None
198 self._slicenode = False
199 self._hostname = False
201 def _skip_provision(self):
202 pl_user = self.get("pluser")
203 pl_pass = self.get("plpassword")
204 if not pl_user and not pl_pass:
211 pl_user = self.get("pluser")
212 pl_pass = self.get("plpassword")
213 pl_url = self.get("plcApiUrl")
214 pl_ptn = self.get("plcApiPattern")
216 self._plapi = PLCAPIFactory.get_api(pl_user, pl_pass, pl_url,
224 def do_discover(self):
226 Based on the attributes defined by the user, discover the suitable
229 if self._skip_provision():
230 super(PlanetlabNode, self).do_discover()
233 hostname = self._get_hostname()
235 # the user specified one particular node to be provisioned
236 # check with PLCAPI if it is alive
237 self._hostname = True
238 node_id = self._query_if_alive(hostname=hostname)
239 node_id = node_id.pop()
241 # check that the node is not blacklisted or being provisioned
243 with PlanetlabNode.lock:
244 plist = self.plapi.reserved()
245 blist = self.plapi.blacklisted()
246 if node_id not in blist and node_id not in plist:
248 # check that is really alive, by performing ping
249 ping_ok = self._do_ping(node_id)
251 self._blacklist_node(node_id)
252 self.fail_node_not_alive(hostname)
254 if self._check_if_in_slice([node_id]):
255 self._slicenode = True
256 self._put_node_in_provision(node_id)
257 self._node_to_provision = node_id
259 self.fail_node_not_available(hostname)
260 super(PlanetlabNode, self).do_discover()
263 # the user specifies constraints based on attributes, zero, one or
264 # more nodes can match these constraints
265 nodes = self._filter_based_on_attributes()
266 nodes_alive = self._query_if_alive(nodes)
268 # nodes that are already part of user's slice have the priority to
270 nodes_inslice = self._check_if_in_slice(nodes_alive)
271 nodes_not_inslice = list(set(nodes_alive) - set(nodes_inslice))
275 node_id = self._choose_random_node(nodes_inslice)
276 self._slicenode = True
279 # Either there were no matching nodes in the user's slice, or
280 # the nodes in the slice were blacklisted or being provisioned
281 # by other RM. Note nodes_not_inslice is never empty
282 node_id = self._choose_random_node(nodes_not_inslice)
283 self._slicenode = False
286 self._node_to_provision = node_id
288 self._set_hostname_attr(node_id)
289 self.info(" Selected node to provision ")
290 super(PlanetlabNode, self).do_discover()
292 with PlanetlabNode.lock:
293 self._blacklist_node(node_id)
296 self.fail_not_enough_nodes()
298 def do_provision(self):
300 Add node to user's slice after verifing that the node is functioning
303 if self._skip_provision():
304 super(PlanetlabNode, self).do_provision()
312 while not provision_ok:
313 node = self._node_to_provision
314 if not self._slicenode:
315 self._add_node_to_slice(node)
317 # check ssh connection
319 while t < timeout and not ssh_ok:
321 cmd = 'echo \'GOOD NODE\''
322 ((out, err), proc) = self.execute(cmd)
323 if out.find("GOOD NODE") < 0:
331 cmd = 'echo \'GOOD NODE\''
332 ((out, err), proc) = self.execute(cmd)
333 if not out.find("GOOD NODE") < 0:
337 # the timeout was reach without establishing ssh connection
338 # the node is blacklisted, deleted from the slice, and a new
339 # node to provision is discovered
340 with PlanetlabNode.lock:
341 self.warn(" Could not SSH login ")
342 self._blacklist_node(node)
343 #self._delete_node_from_slice(node)
344 #self.set('hostname', None)
348 # check /proc directory is mounted (ssh_ok = True)
350 cmd = 'mount |grep proc'
351 ((out, err), proc) = self.execute(cmd)
352 if out.find("/proc type proc") < 0:
353 with PlanetlabNode.lock:
354 self.warn(" Could not find directory /proc ")
355 self._blacklist_node(node)
356 #self._delete_node_from_slice(node)
357 #self.set('hostname', None)
363 if not self.get('hostname'):
364 self._set_hostname_attr(node)
366 ip = self._get_ip(node)
369 super(PlanetlabNode, self).do_provision()
371 def _filter_based_on_attributes(self):
373 Retrive the list of nodes ids that match user's constraints
375 # Map user's defined attributes with tagnames of PlanetLab
376 timeframe = self.get("timeframe")[0]
379 'country' : 'country',
381 'architecture' : 'arch',
382 'operatingSystem' : 'fcdistro',
383 #'site' : 'pldistro',
384 'minReliability' : 'reliability%s' % timeframe,
385 'maxReliability' : 'reliability%s' % timeframe,
386 'minBandwidth' : 'bw%s' % timeframe,
387 'maxBandwidth' : 'bw%s' % timeframe,
388 'minLoad' : 'load%s' % timeframe,
389 'maxLoad' : 'load%s' % timeframe,
390 'minCpu' : 'cpu%s' % timeframe,
391 'maxCpu' : 'cpu%s' % timeframe,
397 for attr_name, attr_obj in self._attrs.iteritems():
398 attr_value = self.get(attr_name)
400 if attr_value is not None and attr_obj.flags == 8 and \
401 attr_name != 'timeframe':
403 attr_tag = attr_to_tags[attr_name]
404 filters['tagname'] = attr_tag
406 # filter nodes by fixed constraints e.g. operating system
407 if not 'min' in attr_name and not 'max' in attr_name:
408 filters['value'] = attr_value
409 nodes_id = self._filter_by_fixed_attr(filters, nodes_id)
411 # filter nodes by range constraints e.g. max bandwidth
412 elif ('min' or 'max') in attr_name:
413 nodes_id = self._filter_by_range_attr(attr_name, attr_value, filters, nodes_id)
416 nodes = self.plapi.get_nodes()
418 nodes_id.append(node['node_id'])
423 def _filter_by_fixed_attr(self, filters, nodes_id):
425 Query PLCAPI for nodes ids matching fixed attributes defined by the
428 node_tags = self.plapi.get_node_tags(filters)
429 if node_tags is not None:
431 if len(nodes_id) == 0:
432 # first attribute being matched
433 for node_tag in node_tags:
434 nodes_id.append(node_tag['node_id'])
436 # remove the nodes ids that don't match the new attribute
437 # that is being match
440 for node_tag in node_tags:
441 if node_tag['node_id'] in nodes_id:
442 nodes_id_tmp.append(node_tag['node_id'])
444 if len(nodes_id_tmp):
445 nodes_id = set(nodes_id) & set(nodes_id_tmp)
447 # no node from before match the new constraint
448 self.fail_discovery()
450 # no nodes match the filter applied
451 self.fail_discovery()
455 def _filter_by_range_attr(self, attr_name, attr_value, filters, nodes_id):
457 Query PLCAPI for nodes ids matching attributes defined in a certain
460 node_tags = self.plapi.get_node_tags(filters)
463 if len(nodes_id) == 0:
464 # first attribute being matched
465 for node_tag in node_tags:
467 # check that matches the min or max restriction
468 if 'min' in attr_name and node_tag['value'] != 'n/a' and \
469 float(node_tag['value']) > attr_value:
470 nodes_id.append(node_tag['node_id'])
472 elif 'max' in attr_name and node_tag['value'] != 'n/a' and \
473 float(node_tag['value']) < attr_value:
474 nodes_id.append(node_tag['node_id'])
477 # remove the nodes ids that don't match the new attribute
478 # that is being match
480 for node_tag in node_tags:
482 # check that matches the min or max restriction and was a
483 # matching previous filters
484 if 'min' in attr_name and node_tag['value'] != 'n/a' and \
485 float(node_tag['value']) > attr_value and \
486 node_tag['node_id'] in nodes_id:
487 nodes_id_tmp.append(node_tag['node_id'])
489 elif 'max' in attr_name and node_tag['value'] != 'n/a' and \
490 float(node_tag['value']) < attr_value and \
491 node_tag['node_id'] in nodes_id:
492 nodes_id_tmp.append(node_tag['node_id'])
494 if len(nodes_id_tmp):
495 nodes_id = set(nodes_id) & set(nodes_id_tmp)
497 # no node from before match the new constraint
498 self.fail_discovery()
501 # no nodes match the filter applied
502 self.fail_discovery()
506 def _query_if_alive(self, nodes_id=None, hostname=None):
508 Query PLCAPI for nodes that register activity recently, using filters
509 related to the state of the node, e.g. last time it was contacted
511 if nodes_id is None and hostname is None:
512 msg = "Specify nodes_id or hostname"
513 raise RuntimeError, msg
515 if nodes_id is not None and hostname is not None:
516 msg = "Specify either nodes_id or hostname"
517 raise RuntimeError, msg
519 # define PL filters to check the node is alive
521 filters['run_level'] = 'boot'
522 filters['boot_state'] = 'boot'
523 filters['node_type'] = 'regular'
524 #filters['>last_contact'] = int(time.time()) - 2*3600
526 # adding node_id or hostname to the filters to check for the particular
529 filters['node_id'] = list(nodes_id)
530 alive_nodes_id = self._get_nodes_id(filters)
532 filters['hostname'] = hostname
533 alive_nodes_id = self._get_nodes_id(filters)
535 if len(alive_nodes_id) == 0:
536 self.fail_node_not_alive(hostname)
539 for node_id in alive_nodes_id:
540 nid = node_id['node_id']
545 def _choose_random_node(self, nodes):
547 From the possible nodes for provision, choose randomly to decrese the
548 probability of different RMs choosing the same node for provision
553 index = randint(0, size)
554 node_id = nodes[index]
555 nodes[index] = nodes[size]
557 # check the node is not blacklisted or being provision by other RM
558 # and perform ping to check that is really alive
559 with PlanetlabNode.lock:
561 blist = self.plapi.blacklisted()
562 plist = self.plapi.reserved()
563 if node_id not in blist and node_id not in plist:
564 ping_ok = self._do_ping(node_id)
566 self._set_hostname_attr(node_id)
567 self.warn(" Node not responding PING ")
568 self._blacklist_node(node_id)
569 #self.set('hostname', None)
571 # discovered node for provision, added to provision list
572 self._put_node_in_provision(node_id)
575 def _get_nodes_id(self, filters):
576 return self.plapi.get_nodes(filters, fields=['node_id'])
578 def _add_node_to_slice(self, node_id):
579 self.info(" Adding node to slice ")
580 slicename = self.get("username")
581 with PlanetlabNode.lock:
582 slice_nodes = self.plapi.get_slice_nodes(slicename)
583 slice_nodes.append(node_id)
584 self.plapi.add_slice_nodes(slicename, slice_nodes)
586 def _delete_node_from_slice(self, node):
587 self.warn(" Deleting node from slice ")
588 slicename = self.get("username")
589 self.plapi.delete_slice_node(slicename, [node])
591 def _get_hostname(self):
592 hostname = self.get("hostname")
597 hostname = socket.gethostbyaddr(ip)[0]
598 self.set('hostname', hostname)
603 def _set_hostname_attr(self, node):
605 Query PLCAPI for the hostname of a certain node id and sets the
606 attribute hostname, it will over write the previous value
608 hostname = self.plapi.get_nodes(node, ['hostname'])
609 self.set("hostname", hostname[0]['hostname'])
611 def _check_if_in_slice(self, nodes_id):
613 Query PLCAPI to find out if any node id from nodes_id is in the user's
616 slicename = self.get("username")
617 slice_nodes = self.plapi.get_slice_nodes(slicename)
618 nodes_inslice = list(set(nodes_id) & set(slice_nodes))
621 def _do_ping(self, node_id):
623 Perform ping command on node's IP matching node id
626 ip = self._get_ip(node_id)
627 if not ip: return ping_ok
629 command = "ping -c4 %s" % ip
631 (out, err) = lexec(command)
632 if not out.find("2 received") or not out.find("3 received") or not \
633 out.find("4 received") < 0:
638 def _blacklist_node(self, node):
640 Add node mal functioning node to blacklist
642 self.warn(" Blacklisting malfunctioning node ")
643 self.plapi.blacklist_host(node)
644 if not self._hostname:
645 self.set('hostname', None)
647 def _put_node_in_provision(self, node):
649 Add node to the list of nodes being provisioned, in order for other RMs
650 to not try to provision the same one again
652 self.plapi.reserve_host(node)
654 def _get_ip(self, node_id):
656 Query PLCAPI for the IP of a node with certain node id
658 hostname = self.plapi.get_nodes(node_id, ['hostname'])[0]
660 ip = sshfuncs.gethostbyname(hostname['hostname'])
662 # Fail while trying to find the IP
666 def fail_discovery(self):
667 msg = "Discovery failed. No candidates found for node"
669 raise RuntimeError, msg
671 def fail_node_not_alive(self, hostname=None):
672 msg = "Node %s not alive" % hostname
673 raise RuntimeError, msg
675 def fail_node_not_available(self, hostname):
676 msg = "Node %s not available for provisioning" % hostname
677 raise RuntimeError, msg
679 def fail_not_enough_nodes(self):
680 msg = "Not enough nodes available for provisioning"
681 raise RuntimeError, msg
683 def fail_plapi(self):
684 msg = "Failing while trying to instanciate the PLC API.\nSet the" + \
685 " attributes pluser and plpassword."
686 raise RuntimeError, msg
688 def valid_connection(self, guid):