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 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 if self.get("gateway") or self.get("gatewayUser"):
202 self.set("gateway", None)
203 self.set("gatewayUser", None)
205 def _skip_provision(self):
206 pl_user = self.get("pluser")
207 pl_pass = self.get("plpassword")
208 if not pl_user and not pl_pass:
215 pl_user = self.get("pluser")
216 pl_pass = self.get("plpassword")
217 pl_url = self.get("plcApiUrl")
218 pl_ptn = self.get("plcApiPattern")
220 self._plapi = PLCAPIFactory.get_api(pl_user, pl_pass, pl_url,
228 def do_discover(self):
230 Based on the attributes defined by the user, discover the suitable
233 if self._skip_provision():
234 super(PlanetlabNode, self).do_discover()
237 hostname = self._get_hostname()
239 # the user specified one particular node to be provisioned
240 self._hostname = True
241 node_id = self._get_nodes_id({'hostname':hostname})
242 node_id = node_id.pop()['node_id']
244 # check that the node is not blacklisted or being provisioned
246 with PlanetlabNode.lock:
247 plist = self.plapi.reserved()
248 blist = self.plapi.blacklisted()
249 if node_id not in blist and node_id not in plist:
251 # check that is really alive, by performing ping
252 ping_ok = self._do_ping(node_id)
254 self._blacklist_node(node_id)
255 self.fail_node_not_alive(hostname)
257 if self._check_if_in_slice([node_id]):
258 self._slicenode = True
259 self._put_node_in_provision(node_id)
260 self._node_to_provision = node_id
262 self.fail_node_not_available(hostname)
263 super(PlanetlabNode, self).do_discover()
266 # the user specifies constraints based on attributes, zero, one or
267 # more nodes can match these constraints
268 nodes = self._filter_based_on_attributes()
270 # nodes that are already part of user's slice have the priority to
272 nodes_inslice = self._check_if_in_slice(nodes)
273 nodes_not_inslice = list(set(nodes) - set(nodes_inslice))
277 node_id = self._choose_random_node(nodes_inslice)
278 self._slicenode = True
281 # Either there were no matching nodes in the user's slice, or
282 # the nodes in the slice were blacklisted or being provisioned
283 # by other RM. Note nodes_not_inslice is never empty
284 node_id = self._choose_random_node(nodes_not_inslice)
285 self._slicenode = False
288 self._node_to_provision = node_id
290 self._set_hostname_attr(node_id)
291 self.info(" Selected node to provision ")
292 super(PlanetlabNode, self).do_discover()
294 with PlanetlabNode.lock:
295 self._blacklist_node(node_id)
298 self.fail_not_enough_nodes()
300 def do_provision(self):
302 Add node to user's slice after verifing that the node is functioning
305 if self._skip_provision():
306 super(PlanetlabNode, self).do_provision()
314 while not provision_ok:
315 node = self._node_to_provision
316 if not self._slicenode:
317 self._add_node_to_slice(node)
319 # check ssh connection
321 while t < timeout and not ssh_ok:
323 cmd = 'echo \'GOOD NODE\''
324 ((out, err), proc) = self.execute(cmd)
325 if out.find("GOOD NODE") < 0:
333 cmd = 'echo \'GOOD NODE\''
334 ((out, err), proc) = self.execute(cmd)
335 if not out.find("GOOD NODE") < 0:
339 # the timeout was reach without establishing ssh connection
340 # the node is blacklisted, deleted from the slice, and a new
341 # node to provision is discovered
342 with PlanetlabNode.lock:
343 self.warn(" Could not SSH login ")
344 self._blacklist_node(node)
345 #self._delete_node_from_slice(node)
349 # check /proc directory is mounted (ssh_ok = True)
350 # and file system is not read only
352 cmd = 'mount |grep proc'
353 ((out1, err1), proc1) = self.execute(cmd)
354 cmd = 'touch /tmp/tmpfile; rm /tmp/tmpfile'
355 ((out2, err2), proc2) = self.execute(cmd)
356 if out1.find("/proc type proc") < 0 or \
357 "Read-only file system".lower() in err2.lower():
358 with PlanetlabNode.lock:
359 self.warn(" Corrupted file system ")
360 self._blacklist_node(node)
361 #self._delete_node_from_slice(node)
367 if not self.get('hostname'):
368 self._set_hostname_attr(node)
370 ip = self._get_ip(node)
373 super(PlanetlabNode, self).do_provision()
375 def _filter_based_on_attributes(self):
377 Retrive the list of nodes ids that match user's constraints
379 # Map user's defined attributes with tagnames of PlanetLab
380 timeframe = self.get("timeframe")[0]
383 'country' : 'country',
385 'architecture' : 'arch',
386 'operatingSystem' : 'fcdistro',
387 #'site' : 'pldistro',
388 'minReliability' : 'reliability%s' % timeframe,
389 'maxReliability' : 'reliability%s' % timeframe,
390 'minBandwidth' : 'bw%s' % timeframe,
391 'maxBandwidth' : 'bw%s' % timeframe,
392 'minLoad' : 'load%s' % timeframe,
393 'maxLoad' : 'load%s' % timeframe,
394 'minCpu' : 'cpu%s' % timeframe,
395 'maxCpu' : 'cpu%s' % timeframe,
401 for attr_name, attr_obj in self._attrs.iteritems():
402 attr_value = self.get(attr_name)
404 if attr_value is not None and attr_obj.flags == 8 and \
405 attr_name != 'timeframe':
407 attr_tag = attr_to_tags[attr_name]
408 filters['tagname'] = attr_tag
410 # filter nodes by fixed constraints e.g. operating system
411 if not 'min' in attr_name and not 'max' in attr_name:
412 filters['value'] = attr_value
413 nodes_id = self._filter_by_fixed_attr(filters, nodes_id)
415 # filter nodes by range constraints e.g. max bandwidth
416 elif ('min' or 'max') in attr_name:
417 nodes_id = self._filter_by_range_attr(attr_name, attr_value, filters, nodes_id)
420 nodes = self._get_nodes_id()
422 nodes_id.append(node['node_id'])
425 def _filter_by_fixed_attr(self, filters, nodes_id):
427 Query PLCAPI for nodes ids matching fixed attributes defined by the
430 node_tags = self.plapi.get_node_tags(filters)
431 if node_tags is not None:
433 if len(nodes_id) == 0:
434 # first attribute being matched
435 for node_tag in node_tags:
436 nodes_id.append(node_tag['node_id'])
438 # remove the nodes ids that don't match the new attribute
439 # that is being match
442 for node_tag in node_tags:
443 if node_tag['node_id'] in nodes_id:
444 nodes_id_tmp.append(node_tag['node_id'])
446 if len(nodes_id_tmp):
447 nodes_id = set(nodes_id) & set(nodes_id_tmp)
449 # no node from before match the new constraint
450 self.fail_discovery()
452 # no nodes match the filter applied
453 self.fail_discovery()
457 def _filter_by_range_attr(self, attr_name, attr_value, filters, nodes_id):
459 Query PLCAPI for nodes ids matching attributes defined in a certain
462 node_tags = self.plapi.get_node_tags(filters)
465 if len(nodes_id) == 0:
466 # first attribute being matched
467 for node_tag in node_tags:
469 # check that matches the min or max restriction
470 if 'min' in attr_name and node_tag['value'] != 'n/a' and \
471 float(node_tag['value']) > attr_value:
472 nodes_id.append(node_tag['node_id'])
474 elif 'max' in attr_name and node_tag['value'] != 'n/a' and \
475 float(node_tag['value']) < attr_value:
476 nodes_id.append(node_tag['node_id'])
479 # remove the nodes ids that don't match the new attribute
480 # that is being match
482 for node_tag in node_tags:
484 # check that matches the min or max restriction and was a
485 # matching previous filters
486 if 'min' in attr_name and node_tag['value'] != 'n/a' and \
487 float(node_tag['value']) > attr_value and \
488 node_tag['node_id'] in nodes_id:
489 nodes_id_tmp.append(node_tag['node_id'])
491 elif 'max' in attr_name and node_tag['value'] != 'n/a' and \
492 float(node_tag['value']) < attr_value and \
493 node_tag['node_id'] in nodes_id:
494 nodes_id_tmp.append(node_tag['node_id'])
496 if len(nodes_id_tmp):
497 nodes_id = set(nodes_id) & set(nodes_id_tmp)
499 # no node from before match the new constraint
500 self.fail_discovery()
503 # no nodes match the filter applied
504 self.fail_discovery()
508 def _choose_random_node(self, nodes):
510 From the possible nodes for provision, choose randomly to decrese the
511 probability of different RMs choosing the same node for provision
516 index = randint(0, size)
517 node_id = nodes[index]
518 nodes[index] = nodes[size]
520 # check the node is not blacklisted or being provision by other RM
521 # and perform ping to check that is really alive
522 with PlanetlabNode.lock:
524 blist = self.plapi.blacklisted()
525 plist = self.plapi.reserved()
526 if node_id not in blist and node_id not in plist:
527 ping_ok = self._do_ping(node_id)
529 self._set_hostname_attr(node_id)
530 self.warn(" Node not responding PING ")
531 self._blacklist_node(node_id)
533 # discovered node for provision, added to provision list
534 self._put_node_in_provision(node_id)
537 def _get_nodes_id(self, filters=None):
538 return self.plapi.get_nodes(filters, fields=['node_id'])
540 def _add_node_to_slice(self, node_id):
541 self.info(" Adding node to slice ")
542 slicename = self.get("username")
543 with PlanetlabNode.lock:
544 slice_nodes = self.plapi.get_slice_nodes(slicename)
545 slice_nodes.append(node_id)
546 self.plapi.add_slice_nodes(slicename, slice_nodes)
548 def _delete_node_from_slice(self, node):
549 self.warn(" Deleting node from slice ")
550 slicename = self.get("username")
551 self.plapi.delete_slice_node(slicename, [node])
553 def _get_hostname(self):
554 hostname = self.get("hostname")
559 hostname = socket.gethostbyaddr(ip)[0]
560 self.set('hostname', hostname)
565 def _set_hostname_attr(self, node):
567 Query PLCAPI for the hostname of a certain node id and sets the
568 attribute hostname, it will over write the previous value
570 hostname = self.plapi.get_nodes(node, ['hostname'])
571 self.set("hostname", hostname[0]['hostname'])
573 def _check_if_in_slice(self, nodes_id):
575 Query PLCAPI to find out if any node id from nodes_id is in the user's
578 slicename = self.get("username")
579 slice_nodes = self.plapi.get_slice_nodes(slicename)
580 nodes_inslice = list(set(nodes_id) & set(slice_nodes))
583 def _do_ping(self, node_id):
585 Perform ping command on node's IP matching node id
588 ip = self._get_ip(node_id)
589 if not ip: return ping_ok
591 command = ['ping', '-c4']
594 (out, err) = lexec(command)
595 if not out.find("2 received") or not out.find("3 received") or not \
596 out.find("4 received") < 0:
601 def _blacklist_node(self, node):
603 Add node mal functioning node to blacklist
605 self.warn(" Blacklisting malfunctioning node ")
606 self.plapi.blacklist_host(node)
607 if not self._hostname:
608 self.set('hostname', None)
610 def _put_node_in_provision(self, node):
612 Add node to the list of nodes being provisioned, in order for other RMs
613 to not try to provision the same one again
615 self.plapi.reserve_host(node)
617 def _get_ip(self, node_id):
619 Query PLCAPI for the IP of a node with certain node id
621 hostname = self.plapi.get_nodes(node_id, ['hostname'])[0]
623 ip = sshfuncs.gethostbyname(hostname['hostname'])
625 # Fail while trying to find the IP
629 def fail_discovery(self):
630 msg = "Discovery failed. No candidates found for node"
632 raise RuntimeError, msg
634 def fail_node_not_alive(self, hostname=None):
635 msg = "Node %s not alive" % hostname
636 raise RuntimeError, msg
638 def fail_node_not_available(self, hostname):
639 msg = "Node %s not available for provisioning" % hostname
640 raise RuntimeError, msg
642 def fail_not_enough_nodes(self):
643 msg = "Not enough nodes available for provisioning"
644 raise RuntimeError, msg
646 def fail_plapi(self):
647 msg = "Failing while trying to instanciate the PLC API.\nSet the" + \
648 " attributes pluser and plpassword."
649 raise RuntimeError, msg
651 def valid_connection(self, guid):