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: Lucia Guevgeozian <lucia.guevgeozian_odizzio@inria.fr>
20 from nepi.execution.attribute import Attribute, Flags, Types
21 from nepi.execution.resource import ResourceManager, clsinit_copy, \
22 ResourceState, reschedule_delay
23 from nepi.resources.linux.node import LinuxNode
24 from nepi.util.sfaapi import SFAAPIFactory
25 from nepi.util.execfuncs import lexec
26 from nepi.util import sshfuncs
28 from random import randint
37 class PlanetlabSfaNode(LinuxNode):
38 _rtype = "PlanetlabSfaNode"
39 _help = "Controls a PlanetLab host accessible using a SSH key " \
40 "and provisioned using SFA"
41 _backend = "planetlab"
44 def _register_attributes(cls):
46 sfa_user = Attribute("sfauser", "SFA user",
47 flags = Flags.Credential)
49 sfa_private_key = Attribute("sfaPrivateKey", "SFA path to the private key \
50 used to generate the user credential")
52 city = Attribute("city", "Constrain location (city) during resource \
53 discovery. May use wildcards.",
56 country = Attribute("country", "Constrain location (country) during \
57 resource discovery. May use wildcards.",
60 region = Attribute("region", "Constrain location (region) during \
61 resource discovery. May use wildcards.",
64 architecture = Attribute("architecture", "Constrain architecture \
65 during resource discovery.",
66 type = Types.Enumerate,
71 operating_system = Attribute("operatingSystem", "Constrain operating \
72 system during resource discovery.",
73 type = Types.Enumerate,
81 min_reliability = Attribute("minReliability", "Constrain reliability \
82 while picking PlanetLab nodes. Specifies a lower \
88 max_reliability = Attribute("maxReliability", "Constrain reliability \
89 while picking PlanetLab nodes. Specifies an upper \
95 min_bandwidth = Attribute("minBandwidth", "Constrain available \
96 bandwidth while picking PlanetLab nodes. \
97 Specifies a lower acceptable bound.",
100 flags = Flags.Filter)
102 max_bandwidth = Attribute("maxBandwidth", "Constrain available \
103 bandwidth while picking PlanetLab nodes. \
104 Specifies an upper acceptable bound.",
107 flags = Flags.Filter)
109 min_load = Attribute("minLoad", "Constrain node load average while \
110 picking PlanetLab nodes. Specifies a lower acceptable \
114 flags = Flags.Filter)
116 max_load = Attribute("maxLoad", "Constrain node load average while \
117 picking PlanetLab nodes. Specifies an upper acceptable \
121 flags = Flags.Filter)
123 min_cpu = Attribute("minCpu", "Constrain available cpu time while \
124 picking PlanetLab nodes. Specifies a lower acceptable \
128 flags = Flags.Filter)
130 max_cpu = Attribute("maxCpu", "Constrain available cpu time while \
131 picking PlanetLab nodes. Specifies an upper acceptable \
135 flags = Flags.Filter)
137 timeframe = Attribute("timeframe", "Past time period in which to check\
138 information about the node. Values are year,month, \
141 type = Types.Enumerate,
146 flags = Flags.Filter)
148 plblacklist = Attribute("persist_blacklist", "Take into account the file plblacklist \
149 in the user's home directory under .nepi directory. This file \
150 contains a list of PL nodes to blacklist, and at the end \
151 of the experiment execution the new blacklisted nodes are added.",
154 flags = Flags.Global)
156 cls._register_attribute(sfa_user)
157 cls._register_attribute(sfa_private_key)
158 cls._register_attribute(city)
159 cls._register_attribute(country)
160 cls._register_attribute(region)
161 cls._register_attribute(architecture)
162 cls._register_attribute(operating_system)
163 cls._register_attribute(min_reliability)
164 cls._register_attribute(max_reliability)
165 cls._register_attribute(min_bandwidth)
166 cls._register_attribute(max_bandwidth)
167 cls._register_attribute(min_load)
168 cls._register_attribute(max_load)
169 cls._register_attribute(min_cpu)
170 cls._register_attribute(max_cpu)
171 cls._register_attribute(timeframe)
172 cls._register_attribute(plblacklist)
174 def __init__(self, ec, guid):
175 super(PlanetlabSfaNode, self).__init__(ec, guid)
177 self._ecobj = weakref.ref(ec)
179 self._node_to_provision = None
180 self._slicenode = False
181 self._hostname = False
183 if self.get("gateway") or self.get("gatewayUser"):
184 self.set("gateway", None)
185 self.set("gatewayUser", None)
187 def _skip_provision(self):
188 sfa_user = self.get("sfauser")
196 sfa_user = self.get("sfauser")
197 sfa_sm = "http://sfa3.planet-lab.eu:12346/"
198 sfa_auth = '.'.join(sfa_user.split('.')[:2])
199 sfa_registry = "http://sfa3.planet-lab.eu:12345/"
200 sfa_private_key = self.get("sfaPrivateKey")
202 _sfaapi = SFAAPIFactory.get_api(sfa_user, sfa_auth,
203 sfa_registry, sfa_sm, sfa_private_key, self._ecobj())
208 self._sfaapi = weakref.ref(_sfaapi)
210 return self._sfaapi()
212 def do_discover(self):
214 Based on the attributes defined by the user, discover the suitable
217 if self._skip_provision():
218 super(PlanetlabSfaNode, self).do_discover()
221 nodes = self.sfaapi.get_resources_hrn()
223 hostname = self._get_hostname()
225 # the user specified one particular node to be provisioned
226 self._hostname = True
227 host_hrn = nodes[hostname]
229 # check that the node is not blacklisted or being provisioned
231 if not self._blacklisted(host_hrn):
232 if not self._reserved(host_hrn):
233 # Node in reservation
234 ping_ok = self._do_ping(hostname)
236 self._blacklist_node(host_hrn)
237 self.fail_node_not_alive(hostname)
239 if self._check_if_in_slice([host_hrn]):
240 self._slicenode = True
241 self._node_to_provision = host_hrn
242 super(PlanetlabSfaNode, self).do_discover()
245 # # the user specifies constraints based on attributes, zero, one or
246 # # more nodes can match these constraints
247 # nodes = self._filter_based_on_attributes()
249 # # nodes that are already part of user's slice have the priority to
251 # nodes_inslice = self._check_if_in_slice(nodes)
252 # nodes_not_inslice = list(set(nodes) - set(nodes_inslice))
256 # node_id = self._choose_random_node(nodes_inslice)
257 # self._slicenode = True
260 # # Either there were no matching nodes in the user's slice, or
261 # # the nodes in the slice were blacklisted or being provisioned
262 # # by other RM. Note nodes_not_inslice is never empty
263 # node_id = self._choose_random_node(nodes_not_inslice)
264 # self._slicenode = False
267 # self._node_to_provision = node_id
269 # self._set_hostname_attr(node_id)
270 # self.info(" Selected node to provision ")
271 # super(PlanetlabSfaNode, self).do_discover()
273 # with PlanetlabSfaNode.lock:
274 # self._blacklist_node(node_id)
277 # self.fail_not_enough_nodes()
279 def _blacklisted(self, host_hrn):
280 if self.sfaapi.blacklisted(host_hrn):
281 self.fail_node_not_available(hostname)
284 def _reserved(self, host_hrn):
285 if self.sfaapi.reserved(host_hrn):
286 self.fail_node_not_available(hostname)
289 def do_provision(self):
291 Add node to user's slice after verifing that the node is functioning
294 if self._skip_provision():
295 super(PlanetlabSfaNode, self).do_provision()
303 while not provision_ok:
304 node = self._node_to_provision
305 if not self._slicenode:
306 self._add_node_to_slice(node)
308 # check ssh connection
310 while t < timeout and not ssh_ok:
312 cmd = 'echo \'GOOD NODE\''
313 ((out, err), proc) = self.execute(cmd)
314 if out.find("GOOD NODE") < 0:
315 self.debug( "No SSH connection, waiting 60s" )
320 self.debug( "SSH OK" )
324 cmd = 'echo \'GOOD NODE\''
325 ((out, err), proc) = self.execute(cmd)
326 if not out.find("GOOD NODE") < 0:
330 # the timeout was reach without establishing ssh connection
331 # the node is blacklisted, deleted from the slice, and a new
332 # node to provision is discovered
333 self.warn(" Could not SSH login ")
334 self._blacklist_node(node)
338 # check /proc directory is mounted (ssh_ok = True)
339 # and file system is not read only
341 cmd = 'mount |grep proc'
342 ((out1, err1), proc1) = self.execute(cmd)
343 cmd = 'touch /tmp/tmpfile; rm /tmp/tmpfile'
344 ((out2, err2), proc2) = self.execute(cmd)
345 if out1.find("/proc type proc") < 0 or \
346 "Read-only file system".lower() in err2.lower():
347 self.warn(" Corrupted file system ")
348 self._blacklist_node(node)
354 if not self.get('hostname'):
355 self._set_hostname_attr(node)
356 self.info(" Node provisioned ")
358 super(PlanetlabSfaNode, self).do_provision()
360 # def _filter_based_on_attributes(self):
362 # Retrive the list of nodes hrn that match user's constraints
364 # # Map user's defined attributes with tagnames of PlanetLab
365 # timeframe = self.get("timeframe")[0]
368 # 'country' : 'country',
369 # 'region' : 'region',
370 # 'architecture' : 'arch',
371 # 'operatingSystem' : 'fcdistro',
372 # 'minReliability' : 'reliability%s' % timeframe,
373 # 'maxReliability' : 'reliability%s' % timeframe,
374 # 'minBandwidth' : 'bw%s' % timeframe,
375 # 'maxBandwidth' : 'bw%s' % timeframe,
376 # 'minLoad' : 'load%s' % timeframe,
377 # 'maxLoad' : 'load%s' % timeframe,
378 # 'minCpu' : 'cpu%s' % timeframe,
379 # 'maxCpu' : 'cpu%s' % timeframe,
385 # for attr_name, attr_obj in self._attrs.iteritems():
386 # attr_value = self.get(attr_name)
388 # if attr_value is not None and attr_obj.has_flag(Flags.Filter) and \
389 # attr_name != 'timeframe':
391 # attr_tag = attr_to_tags[attr_name]
392 # filters['tagname'] = attr_tag
394 # # filter nodes by fixed constraints e.g. operating system
395 # if not 'min' in attr_name and not 'max' in attr_name:
396 # filters['value'] = attr_value
397 # nodes_hrn = self._filter_by_fixed_attr(filters, nodes_hrn)
399 # # filter nodes by range constraints e.g. max bandwidth
400 # elif ('min' or 'max') in attr_name:
401 # nodes_hrn = self._filter_by_range_attr(attr_name, attr_value, filters, nodes_hrn)
404 # nodes = self.sfaapi.get_resources_hrn()
406 # nodes_hrn.append(node[node.key()])
409 # def _filter_by_fixed_attr(self, filters, nodes_hrn):
411 # Query SFA API for nodes matching fixed attributes defined by the
415 ## node_tags = self.sfaapi.get_resources_tags(filters)
416 ## if node_tags is not None:
418 ## if len(nodes_id) == 0:
419 ## # first attribute being matched
420 ## for node_tag in node_tags:
421 ## nodes_id.append(node_tag['node_id'])
423 ## # remove the nodes ids that don't match the new attribute
424 ## # that is being match
427 ## for node_tag in node_tags:
428 ## if node_tag['node_id'] in nodes_id:
429 ## nodes_id_tmp.append(node_tag['node_id'])
431 ## if len(nodes_id_tmp):
432 ## nodes_id = set(nodes_id) & set(nodes_id_tmp)
434 ## # no node from before match the new constraint
435 ## self.fail_discovery()
437 ## # no nodes match the filter applied
438 ## self.fail_discovery()
442 # def _filter_by_range_attr(self, attr_name, attr_value, filters, nodes_id):
444 # Query PLCAPI for nodes ids matching attributes defined in a certain
448 ## node_tags = self.plapi.get_node_tags(filters)
451 ## if len(nodes_id) == 0:
452 ## # first attribute being matched
453 ## for node_tag in node_tags:
455 ## # check that matches the min or max restriction
456 ## if 'min' in attr_name and node_tag['value'] != 'n/a' and \
457 ## float(node_tag['value']) > attr_value:
458 ## nodes_id.append(node_tag['node_id'])
460 ## elif 'max' in attr_name and node_tag['value'] != 'n/a' and \
461 ## float(node_tag['value']) < attr_value:
462 ## nodes_id.append(node_tag['node_id'])
465 ## # remove the nodes ids that don't match the new attribute
466 ## # that is being match
468 ## for node_tag in node_tags:
470 ## # check that matches the min or max restriction and was a
471 ## # matching previous filters
472 ## if 'min' in attr_name and node_tag['value'] != 'n/a' and \
473 ## float(node_tag['value']) > attr_value and \
474 ## node_tag['node_id'] in nodes_id:
475 ## nodes_id_tmp.append(node_tag['node_id'])
477 ## elif 'max' in attr_name and node_tag['value'] != 'n/a' and \
478 ## float(node_tag['value']) < attr_value and \
479 ## node_tag['node_id'] in nodes_id:
480 ## nodes_id_tmp.append(node_tag['node_id'])
482 ## if len(nodes_id_tmp):
483 ## nodes_id = set(nodes_id) & set(nodes_id_tmp)
485 ## # no node from before match the new constraint
486 ## self.fail_discovery()
489 ## # no nodes match the filter applied
490 ## self.fail_discovery()
494 # def _choose_random_node(self, nodes):
496 # From the possible nodes for provision, choose randomly to decrese the
497 # probability of different RMs choosing the same node for provision
502 # index = randint(0, size)
503 # node_id = nodes[index]
504 # nodes[index] = nodes[size]
506 # # check the node is not blacklisted or being provision by other RM
507 # # and perform ping to check that is really alive
508 # with PlanetlabNode.lock:
510 # blist = self.plapi.blacklisted()
511 # plist = self.plapi.reserved()
512 # if node_id not in blist and node_id not in plist:
513 # ping_ok = self._do_ping(node_id)
515 # self._set_hostname_attr(node_id)
516 # self.warn(" Node not responding PING ")
517 # self._blacklist_node(node_id)
519 # # discovered node for provision, added to provision list
520 # self._put_node_in_provision(node_id)
523 # def _get_nodes_id(self, filters=None):
524 # return self.plapi.get_nodes(filters, fields=['node_id'])
526 def _add_node_to_slice(self, host_hrn):
527 self.info(" Adding node to slice ")
528 slicename = self.get("username").replace('_', '.')
529 slicename = 'ple.' + slicename
530 self.sfaapi.add_resource_to_slice(slicename, host_hrn)
532 # def _delete_node_from_slice(self, node):
533 # self.warn(" Deleting node from slice ")
534 # slicename = self.get("username")
535 # self.plapi.delete_slice_node(slicename, [node])
537 def _get_hostname(self):
538 hostname = self.get("hostname")
544 def _set_hostname_attr(self, node):
546 Query SFAAPI for the hostname of a certain host hrn and sets the
547 attribute hostname, it will over write the previous value
549 hosts_hrn = self.sfaapi.get_resources_hrn()
550 for hostname, hrn in hosts_hrn.iteritems():
552 self.set("hostname", hostname)
554 def _check_if_in_slice(self, hosts_hrn):
556 Check using SFA API if any host hrn from hosts_hrn is in the user's
559 slicename = self.get("username").replace('_', '.')
560 slicename = 'ple.' + slicename
561 slice_nodes = self.sfaapi.get_slice_resources(slicename)['resource']
562 slice_nodes_hrn = self.sfaapi.get_resources_hrn(slice_nodes)
563 nodes_inslice = list(set(hosts_hrn) & set(slice_nodes_hrn.values()))
566 def _do_ping(self, hostname):
568 Perform ping command on node's IP matching hostname
571 ip = self._get_ip(hostname)
573 command = "ping -c4 %s" % ip
574 (out, err) = lexec(command)
576 m = re.search("(\d+)% packet loss", str(out))
577 if m and int(m.groups()[0]) < 50:
582 def _blacklist_node(self, host_hrn):
584 Add node mal functioning node to blacklist
586 self.warning(" Blacklisting malfunctioning node ")
587 self.sfaapi.blacklist_resource(host_hrn)
588 if not self._hostname:
589 self.set('hostname', None)
591 def _reserve(self, host_hrn):
593 Add node to the list of nodes being provisioned, in order for other RMs
594 to not try to provision the same one again.
596 self.sfaapi.reserve_resource(host_hrn)
598 def _get_ip(self, hostname):
600 Query cache for the IP of a node with certain hostname
603 ip = sshfuncs.gethostbyname(hostname)
605 # Fail while trying to find the IP
609 def fail_discovery(self):
610 msg = "Discovery failed. No candidates found for node"
612 raise RuntimeError, msg
614 def fail_node_not_alive(self, hostname=None):
615 msg = "Node %s not alive" % hostname
616 raise RuntimeError, msg
618 def fail_node_not_available(self, hostname):
619 msg = "Node %s not available for provisioning" % hostname
620 raise RuntimeError, msg
622 def fail_not_enough_nodes(self):
623 msg = "Not enough nodes available for provisioning"
624 raise RuntimeError, msg
626 def fail_sfaapi(self):
627 msg = "Failing while trying to instanciate the SFA API.\nSet the" + \
628 " attributes sfauser and sfaPrivateKey."
629 raise RuntimeError, msg
631 def valid_connection(self, guid):