X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=src%2Fnepi%2Fresources%2Fplanetlab%2Fnode.py;h=c6f7bde861087ddd15ae19df13fc122da88ca7e4;hb=6285ca51026efb69642eea9dfc7c480e722d84a9;hp=d2461bccd7a9438cf300d670451dad1d47a68ba6;hpb=52cf8298f89325a9b64e04b5db85f862134f09eb;p=nepi.git diff --git a/src/nepi/resources/planetlab/node.py b/src/nepi/resources/planetlab/node.py index d2461bcc..c6f7bde8 100644 --- a/src/nepi/resources/planetlab/node.py +++ b/src/nepi/resources/planetlab/node.py @@ -3,9 +3,8 @@ # Copyright (C) 2013 INRIA # # This program is free software: you can redistribute it and/or modify -# it under the terms of the GNU General Public License as published by -# the Free Software Foundation, either version 3 of the License, or -# (at your option) any later version. +# it under the terms of the GNU General Public License version 2 as +# published by the Free Software Foundation; # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of @@ -20,24 +19,27 @@ from nepi.execution.attribute import Attribute, Flags, Types from nepi.execution.resource import ResourceManager, clsinit_copy, \ - ResourceState, reschedule_delay + ResourceState from nepi.resources.linux.node import LinuxNode from nepi.resources.planetlab.plcapi import PLCAPIFactory from nepi.util.execfuncs import lexec from nepi.util import sshfuncs from random import randint +import re +import os import time import socket import threading import datetime +import weakref @clsinit_copy class PlanetlabNode(LinuxNode): - _rtype = "PlanetlabNode" + _rtype = "planetlab::Node" _help = "Controls a PlanetLab host accessible using a SSH key " \ "associated to a PlanetLab user account" - _backend = "planetlab" + _platform = "planetlab" lock = threading.Lock() @@ -94,14 +96,6 @@ class PlanetlabNode(LinuxNode): "other"], flags = Flags.Filter) - #site = Attribute("site", "Constrain the PlanetLab site this node \ - # should reside on.", - # type = Types.Enumerate, - # allowed = ["PLE", - # "PLC", - # "PLJ"], - # flags = Flags.Filter) - min_reliability = Attribute("minReliability", "Constrain reliability \ while picking PlanetLab nodes. Specifies a lower \ acceptable bound.", @@ -169,21 +163,19 @@ class PlanetlabNode(LinuxNode): "year"], flags = Flags.Filter) -# plblacklist = Attribute("blacklist", "Take into account the file plblacklist \ -# in the user's home directory under .nepi directory. This file \ -# contains a list of PL nodes to blacklist, and at the end \ -# of the experiment execution the new blacklisted nodes are added.", -# type = Types.Bool, -# default = True, -# flags = Flags.ReadOnly) -# + plblacklist = Attribute("persist_blacklist", "Take into account the file plblacklist \ + in the user's home directory under .nepi directory. This file \ + contains a list of PL nodes to blacklist, and at the end \ + of the experiment execution the new blacklisted nodes are added.", + type = Types.Bool, + default = False, + flags = Flags.Global) cls._register_attribute(ip) cls._register_attribute(pl_url) cls._register_attribute(pl_ptn) cls._register_attribute(pl_user) cls._register_attribute(pl_password) - #cls._register_attribute(site) cls._register_attribute(city) cls._register_attribute(country) cls._register_attribute(region) @@ -198,10 +190,12 @@ class PlanetlabNode(LinuxNode): cls._register_attribute(min_cpu) cls._register_attribute(max_cpu) cls._register_attribute(timeframe) + cls._register_attribute(plblacklist) def __init__(self, ec, guid): super(PlanetlabNode, self).__init__(ec, guid) + self._ecobj = weakref.ref(ec) self._plapi = None self._node_to_provision = None self._slicenode = False @@ -211,6 +205,18 @@ class PlanetlabNode(LinuxNode): self.set("gateway", None) self.set("gatewayUser", None) + # Blacklist file + nepi_home = os.path.join(os.path.expanduser("~"), ".nepi") + plblacklist_file = os.path.join(nepi_home, "plblacklist.txt") + if not os.path.exists(plblacklist_file): + if os.path.isdir(nepi_home): + with open(plblacklist_file, 'w') as clear: + pass + else: + os.makedirs(nepi_home) + with open(plblacklist_file, 'w') as clear: + pass + def _skip_provision(self): pl_user = self.get("pluser") pl_pass = self.get("plpassword") @@ -225,14 +231,15 @@ class PlanetlabNode(LinuxNode): pl_pass = self.get("plpassword") pl_url = self.get("plcApiUrl") pl_ptn = self.get("plcApiPattern") - - self._plapi = PLCAPIFactory.get_api(pl_user, pl_pass, pl_url, - pl_ptn) + _plapi = PLCAPIFactory.get_api(pl_user, pl_pass, pl_url, + pl_ptn, self._ecobj()) - if not self._plapi: + if not _plapi: self.fail_plapi() + + self._plapi = weakref.ref(_plapi) - return self._plapi + return self._plapi() def do_discover(self): """ @@ -318,14 +325,21 @@ class PlanetlabNode(LinuxNode): provision_ok = False ssh_ok = False proc_ok = False - timeout = 3600 + timeout = 1800 while not provision_ok: node = self._node_to_provision if not self._slicenode: self._add_node_to_slice(node) - self.info( " Node added to slice ") - + if self._check_if_in_slice([node]): + self.debug( "Node added to slice" ) + else: + self.warning(" Could not add to slice ") + with PlanetlabNode.lock: + self._blacklist_node(node) + self.do_discover() + continue + # check ssh connection t = 0 while t < timeout and not ssh_ok: @@ -333,12 +347,12 @@ class PlanetlabNode(LinuxNode): cmd = 'echo \'GOOD NODE\'' ((out, err), proc) = self.execute(cmd) if out.find("GOOD NODE") < 0: - self.info(" No SSH login, sleeping for 60 seconds ") + self.debug( "No SSH connection, waiting 60s" ) t = t + 60 time.sleep(60) continue else: - self.info(" SSH login OK ") + self.debug( "SSH OK" ) ssh_ok = True continue else: @@ -385,6 +399,12 @@ class PlanetlabNode(LinuxNode): super(PlanetlabNode, self).do_provision() + def do_release(self): + super(PlanetlabNode, self).do_release() + if self.state == ResourceState.RELEASED and not self._skip_provision(): + self.debug(" Releasing PLC API ") + self.plapi.release() + def _filter_based_on_attributes(self): """ Retrive the list of nodes ids that match user's constraints @@ -397,7 +417,6 @@ class PlanetlabNode(LinuxNode): 'region' : 'region', 'architecture' : 'arch', 'operatingSystem' : 'fcdistro', - #'site' : 'pldistro', 'minReliability' : 'reliability%s' % timeframe, 'maxReliability' : 'reliability%s' % timeframe, 'minBandwidth' : 'bw%s' % timeframe, @@ -555,6 +574,7 @@ class PlanetlabNode(LinuxNode): slicename = self.get("username") with PlanetlabNode.lock: slice_nodes = self.plapi.get_slice_nodes(slicename) + self.debug(" Previous slice nodes %s " % slice_nodes) slice_nodes.append(node_id) self.plapi.add_slice_nodes(slicename, slice_nodes) @@ -599,14 +619,13 @@ class PlanetlabNode(LinuxNode): """ ping_ok = False ip = self._get_ip(node_id) - if not ip: return ping_ok - - command = "ping -c4 %s" % ip + if ip: + command = "ping -c4 %s" % ip + (out, err) = lexec(command) - (out, err) = lexec(command) - if not str(out).find("2 received") < 0 or not str(out).find("3 received") < 0 or not \ - str(out).find("4 received") < 0: - ping_ok = True + m = re.search("(\d+)% packet loss", str(out)) + if m and int(m.groups()[0]) < 50: + ping_ok = True return ping_ok @@ -642,24 +661,24 @@ class PlanetlabNode(LinuxNode): def fail_discovery(self): msg = "Discovery failed. No candidates found for node" self.error(msg) - raise RuntimeError, msg + raise RuntimeError(msg) def fail_node_not_alive(self, hostname=None): msg = "Node %s not alive" % hostname - raise RuntimeError, msg + raise RuntimeError(msg) def fail_node_not_available(self, hostname): msg = "Node %s not available for provisioning" % hostname - raise RuntimeError, msg + raise RuntimeError(msg) def fail_not_enough_nodes(self): msg = "Not enough nodes available for provisioning" - raise RuntimeError, msg + raise RuntimeError(msg) def fail_plapi(self): msg = "Failing while trying to instanciate the PLC API.\nSet the" + \ " attributes pluser and plpassword." - raise RuntimeError, msg + raise RuntimeError(msg) def valid_connection(self, guid): # TODO: Validate!