#
# Author: Alina Quereilhac <alina.quereilhac@inria.fr>
-from nepi.util.timefuncs import strfnow, strfdiff, strfvalid
+from nepi.util.timefuncs import tnow, tdiff, tdiffsec, stabsformat
from nepi.util.logger import Logger
from nepi.execution.trace import TraceAttr
import pkgutil
import weakref
-reschedule_delay = "0.5s"
+reschedule_delay = "1s"
class ResourceAction:
+ """ Action that a user can order to a Resource Manager
+
+ """
DEPLOY = 0
START = 1
STOP = 2
class ResourceState:
+ """ State of a Resource Manager
+
+ """
NEW = 0
DISCOVERED = 1
PROVISIONED = 2
@classmethod
def rtype(cls):
+ """ Returns the type of the Resource Manager
+
+ """
return cls._rtype
@classmethod
# the resource instance gets a copy of all traces
self._trcs = copy.deepcopy(self._traces)
- self._state = ResourceState.NEW
+ # Each resource is placed on a deployment group by the EC
+ # during deployment
+ self.deployment_group = None
self._start_time = None
self._stop_time = None
self._provision_time = None
self._ready_time = None
self._release_time = None
+ self._finish_time = None
+ self._failed_time = None
+
+ self._state = ResourceState.NEW
@property
def guid(self):
+ """ Returns the global unique identifier of the RM """
return self._guid
@property
def ec(self):
+ """ Returns the Experiment Controller """
return self._ec()
@property
def connections(self):
+ """ Returns the set of guids of connected RMs"""
return self._connections
@property
def conditions(self):
+ """ Returns the conditions to which the RM is subjected to.
+
+ The object returned by this method is a dictionary indexed by
+ ResourceAction."""
return self._conditions
@property
def start_time(self):
- """ Returns timestamp with the time the RM started """
+ """ Returns the start time of the RM as a timestamp"""
return self._start_time
@property
def stop_time(self):
- """ Returns timestamp with the time the RM stopped """
+ """ Returns the stop time of the RM as a timestamp"""
return self._stop_time
@property
def discover_time(self):
- """ Returns timestamp with the time the RM passed to state discovered """
+ """ Returns the time discovering was finished for the RM as a timestamp"""
return self._discover_time
@property
def provision_time(self):
- """ Returns timestamp with the time the RM passed to state provisioned """
+ """ Returns the time provisioning was finished for the RM as a timestamp"""
return self._provision_time
@property
def ready_time(self):
- """ Returns timestamp with the time the RM passed to state ready """
+ """ Returns the time deployment was finished for the RM as a timestamp"""
return self._ready_time
@property
def release_time(self):
- """ Returns timestamp with the time the RM was released """
+ """ Returns the release time of the RM as a timestamp"""
return self._release_time
+ @property
+ def finish_time(self):
+ """ Returns the finalization time of the RM as a timestamp"""
+ return self._finish_time
+
+ @property
+ def failed_time(self):
+ """ Returns the time failure occured for the RM as a timestamp"""
+ return self._failed_time
+
@property
def state(self):
+ """ Get the state of the current RM """
return self._state
def log_message(self, msg):
+ """ Returns the log message formatted with added information.
+
+ :param msg: text message
+ :type msg: str
+ :rtype: str
+ """
return " %s guid: %d - %s " % (self._rtype, self.guid, msg)
- def connect(self, guid):
+ def register_connection(self, guid):
+ """ Registers a connection to the RM identified by guid
+
+ :param guid: Global unique identified of the RM to connect to
+ :type guid: int
+ """
if self.valid_connection(guid):
+ self.connect(guid)
self._connections.add(guid)
+ def unregister_connection(self, guid):
+ """ Removes a registered connection to the RM identified by guid
+
+ :param guid: Global unique identified of the RM to connect to
+ :type guid: int
+ """
+ if guid in self._connections:
+ self.disconnect(guid)
+ self._connections.remove(guid)
+
def discover(self):
- self._discover_time = strfnow()
- self._state = ResourceState.DISCOVERED
+ """ Performs resource discovery.
+
+ This method is resposible for selecting an individual resource
+ matching user requirements.
+ This method should be redefined when necessary in child classes.
+ """
+ self.set_discovered()
def provision(self):
- self._provision_time = strfnow()
- self._state = ResourceState.PROVISIONED
+ """ Performs resource provisioning.
- def start(self):
- """ Start the Resource Manager
+ This method is resposible for provisioning one resource.
+ After this method has been successfully invoked, the resource
+ should be acccesible/controllable by the RM.
+ This method should be redefined when necessary in child classes.
+ """
+ self.set_provisioned()
+ def start(self):
+ """ Starts the resource.
+
+ There is no generic start behavior for all resources.
+ This method should be redefined when necessary in child classes.
"""
- if not self._state in [ResourceState.READY, ResourceState.STOPPED]:
+ if not self.state in [ResourceState.READY, ResourceState.STOPPED]:
self.error("Wrong state %s for start" % self.state)
return
- self._start_time = strfnow()
- self._state = ResourceState.STARTED
+ self.set_started()
def stop(self):
- """ Start the Resource Manager
-
+ """ Stops the resource.
+
+ There is no generic stop behavior for all resources.
+ This method should be redefined when necessary in child classes.
"""
- if not self._state in [ResourceState.STARTED]:
+ if not self.state in [ResourceState.STARTED]:
self.error("Wrong state %s for stop" % self.state)
return
+
+ self.set_stopped()
- self._stop_time = strfnow()
- self._state = ResourceState.STOPPED
+ def deploy(self):
+ """ Execute all steps required for the RM to reach the state READY
+
+ """
+ if self.state > ResourceState.READY:
+ self.error("Wrong state %s for deploy" % self.state)
+ return
+
+ self.debug("----- READY ---- ")
+ self.set_ready()
+
+ def release(self):
+ self.set_released()
+
+ def finish(self):
+ self.set_finished()
+
+ def fail(self):
+ self.set_failed()
def set(self, name, value):
""" Set the value of the attribute
attr.value = value
def get(self, name):
- """ Start the Resource Manager
+ """ Returns the value of the attribute
:param name: Name of the attribute
:type name: str
attr = self._attrs[name]
return attr.value
- def register_trace(self, name):
- """ Enable trace
+ def enable_trace(self, name):
+ """ Explicitly enable trace generation
:param name: Name of the trace
:type name: str
"""
trace = self._trcs[name]
trace.enabled = True
+
+ def trace_enabled(self, name):
+ """Returns True if trace is enables
+ :param name: Name of the trace
+ :type name: str
+ """
+ trace = self._trcs[name]
+ return trace.enabled
+
def trace(self, name, attr = TraceAttr.ALL, block = 512, offset = 0):
""" Get information on collected trace
"""
pass
- def register_condition(self, action, group, state,
- time = None):
+ def register_condition(self, action, group, state, time = None):
""" Registers a condition on the resource manager to allow execution
of 'action' only after 'time' has elapsed from the moment all resources
in 'group' reached state 'state'
:type time: str
"""
+
+ if not action in self.conditions:
+ self._conditions[action] = list()
+
conditions = self.conditions.get(action)
- if not conditions:
- conditions = list()
- self._conditions[action] = conditions
# For each condition to register a tuple of (group, state, time) is
# added to the 'action' list
conditions.append((group, state, time))
- def get_connected(self, rtype):
+ def unregister_condition(self, group, action = None):
+ """ Removed conditions for a certain group of guids
+
+ :param action: Action to restrict to condition (either 'START' or 'STOP')
+ :type action: str
+
+ :param group: Group of RMs to wait for (list of guids)
+ :type group: int or list of int
+
+ """
+ # For each condition a tuple of (group, state, time) is
+ # added to the 'action' list
+ if not isinstance(group, list):
+ group = [group]
+
+ for act, conditions in self.conditions.iteritems():
+ if action and act != action:
+ continue
+
+ for condition in list(conditions):
+ (grp, state, time) = condition
+
+ # If there is an intersection between grp and group,
+ # then remove intersected elements
+ intsec = set(group).intersection(set(grp))
+ if intsec:
+ idx = conditions.index(condition)
+ newgrp = set(grp)
+ newgrp.difference_update(intsec)
+ conditions[idx] = (newgrp, state, time)
+
+ def get_connected(self, rtype = None):
+ """ Returns the list of RM with the type 'rtype'
+
+ :param rtype: Type of the RM we look for
+ :type rtype: str
+ :return: list of guid
+ """
connected = []
+ rclass = ResourceFactory.get_resource_type(rtype)
for guid in self.connections:
rm = self.ec.get_resource(guid)
- if rm.rtype() == rtype:
+ if not rtype or isinstance(rm, rclass):
connected.append(rm)
return connected
for guid in group:
rm = self.ec.get_resource(guid)
# If the RM state is lower than the requested state we must
- # reschedule (e.g. if RM is READY but we required STARTED)
+ # reschedule (e.g. if RM is READY but we required STARTED).
if rm.state < state:
reschedule = True
break
# Only keep time information for START and STOP
break
- d = strfdiff(strfnow(), t)
- wait = strfdiff(strfvalid(time),strfvalid(str(d)+"s"))
+ # time already elapsed since RM changed state
+ waited = "%fs" % tdiffsec(tnow(), t)
+
+ # time still to wait
+ wait = tdiffsec(stabsformat(time), stabsformat(waited))
+
if wait > 0.001:
reschedule = True
delay = "%fs" % wait
break
+
return reschedule, delay
def set_with_conditions(self, name, value, group, state, time):
""" Set value 'value' on attribute with name 'name' when 'time'
- has elapsed since all elements in 'group' have reached state
- 'state'
+ has elapsed since all elements in 'group' have reached state
+ 'state'
:param name: Name of the attribute to set
:type name: str
:type state: str
:param time: Time to wait after 'state' is reached on all RMs in group. (e.g. '2s')
:type time: str
-
"""
reschedule = False
self.debug(" ----- STOPPING ---- ")
self.stop()
- def deploy(self):
- """ Execute all steps required for the RM to reach the state READY
-
+ def connect(self, guid):
+ """ Performs actions that need to be taken upon associating RMs.
+ This method should be redefined when necessary in child classes.
"""
- if self._state > ResourceState.READY:
- self.error("Wrong state %s for deploy" % self.state)
- return
-
- self.debug("----- READY ---- ")
- self._ready_time = strfnow()
- self._state = ResourceState.READY
-
- def release(self):
- """Clean the resource at the end of the Experiment and change the status
+ pass
+ def disconnect(self, guid):
+ """ Performs actions that need to be taken upon disassociating RMs.
+ This method should be redefined when necessary in child classes.
"""
- self._release_time = strfnow()
- self._state = ResourceState.RELEASED
+ pass
def valid_connection(self, guid):
- """Check if the connection is available.
+ """Checks whether a connection with the other RM
+ is valid.
+ This method need to be redefined by each new Resource Manager.
:param guid: Guid of the current Resource Manager
:type guid: int
"""
# TODO: Validate!
return True
+
+ def set_started(self):
+ """ Mark ResourceManager as STARTED """
+ self._start_time = tnow()
+ self._state = ResourceState.STARTED
+
+ def set_stopped(self):
+ """ Mark ResourceManager as STOPPED """
+ self._stop_time = tnow()
+ self._state = ResourceState.STOPPED
+
+ def set_ready(self):
+ """ Mark ResourceManager as READY """
+ self._ready_time = tnow()
+ self._state = ResourceState.READY
+
+ def set_released(self):
+ """ Mark ResourceManager as REALEASED """
+ self._release_time = tnow()
+ self._state = ResourceState.RELEASED
+
+ def set_finished(self):
+ """ Mark ResourceManager as FINISHED """
+ self._finish_time = tnow()
+ self._state = ResourceState.FINISHED
+
+ def set_failed(self):
+ """ Mark ResourceManager as FAILED """
+ self._failed_time = tnow()
+ self._state = ResourceState.FAILED
+
+ def set_discovered(self):
+ """ Mark ResourceManager as DISCOVERED """
+ self._discover_time = tnow()
+ self._state = ResourceState.DISCOVERED
+
+ def set_provisioned(self):
+ """ Mark ResourceManager as PROVISIONED """
+ self._provision_time = tnow()
+ self._state = ResourceState.PROVISIONED
class ResourceFactory(object):
_resource_types = dict()
@classmethod
def resource_types(cls):
+ """Return the type of the Class"""
return cls._resource_types
+ @classmethod
+ def get_resource_type(cls, rtype):
+ """Return the type of the Class"""
+ return cls._resource_types.get(rtype)
+
@classmethod
def register_type(cls, rclass):
+ """Register a new Ressource Manager"""
cls._resource_types[rclass.rtype()] = rclass
@classmethod
def create(cls, rtype, ec, guid):
+ """Create a new instance of a Ressource Manager"""
rclass = cls._resource_types[rtype]
return rclass(ec, guid)
def populate_factory():
- for rclass in find_types():
- ResourceFactory.register_type(rclass)
+ """Register all the possible RM that exists in the current version of Nepi.
+ """
+ # Once the factory is populated, don't repopulate
+ if not ResourceFactory.resource_types():
+ for rclass in find_types():
+ ResourceFactory.register_type(rclass)
def find_types():
+ """Look into the different folders to find all the
+ availables Resources Managers
+ """
search_path = os.environ.get("NEPI_SEARCH_PATH", "")
search_path = set(search_path.split(" "))
types = []
- for importer, modname, ispkg in pkgutil.walk_packages(search_path):
+ for importer, modname, ispkg in pkgutil.walk_packages(search_path,
+ prefix = "nepi.resources."):
+
loader = importer.find_module(modname)
+
try:
- module = loader.load_module(loader.fullname)
+ # Notice: Repeated calls to load_module will act as a reload of teh module
+ module = loader.load_module(modname)
+
for attrname in dir(module):
if attrname.startswith("_"):
continue
import logging
err = traceback.format_exc()
logger = logging.getLogger("Resource.find_types()")
- logger.error("Error while lading Resource Managers %s" % err)
+ logger.error("Error while loading Resource Managers %s" % err)
return types
git://git.onelab.eu / nepi.git / blobdiff