Example #1
0
    def work_seek(self):
        try:
            while True:
                job = self.web_resource.queue.get(block=False)
                if job is None:
                    raise Queue.Empty()

                yield self.send(
                    self.queue_name_work, 'work', {
                        "work_amount": job.length,
                        "batchid": job.batchid,
                        "jobid": job.jobid
                    })

                extradict = {
                    "batchid": job.batchid,
                    "jobid": job.jobid,
                    "work_amount": job.length
                }
                cei_events.event("workproducer", "job_sent", extra=extradict)

                # This is an unfortunate hack to work around a memory leak in ion.
                # Some caches are only cleared after a received message is handled.
                # Since this process sends messages "spontaneously" -- triggered by a
                # LoopingCall -- we must manually clear the cache.
                self.message_client.workbench.manage_workbench_cache(
                    'Default Context')

        except Queue.Empty:
            return
        except Exception, e:
            # unhandled exceptions will terminate the LoopingCall
            log.error("Error adding work: %s", e, exc_info=True)
Example #2
0
    def work_seek(self):
        try:
            while True:
                job = self.web_resource.queue.get(block=False)
                if job is None:
                    raise Queue.Empty()

                yield self.send(self.queue_name_work, 'work', {"work_amount":job.length, "batchid":job.batchid, "jobid":job.jobid})

                extradict = {"batchid":job.batchid,
                             "jobid":job.jobid,
                             "work_amount":job.length}
                cei_events.event("workproducer", "job_sent", extra=extradict)

                # This is an unfortunate hack to work around a memory leak in ion.
                # Some caches are only cleared after a received message is handled.
                # Since this process sends messages "spontaneously" -- triggered by a
                # LoopingCall -- we must manually clear the cache.
                self.message_client.workbench.manage_workbench_cache('Default Context')

        except Queue.Empty:
            return
        except Exception,e:
            # unhandled exceptions will terminate the LoopingCall
            log.error("Error adding work: %s", e, exc_info=True)
Example #3
0
    def launch(self, deployable_type_id, launch_description, extravars=None):
        """Choose instance IDs for each instance desired, a launch ID and send
        appropriate message to Provisioner.

        Control API method, see the decision engine implementer's guide.

        @param deployable_type_id string identifier of the DP to launch
        @param launch_description See engine implementer's guide
        @param extravars Optional, see engine implementer's guide
        @retval tuple (launch_id, launch_description), see guide
        @exception Exception illegal input
        @exception Exception message not sent
        """

        # right now we are sending some node-specific data in provisioner vars
        # (node_id at least)
        if len(launch_description) != 1:
            raise NotImplementedError("Only single-node launches are supported")

        launch_id = str(uuid.uuid4())
        log.info("Request for DP '%s' is a new launch with id '%s'" % (deployable_type_id, launch_id))
        new_instance_id_list = []
        for group, item in launch_description.iteritems():
            log.info(" - %s is %d %s from %s" % (group, item.num_instances, item.allocation_id, item.site))

            if item.num_instances != 1:
                raise NotImplementedError("Only single-node launches are supported")

            for i in range(item.num_instances):
                new_instance_id = str(uuid.uuid4())
                self.state.new_instance_launch(new_instance_id, launch_id, item.site, item.allocation_id)
                item.instance_ids.append(new_instance_id)
                new_instance_id_list.append(new_instance_id)

        vars_send = self.prov_vars.copy()
        if extravars:
            vars_send.update(extravars)

        # The node_id var is the reason only single-node launches are supported.
        # It could be instead added by the provisioner or something? It also
        # is complicated by the contextualization system.
        vars_send["node_id"] = new_instance_id_list[0]
        vars_send["heartbeat_dest"] = self.controller_name

        # hide passwords from logging
        hide_password = deepcopy(vars_send)
        if "cassandra_password" in hide_password:
            hide_password["cassandra_password"] = "******"
        if "broker_password" in hide_password:
            hide_password["broker_password"] = "******"

        log.debug("Launching with parameters:\n%s" % str(hide_password))

        subscribers = (self.controller_name,)

        self.provisioner.provision(launch_id, deployable_type_id, launch_description, subscribers, vars=vars_send)
        extradict = {"launch_id": launch_id, "new_instance_ids": new_instance_id_list, "subscribers": subscribers}
        cei_events.event("controller", "new_launch", extra=extradict)
        return launch_id, launch_description
Example #4
0
 def test_event_write(self):
     self.log.debug("something")
     cei_events.event("unittest", "TRIAL1", self.log)
     self.log.debug("something-else")
     events = cei_events.events_from_file(self.logfilepath)
     assert len(events) == 1
     assert events[0].source == "unittest"
     assert events[0].name == "TRIAL1"
Example #5
0
 def test_event_write(self):
     log.debug("something")
     cei_events.event("unittest", "TRIAL1", log)
     log.debug("something-else")
     events = cei_events.events_from_file(self.logfilepath)
     assert len(events) == 1
     assert events[0].source == "unittest"
     assert events[0].name == "TRIAL1"
Example #6
0
    def query_one_site(self, site, nodes, driver=None):
        node_driver = driver or self.site_drivers[site]

        log.info('Querying site "%s"', site)
        nimboss_nodes = yield threads.deferToThread(node_driver.list_nodes)
        nimboss_nodes = dict((node.id, node) for node in nimboss_nodes)

        # note we are walking the nodes from datastore, NOT from nimboss
        for node in nodes:
            state = node['state']
            if state < states.PENDING or state >= states.TERMINATED:
                continue

            nimboss_id = node.get('iaas_id')
            nimboss_node = nimboss_nodes.pop(nimboss_id, None)
            if not nimboss_node:
                # this state is unknown to underlying IaaS. What could have
                # happened? IaaS error? Recovery from loss of net to IaaS?

                # Or lazily-updated records. On EC2, there can be a short
                # window where pending instances are not included in query
                # response

                start_time = node.get('pending_timestamp')
                now = time.time()

                if start_time and (now - start_time) <= _IAAS_NODE_QUERY_WINDOW_SECONDS:
                    log.debug('node %s: not in query of IaaS, but within '+
                            'allowed startup window (%d seconds)',
                            node['node_id'], _IAAS_NODE_QUERY_WINDOW_SECONDS)
                else:
                    log.warn('node %s: in data store but unknown to IaaS. '+
                            'Marking as terminated.', node['node_id'])

                    node['state'] = states.FAILED
                    node['state_desc'] = 'NODE_DISAPPEARED'

                    launch = yield self.store.get_launch(node['launch_id'])
                    if launch:
                        yield self.store_and_notify([node], launch['subscribers'])
            else:
                nimboss_state = _NIMBOSS_STATE_MAP[nimboss_node.state]
                if nimboss_state > node['state']:
                    #TODO nimboss could go backwards in state.
                    node['state'] = nimboss_state

                    update_node_ip_info(node, nimboss_node)

                    if nimboss_state == states.STARTED:
                        extradict = {'iaas_id': nimboss_id,
                                     'node_id': node.get('node_id'),
                                     'public_ip': node.get('public_ip'),
                                     'private_ip': node.get('private_ip') }
                        cei_events.event("provisioner", "node_started",
                                         extra=extradict)

                    launch = yield self.store.get_launch(node['launch_id'])
                    yield self.store_and_notify([node], launch['subscribers'])
Example #7
0
    def slc_init(self):
        cei_events.event("provisioner", "init_begin")

        try:
            store = self.spawn_args['store']
            site_drivers = self.spawn_args['site_drivers']
            context_client = self.spawn_args['context_client']
        except KeyError,e:
            raise KeyError("Missing provisioner spawn_arg: " + str(e))
Example #8
0
    def slc_init(self):
        cei_events.event("provisioner", "init_begin")

        try:
            store = self.spawn_args['store']
            site_drivers = self.spawn_args['site_drivers']
            context_client = self.spawn_args['context_client']
        except KeyError, e:
            raise KeyError("Missing provisioner spawn_arg: " + str(e))
Example #9
0
    def start(self):

        self.dashi.handle(self.work)
        extradict = {"queue_name_work": self.queue_name_work}
        cei_events.event("worker", "init_end", extra=extradict)

        try:
            self.dashi.consume()
        except KeyboardInterrupt:
            self.log.info("Caught terminate signal. Bye!")
Example #10
0
    def work(self, work_amount=0, batchid="", jobid=""):

        extradict = {"batchid": batchid,
                     "jobid": jobid,
                     "work_amount": work_amount}
        cei_events.event("worker", "job_begin", extra=extradict)
        self.log.info("WORK: sleeping for %d seconds ---" % work_amount)
        time.sleep(work_amount)
        cei_events.event("worker", "job_end", extra=extradict)
        return {"result": "work_complete"}
Example #11
0
 def slc_init(self):
     queue_name = self.spawn_args["queue_name_work"]
     self.workReceiver = WorkerReceiver(name=queue_name,
                                        label=__name__,
                                        scope=WorkerReceiver.SCOPE_SYSTEM,
                                        handler=self.receive)
     self.queue_name_work = self.workReceiver.xname
     extradict = {"queue_name_work":self.queue_name_work}
     cei_events.event("worker", "init_begin", extra=extradict)
     self.laterinitialized = False
     reactor.callLater(0, self.later_init)
Example #12
0
 def slc_init(self):
     queue_name = self.spawn_args["queue_name_work"]
     self.workReceiver = WorkerReceiver(name=queue_name,
                                        label=__name__,
                                        scope=WorkerReceiver.SCOPE_SYSTEM,
                                        handler=self.receive)
     self.queue_name_work = self.workReceiver.xname
     extradict = {"queue_name_work": self.queue_name_work}
     cei_events.event("worker", "init_begin", extra=extradict)
     self.laterinitialized = False
     reactor.callLater(0, self.later_init)
Example #13
0
 def test_extra_hierarchy(self):
     # note the conflicting "hello3" key in higher level:
     innerdict = {"hello3": "hello4"}
     adict = {"hello1": "hello2", "hello5": innerdict, "hello3": "hello6"}
     cei_events.event("unittest", "TRIAL1", self.log, extra=adict)
     events = cei_events.events_from_file(self.logfilepath)
     event = events[0]
     assert event.extra["hello1"] == "hello2"
     assert event.extra["hello3"] == "hello6"
     innerdict = event.extra["hello5"]
     assert isinstance(innerdict, dict)
     assert innerdict["hello3"] == "hello4"
Example #14
0
 def op_work(self, content, headers, msg):
     if not self.laterinitialized:
         log.error("message got here without the later-init")
     sleepsecs = int(content['work_amount'])
     extradict = {"batchid":content['batchid'],
                  "jobid":content['jobid'],
                  "work_amount":sleepsecs}
     cei_events.event("worker", "job_begin", extra=extradict)
     log.info("WORK: sleeping for %d seconds ---" % sleepsecs)
     yield pu.asleep(sleepsecs)
     yield self.reply(msg, 'result', {'result':'work_complete'}, {})
     cei_events.event("worker", "job_end", extra=extradict)
Example #15
0
 def test_timestamp(self):
     utc_now = datetime.datetime.utcnow()
     cei_events.event("unittest", "TRIAL1", self.log)
     events = cei_events.events_from_file(self.logfilepath)
     assert len(events) == 1
     ts = events[0].timestamp
     # It is possible that any of these values could have rolled over
     # between acquiring utc_now and recording the event.  But this is
     # unlikely enough that we'll keep this important UTC sanity check:
     assert ts.year == utc_now.year
     assert ts.month == utc_now.month
     assert ts.day == utc_now.day
     assert ts.hour == utc_now.hour
Example #16
0
    def test_extra_hierarchy(self):
        # note the conflicting "hello3" key in higher level:
        innerdict = {"hello3": "hello4"}
        adict = {"hello1": "hello2", "hello5": innerdict, "hello3": "hello6"}

        cei_events.event("unittest", "TRIAL1", log, extra=adict)
        events = cei_events.events_from_file(self.logfilepath)
        assert len(events) == 1
        assert events[0].extra["hello1"] == "hello2"
        assert events[0].extra["hello3"] == "hello6"

        innerdict = events[0].extra["hello5"]
        assert isinstance(innerdict, dict)
        assert innerdict["hello3"] == "hello4"
Example #17
0
    def test_timestamp(self):
        utc_now = datetime.datetime.utcnow()
        cei_events.event("unittest", "TRIAL1", log)
        events = cei_events.events_from_file(self.logfilepath)
        assert len(events) == 1
        ts = events[0].timestamp

        # It is possible that any of these values could have rolled over
        # between acquiring utc_now and recording the event.  But this is
        # unlikely enough that we'll keep this important UTC sanity check:
        assert ts.year == utc_now.year
        assert ts.month == utc_now.month
        assert ts.day == utc_now.day
        assert ts.hour == utc_now.hour
Example #18
0
 def op_work(self, content, headers, msg):
     if not self.laterinitialized:
         log.error("message got here without the later-init")
     sleepsecs = int(content['work_amount'])
     extradict = {
         "batchid": content['batchid'],
         "jobid": content['jobid'],
         "work_amount": sleepsecs
     }
     cei_events.event("worker", "job_begin", extra=extradict)
     log.info("WORK: sleeping for %d seconds ---" % sleepsecs)
     yield pu.asleep(sleepsecs)
     yield self.reply(msg, 'result', {'result': 'work_complete'}, {})
     cei_events.event("worker", "job_end", extra=extradict)
Example #19
0
    def add_domain(self, caller, domain_id, definition_id, config,
                   subscriber_name=None, subscriber_op=None):
        """See: EPUManagement.msg_add_domain()
        """
        # TODO: parameters are from messages, do legality checks here
        # assert that engine_conf['epuworker_type']['sleeper'] is owned by caller
        log.debug("ADD Domain: %s", config)

        # Make a copy of the definition and merge the config inside.
        # If a definition is changed or deleted, it has no impact over any of
        # the domains.
        try:
            definition = self.store.get_domain_definition(definition_id)
        except NotFoundError:
            raise ValueError("Domain definition does not exist: %s" % definition_id)

        merged_config = copy.copy(definition.get_definition())

        # Hand-made deep merge of config into definition
        if EPUM_CONF_GENERAL in config:
            if EPUM_CONF_GENERAL in merged_config:
                merged_config[EPUM_CONF_GENERAL].update(config[EPUM_CONF_GENERAL])
            else:
                merged_config[EPUM_CONF_GENERAL] = config[EPUM_CONF_GENERAL]

        if EPUM_CONF_HEALTH in config:
            if EPUM_CONF_HEALTH in merged_config:
                merged_config[EPUM_CONF_HEALTH].update(config[EPUM_CONF_HEALTH])
            else:
                merged_config[EPUM_CONF_HEALTH] = config[EPUM_CONF_HEALTH]

        if EPUM_CONF_ENGINE in config:
            if EPUM_CONF_ENGINE in merged_config:
                merged_config[EPUM_CONF_ENGINE].update(config[EPUM_CONF_ENGINE])
            else:
                merged_config[EPUM_CONF_ENGINE] = config[EPUM_CONF_ENGINE]

        self._validate_engine_config(merged_config)

        with EpuLoggerThreadSpecific(domain=domain_id, user=caller):
            domain = self.store.add_domain(caller, domain_id, merged_config)

            if subscriber_name and subscriber_op:
                domain.add_subscriber(subscriber_name, subscriber_op)

        extradict = {'user': caller, 'domain_id': domain_id,
                'definition_id': definition_id}
        cei_events.event("epumanagement", "new_domain", extra=extradict)
Example #20
0
class ProvisionerService(ServiceProcess):
    """Provisioner service interface
    """

    # Declaration of service
    declare = ServiceProcess.service_declare(name='provisioner',
                                             version='0.1.0',
                                             dependencies=[])

    @defer.inlineCallbacks
    def slc_init(self):
        cei_events.event("provisioner", "init_begin")

        try:
            store = self.spawn_args['store']
            site_drivers = self.spawn_args['site_drivers']
            context_client = self.spawn_args['context_client']
        except KeyError, e:
            raise KeyError("Missing provisioner spawn_arg: " + str(e))

        self.store = store

        notifier = self.spawn_args.get('notifier')
        self.notifier = notifier or ProvisionerNotifier(self)
        self.dtrs = DeployableTypeRegistryClient(self)

        self.core = ProvisionerCore(self.store, self.notifier, self.dtrs,
                                    site_drivers, context_client)
        yield self.core.recover()
        cei_events.event("provisioner", "init_end")

        # operator can disable new launches
        self.enabled = True
        self.terminate_all_deferred = None
Example #21
0
    def test_manual_event_write(self):
        cruft = "some cruft %s" % cei_events.event_logtxt("unittest", "TRIAL1")
        log.warning(cruft)
        events = cei_events.events_from_file(self.logfilepath)
        assert len(events) == 1

        cei_events.event("unittest", "TRIAL2", log)
        events = cei_events.events_from_file(self.logfilepath)
        assert len(events) == 2

        cruft = "cruft2 %s" % cei_events.event_logtxt("unittest", "TRIAL3")
        log.warning(cruft)

        events = cei_events.events_from_file(self.logfilepath)
        assert len(events) == 3

        found = {"TRIAL1": False, "TRIAL2": False, "TRIAL3": False}
        for ev in events:
            if found.has_key(ev.name):
                found[ev.name] = True
        for val in found.values():
            assert val
Example #22
0
    def __init__(self, *args, **kwargs):

        configs = ["service", "epu_worker"]
        config_files = get_config_paths(configs)
        self.CFG = bootstrap.configure(config_files)

        self.log = logging.getLogger()

        try:
            if os.environ.get('EPU_USE_GEVENT'):
                bootstrap.enable_gevent()
            else:
                self.log.info("Using standard python Threading")
        except:
            self.log.warning("gevent not available. Falling back to threading")

        self.queue_name_work = self.CFG.queue_name_work
        extradict = {"queue_name_work": self.queue_name_work}

        cei_events.event("worker", "init_begin", extra=extradict)

        self.dashi = bootstrap.dashi_connect(self.topic, self.CFG)
Example #23
0
    def test_manual_event_write(self):
        cruft = "some cruft %s" % cei_events.event_logtxt("unittest", "TRIAL1")
        self.log.warning(cruft)
        events = cei_events.events_from_file(self.logfilepath)
        assert len(events) == 1

        cei_events.event("unittest", "TRIAL2", self.log)
        events = cei_events.events_from_file(self.logfilepath)
        assert len(events) == 2

        cruft = "cruft2 %s" % cei_events.event_logtxt("unittest", "TRIAL3")
        self.log.warning(cruft)

        events = cei_events.events_from_file(self.logfilepath)
        assert len(events) == 3

        found = {"TRIAL1": False, "TRIAL2": False, "TRIAL3": False}
        for ev in events:
            if ev.name in found:
                found[ev.name] = True
        for val in found.values():
            assert val
Example #24
0
File: decider.py Project: ooici/epu
    def launch(self, deployable_type_id, site, allocation, count=1, extravars=None):
        """
        Choose instance IDs for each instance desired, a launch ID and send
        appropriate message to Provisioner.

        @param deployable_type_id string identifier of the DT to launch
        @param site IaaS site to launch on
        @param allocation IaaS allocation (size) to request
        @param count number of instances to launch
        @param extravars Optional, see engine implementer's guide
        @retval tuple (launch_id, instance_ids), see guide
        @exception Exception illegal input
        @exception Exception message not sent
        """

        # right now we are sending some node-specific data in provisioner vars
        # (node_id at least)
        if count != 1:
            raise NotImplementedError("Only single-node launches are supported")

        launch_id = str(uuid.uuid4())
        log.info("Request for DT '%s' is a new launch with id '%s'", deployable_type_id, launch_id)

        if extravars:
            extravars = deepcopy(extravars)
        else:
            extravars = None

        new_instance_id = str(uuid.uuid4())
        instance = self.domain.new_instance_launch(deployable_type_id,
            new_instance_id, launch_id, site, allocation, extravars=extravars)
        new_instance_id_list = (new_instance_id,)

        self.execute_instance_launch(instance)
        extradict = {"launch_id": launch_id,
                     "new_instance_ids": new_instance_id_list}
        cei_events.event("controller", "new_launch", extra=extradict)
        return launch_id, new_instance_id_list
Example #25
0
 def test_unique_keys(self):
     cei_events.event("unittest", "NAME", self.log)
     cei_events.event("unittest", "NAME", self.log)
     cei_events.event("unittest", "NAME", self.log)
     cei_events.event("unittest", "NAME", self.log)
     cei_events.event("unittest", "NAME", self.log)
     cei_events.event("unittest", "NAME", self.log)
     cei_events.event("unittest", "NAME", self.log)
     events = cei_events.events_from_file(self.logfilepath)
     assert len(events) == 7
     uniqs = {}
     for ev in events:
         uniqs[ev.timestamp] = None
     assert len(uniqs) == 7
Example #26
0
 def later_init(self):
     spawnId = yield self.workReceiver.attach()
     log.debug("spawnId: %s" % spawnId)
     self.laterinitialized = True
     extradict = {"queue_name_work":self.queue_name_work}
     cei_events.event("worker", "init_end", extra=extradict)
Example #27
0
    def slc_init(self):

        scoped_name = self.get_scoped_name("system", self.svc_name)
        self.scoped_name = scoped_name

        queue_name_work = self.spawn_args.get("queue_name_work")
        if queue_name_work:
            self.queue_name_work = self.get_scoped_name(
                "system", queue_name_work)

            extradict = {"queue_name_work": self.queue_name_work}
            cei_events.event(self.svc_name, "init_begin", extra=extradict)
            yield self._make_queue(queue_name_work)

            queuestat_client = QueueStatClient(self)
            yield queuestat_client.watch_queue(self.queue_name_work,
                                               self.scoped_name, 'sensor_info')
            cei_events.event(self.svc_name, "queue_watched")

        else:
            self.worker_queue_receiver = None
            self.queue_name_work = None
            extradict = None
            cei_events.event(self.svc_name, "init_begin", extra=extradict)

        engineclass = "epu.decisionengine.impls.NpreservingEngine"
        if self.spawn_args.has_key("engine_class"):
            engineclass = self.spawn_args["engine_class"]
            log.info("Using configured decision engine: %s" % engineclass)
        else:
            log.info("Using default decision engine: %s" % engineclass)

        if self.spawn_args.has_key("engine_conf"):
            engine_conf = self.spawn_args["engine_conf"]
            if isinstance(engine_conf, str):
                engine_conf = json.loads(engine_conf)
        else:
            engine_conf = None

        if self.spawn_args.has_key("cassandra"):
            cass = self.spawn_args["cassandra"]
            host = cass['hostname']
            username = cass['username']
            password = cass['password']
            port = cass['port']
            keyspace = cass['keyspace']

            store = CassandraControllerStore(self.svc_name, host, port,
                                             username, password, keyspace,
                                             CoreInstance, SensorItem)
            store.initialize()
            store.activate()
        elif self.spawn_args.has_key('store'):
            store = self.spawn_args['store']
        else:
            store = ControllerStore()

        self.core = ControllerCore(ProvisionerClient(self),
                                   engineclass,
                                   scoped_name,
                                   conf=engine_conf,
                                   store=store)

        # run state recovery and engine initialization

        # this one needs to run before any messages start arriving. It pulls
        # information from persistence and refreshes local caches.
        yield self.core.run_recovery()

        # temporarily doing this later due to a potential bug in ioncore where
        # queues may not be bound before slc_init runs. This means  if the
        # provisioner is quck to reply to dump_state some messages may be
        # missed.
        reactor.callLater(1, self._delayed_init)
Example #28
0
    def launch(self, deployable_type_id, launch_description, extravars=None):
        """Choose instance IDs for each instance desired, a launch ID and send
        appropriate message to Provisioner.

        Control API method, see the decision engine implementer's guide.

        @param deployable_type_id string identifier of the DP to launch
        @param launch_description See engine implementer's guide
        @param extravars Optional, see engine implementer's guide
        @retval tuple (launch_id, launch_description), see guide
        @exception Exception illegal input
        @exception Exception message not sent
        """

        # right now we are sending some node-specific data in provisioner vars
        # (node_id at least)
        if len(launch_description) != 1:
            raise NotImplementedError(
                "Only single-node launches are supported")

        launch_id = str(uuid.uuid4())
        log.info("Request for DP '%s' is a new launch with id '%s'" %
                 (deployable_type_id, launch_id))
        new_instance_id_list = []
        for group, item in launch_description.iteritems():
            log.info(
                " - %s is %d %s from %s" %
                (group, item.num_instances, item.allocation_id, item.site))

            if item.num_instances != 1:
                raise NotImplementedError(
                    "Only single-node launches are supported")

            for i in range(item.num_instances):
                new_instance_id = str(uuid.uuid4())
                self.state.new_instance_launch(new_instance_id, launch_id,
                                               item.site, item.allocation_id)
                item.instance_ids.append(new_instance_id)
                new_instance_id_list.append(new_instance_id)

        vars_send = self.prov_vars.copy()
        if extravars:
            vars_send.update(extravars)

        # The node_id var is the reason only single-node launches are supported.
        # It could be instead added by the provisioner or something? It also
        # is complicated by the contextualization system.
        vars_send['node_id'] = new_instance_id_list[0]
        vars_send['heartbeat_dest'] = self.controller_name

        # hide passwords from logging
        hide_password = deepcopy(vars_send)
        if 'cassandra_password' in hide_password:
            hide_password['cassandra_password'] = '******'
        if 'broker_password' in hide_password:
            hide_password['broker_password'] = '******'

        log.debug("Launching with parameters:\n%s" % str(hide_password))

        subscribers = (self.controller_name, )

        self.provisioner.provision(launch_id,
                                   deployable_type_id,
                                   launch_description,
                                   subscribers,
                                   vars=vars_send)
        extradict = {
            "launch_id": launch_id,
            "new_instance_ids": new_instance_id_list,
            "subscribers": subscribers
        }
        cei_events.event("controller", "new_launch", extra=extradict)
        return launch_id, launch_description
Example #29
0
 def test_event_sourcefilter2(self):
     cei_events.event("SRC1", "NM1", self.log)
     self.log.debug("something not an event")
     cei_events.event("SRX2", "NM2", self.log)
     cei_events.event("SRC3", "NM3", self.log)
     self.log.debug("something not an event")
     cei_events.event("SRX4", "NM4", self.log)
     cei_events.event("SRC5", "NM5", self.log)
     cei_events.event("SRC6", "NM6", self.log)
     path = self.logfilepath
     events = cei_events.events_from_file(path, sourcefilter="SRC")
     assert len(events) == 4
Example #30
0
 def test_event_namefilter2(self):
     cei_events.event("unittest", "NM1", log)
     log.debug("something not an event")
     cei_events.event("unittest", "XX2", log)
     cei_events.event("unittest", "NM3", log)
     log.debug("something not an event")
     cei_events.event("unittest", "XX4", log)
     cei_events.event("unittest", "NM5", log)
     cei_events.event("unittest", "XX6", log)
     path = self.logfilepath
     events = cei_events.events_from_file(path, namefilter="NM")
     assert len(events) == 3
Example #31
0
 def op_de_state(self, content, headers, msg):
     state = self.core.de_state()
     extradict = {"state": state}
     cei_events.event(self.svc_name, "de_state", extra=extradict)
     yield self.reply_ok(msg, state)
Example #32
0
 def test_event_nameandsourcefilter(self):
     cei_events.event("SRC1", "NX1", log)
     log.debug("something not an event")
     cei_events.event("SRX2", "NM2", log)
     cei_events.event("SRC3", "XX3", log)
     cei_events.event("SRX4", "XX4", log)
     cei_events.event("SRC5", "NM5", log)
     log.debug("something not an event")
     cei_events.event("SRC6", "NM6", log)
     path = self.logfilepath
     events = cei_events.events_from_file(path,
                                          sourcefilter="SRC",
                                          namefilter="NM")
     assert len(events) == 2
Example #33
0
    def _delayed_init(self):
        yield self.core.run_initialize()

        self.core.begin_controlling()
        cei_events.event(self.svc_name, "init_end")
Example #34
0
 def test_extra(self):
     adict = {"hello1":"hello2"}
     cei_events.event("unittest", "TRIAL1", log, extra=adict)
     events = cei_events.events_from_file(self.logfilepath)
     assert len(events) == 1
     assert events[0].extra["hello1"] == "hello2"
Example #35
0
 def test_unique_keys(self):
     cei_events.event("unittest", "NAME", log)
     cei_events.event("unittest", "NAME", log)
     cei_events.event("unittest", "NAME", log)
     cei_events.event("unittest", "NAME", log)
     cei_events.event("unittest", "NAME", log)
     cei_events.event("unittest", "NAME", log)
     cei_events.event("unittest", "NAME", log)
     events = cei_events.events_from_file(self.logfilepath)
     assert len(events) == 7
     uniqs = {}
     for ev in events:
         uniqs[ev.timestamp] = None
     assert len(uniqs) == 7
Example #36
0
 def test_event_nameandsourcefilter(self):
     cei_events.event("SRC1", "NX1", self.log)
     self.log.debug("something not an event")
     cei_events.event("SRX2", "NM2", self.log)
     cei_events.event("SRC3", "XX3", self.log)
     cei_events.event("SRX4", "XX4", self.log)
     cei_events.event("SRC5", "NM5", self.log)
     self.log.debug("something not an event")
     cei_events.event("SRC6", "NM6", self.log)
     path = self.logfilepath
     events = cei_events.events_from_file(path, sourcefilter="SRC", namefilter="NM")
     assert len(events) == 2
Example #37
0
    def _delayed_init(self):
        yield self.core.run_initialize()

        self.core.begin_controlling()
        cei_events.event(self.svc_name, "init_end")
Example #38
0
 def test_extra_integer_values(self):
     adict = {"hello1": 34}
     cei_events.event("unittest", "TRIAL1", log, extra=adict)
     events = cei_events.events_from_file(self.logfilepath)
     assert len(events) == 1
     assert events[0].extra["hello1"] == 34
Example #39
0
 def op_de_state(self, content, headers, msg):
     state = self.core.de_state()
     extradict = {"state":state}
     cei_events.event(self.svc_name, "de_state", extra=extradict)
     yield self.reply_ok(msg, state)
Example #40
0
 def later_init(self):
     spawnId = yield self.workReceiver.attach()
     log.debug("spawnId: %s" % spawnId)
     self.laterinitialized = True
     extradict = {"queue_name_work": self.queue_name_work}
     cei_events.event("worker", "init_end", extra=extradict)
Example #41
0
    def slc_init(self):

        scoped_name = self.get_scoped_name("system", self.svc_name)
        self.scoped_name = scoped_name

        queue_name_work = self.spawn_args.get("queue_name_work")
        if queue_name_work:
            self.queue_name_work = self.get_scoped_name("system", queue_name_work)

            extradict = {"queue_name_work":self.queue_name_work}
            cei_events.event(self.svc_name, "init_begin", extra=extradict)
            yield self._make_queue(queue_name_work)

            queuestat_client = QueueStatClient(self)
            yield queuestat_client.watch_queue(self.queue_name_work, self.scoped_name, 'sensor_info')
            cei_events.event(self.svc_name, "queue_watched")

        else:
            self.worker_queue_receiver = None
            self.queue_name_work = None
            extradict = None
            cei_events.event(self.svc_name, "init_begin", extra=extradict)

        engineclass = "epu.decisionengine.impls.NpreservingEngine"
        if self.spawn_args.has_key("engine_class"):
            engineclass = self.spawn_args["engine_class"]
            log.info("Using configured decision engine: %s" % engineclass)
        else:
            log.info("Using default decision engine: %s" % engineclass)

        if self.spawn_args.has_key("engine_conf"):
            engine_conf = self.spawn_args["engine_conf"]
            if isinstance(engine_conf, str):
                engine_conf = json.loads(engine_conf)
        else:
            engine_conf = None

        if self.spawn_args.has_key("cassandra"):
            cass = self.spawn_args["cassandra"]
            host = cass['hostname']
            username = cass['username']
            password = cass['password']
            port = cass['port']
            keyspace = cass['keyspace']

            store = CassandraControllerStore(self.svc_name, host, port,
                                             username, password, keyspace,
                                             CoreInstance, SensorItem)
            store.initialize()
            store.activate()
        elif self.spawn_args.has_key('store'):
            store = self.spawn_args['store']
        else:
            store = ControllerStore()

        self.core = ControllerCore(ProvisionerClient(self), engineclass,
                                   scoped_name, conf=engine_conf, store=store)

        # run state recovery and engine initialization

        # this one needs to run before any messages start arriving. It pulls
        # information from persistence and refreshes local caches.
        yield self.core.run_recovery()

        # temporarily doing this later due to a potential bug in ioncore where
        # queues may not be bound before slc_init runs. This means  if the
        # provisioner is quck to reply to dump_state some messages may be
        # missed.
        reactor.callLater(1, self._delayed_init)
Example #42
0
            message = '%s nodes from IaaS launch but %s were expected' % (
                    len(iaas_nodes), len(nodes))
            log.error(message)
            raise ProvisioningError('IAAS_PROBLEM '+ message)

        for node_rec, iaas_node in izip(nodes, iaas_nodes):
            node_rec['iaas_id'] = iaas_node.id

            update_node_ip_info(node_rec, iaas_node)

            node_rec['state'] = states.PENDING
            node_rec['pending_timestamp'] = time.time()

            extradict = {'public_ip': node_rec.get('public_ip'),
                         'iaas_id': iaas_node.id, 'node_id': node_rec['node_id']}
            cei_events.event("provisioner", "new_node", extra=extradict)

    @defer.inlineCallbacks
    def store_and_notify(self, records, subscribers):
        """Convenience method to store records and notify subscribers.
        """
        yield self.store.put_nodes(records)
        yield self.notifier.send_records(records, subscribers)

    @defer.inlineCallbacks
    def dump_state(self, nodes, force_subscribe=None):
        """Resends node state information to subscribers

        @param nodes list of node IDs
        @param force_subscribe optional, an extra subscriber that may not be listed in local node records
        """
Example #43
0
 def test_extra_integer_values(self):
     adict = {"hello1": 34}
     cei_events.event("unittest", "TRIAL1", self.log, extra=adict)
     events = cei_events.events_from_file(self.logfilepath)
     assert len(events) == 1
     assert events[0].extra["hello1"] == 34
Example #44
0
 def test_event_namefilter2(self):
     cei_events.event("unittest", "NM1", self.log)
     self.log.debug("something not an event")
     cei_events.event("unittest", "XX2", self.log)
     cei_events.event("unittest", "NM3", self.log)
     self.log.debug("something not an event")
     cei_events.event("unittest", "XX4", self.log)
     cei_events.event("unittest", "NM5", self.log)
     cei_events.event("unittest", "XX6", self.log)
     path = self.logfilepath
     events = cei_events.events_from_file(path, namefilter="NM")
     assert len(events) == 3
Example #45
0
 def test_event_sourcefilter2(self):
     cei_events.event("SRC1", "NM1", log)
     log.debug("something not an event")
     cei_events.event("SRX2", "NM2", log)
     cei_events.event("SRC3", "NM3", log)
     log.debug("something not an event")
     cei_events.event("SRX4", "NM4", log)
     cei_events.event("SRC5", "NM5", log)
     cei_events.event("SRC6", "NM6", log)
     path = self.logfilepath
     events = cei_events.events_from_file(path, sourcefilter="SRC")
     assert len(events) == 4