Python UnitOfWorkDao.update方法代码示例

# 需要导入模块: from synergy.db.dao.unit_of_work_dao import UnitOfWorkDao [as 别名]
# 或者: from synergy.db.dao.unit_of_work_dao.UnitOfWorkDao import update [as 别名]
class AbstractUowAwareWorker(AbstractMqWorker):
    """ Abstract class is inherited by all workers/aggregators
    that are aware of unit_of_work and capable of processing it """

    def __init__(self, process_name, perform_db_logging=False):
        super(AbstractUowAwareWorker, self).__init__(process_name)
        self.perform_db_logging = perform_db_logging
        self.uow_dao = UnitOfWorkDao(self.logger)
        self.mq_transmitter = MqTransmitter(self.logger)

    def __del__(self):
        del self.mq_transmitter
        super(AbstractUowAwareWorker, self).__del__()

    # **************** Abstract Methods ************************
    def _init_performance_tracker(self, logger):
        self.performance_tracker = UowAwareTracker(logger)
        self.performance_tracker.start()

    def _process_uow(self, uow):
        """
        :param uow: unit_of_work to process
        :return: a tuple (number of processed items/documents/etc, desired unit_of_work state) or None
        if None is returned then it is assumed that the return tuple is (0, unit_of_work.STATE_PROCESSED)
        :raise an Exception if the UOW shall be marked as STATE_INVALID
        """
        raise NotImplementedError('method _process_uow must be implemented by {0}'.format(self.__class__.__name__))

    def _clean_up(self):
        """ method is called from the *finally* clause and is suppose to clean up after the uow processing """
        pass

    def _mq_callback(self, message):
        try:
            mq_request = MqTransmission.from_json(message.body)
            uow = self.uow_dao.get_one(mq_request.record_db_id)
            if not uow.is_requested:
                # accept only UOW in STATE_REQUESTED
                self.logger.warning('Skipping UOW: id {0}; state {1};'.format(message.body, uow.state),
                                    exc_info=False)
                self.consumer.acknowledge(message.delivery_tag)
                return
        except Exception:
            self.logger.error('Safety fuse. Can not identify UOW {0}'.format(message.body), exc_info=True)
            self.consumer.acknowledge(message.delivery_tag)
            return

        db_log_handler = UowLogHandler(self.logger, uow.db_id)
        try:
            uow.state = unit_of_work.STATE_IN_PROGRESS
            uow.started_at = datetime.utcnow()
            self.uow_dao.update(uow)
            self.performance_tracker.start_uow(uow)

            if self.perform_db_logging:
                db_log_handler.attach()

            result = self._process_uow(uow)
            if result is None:
                self.logger.warning('method {0}._process_uow returned None. Assuming happy flow.'
                                    .format(self.__class__.__name__))
                number_of_aggregated_objects, target_state = 0, unit_of_work.STATE_PROCESSED
            else:
                number_of_aggregated_objects, target_state = result

            uow.number_of_aggregated_documents = number_of_aggregated_objects
            uow.number_of_processed_documents = self.performance_tracker.success_per_job
            uow.finished_at = datetime.utcnow()
            uow.state = target_state
            self.uow_dao.update(uow)

            if uow.is_finished:
                self.performance_tracker.finish_uow()
            else:
                self.performance_tracker.cancel_uow()

        except Exception as e:
            fresh_uow = self.uow_dao.get_one(mq_request.record_db_id)
            self.performance_tracker.cancel_uow()
            if fresh_uow.is_canceled:
                self.logger.warning('UOW {0} for {1}@{2} was likely marked by MX as SKIPPED. No UOW update performed.'
                                    .format(uow.db_id, uow.process_name, uow.timeperiod), exc_info=False)
            else:
                self.logger.error('Safety fuse while processing UOW {0} for {1}@{2}: {3}'
                                  .format(uow.db_id, uow.process_name, uow.timeperiod, e), exc_info=True)
                uow.state = unit_of_work.STATE_INVALID
                self.uow_dao.update(uow)

        finally:
            self.consumer.acknowledge(message.delivery_tag)
            self.consumer.close()
            self._clean_up()
            db_log_handler.detach()

        try:
            self.mq_transmitter.publish_uow_status(uow)
            self.logger.info('UOW *{0}* status report published into MQ'.format(uow.state))
        except Exception:
            self.logger.error('Error on UOW status report publishing', exc_info=True)

# 需要导入模块: from synergy.db.dao.unit_of_work_dao import UnitOfWorkDao [as 别名]
# 或者: from synergy.db.dao.unit_of_work_dao.UnitOfWorkDao import update [as 别名]
class BashRunnable(threading.Thread):
    """Process starts remote or local bash script job, supervises its execution and updates mq"""

    def __init__(self, logger, message, consumer, performance_ticker):
        self.logger = logger
        self.message = message
        self.mq_request = SynergyMqTransmission.from_json(message.body)
        self.consumer = consumer
        self.performance_ticker = performance_ticker
        self.alive = False
        self.return_code = -1
        self.uow_dao = UnitOfWorkDao(self.logger)

        self.thread_name = str(self.mq_request)
        super(BashRunnable, self).__init__(name=self.thread_name)

    def _poll_process(self):
        return self.alive, self.return_code

    def _start_process(self):
        try:
            uow = self.uow_dao.get_one(self.mq_request.unit_of_work_id)
            if not uow.is_requested:
                # accept only UOW in STATE_REQUESTED
                self.logger.warn('Skipping unit_of_work: id %s; state %s;' % (str(self.message.body), uow.state),
                                 exc_info=False)
                self.consumer.acknowledge(self.message.delivery_tag)
                return
        except Exception:
            self.logger.error('Safety fuse. Can not identify unit_of_work %s' % str(self.message.body), exc_info=True)
            self.consumer.acknowledge(self.message.delivery_tag)
            return

        try:
            self.logger.info('start: %s {' % self.thread_name)
            self.alive = True

            uow.state = unit_of_work.STATE_IN_PROGRESS
            uow.started_at = datetime.utcnow()
            self.uow_dao.update(uow)

            fabric.operations.env.warn_only = True
            fabric.operations.env.abort_on_prompts = True
            fabric.operations.env.use_ssh_config = True
            fabric.operations.env.host_string = uow.arguments[ARGUMENT_CMD_HOST]

            command = os.path.join(uow.arguments[ARGUMENT_CMD_PATH],
                                   uow.arguments[ARGUMENT_CMD_FILE])
            command += ' %s' % uow.arguments[ARGUMENT_CMD_ARGS]

            run_result = fabric.operations.run(command, pty=False)
            if run_result.succeeded:
                self.return_code = 0

            uow.finished_at = datetime.utcnow()
            uow.state = unit_of_work.STATE_PROCESSED
            self.uow_dao.update(uow)

            self.logger.info('Completed %s with result = %r' % (self.thread_name, self.return_code))
        except Exception:
            self.logger.error('Exception on starting: %s' % self.thread_name, exc_info=True)
            uow.state = unit_of_work.STATE_INVALID
            self.uow_dao.update(uow)
        finally:
            self.logger.info('}')
            self.alive = False

    def run(self):
        try:
            self._start_process()
            code = None
            alive = True
            while alive:
                alive, code = self._poll_process()
                time.sleep(0.1)

            if code == 0:
                self.performance_ticker.tracker.increment_success()
            else:
                self.performance_ticker.tracker.increment_failure()

            self.logger.info('BashDriver for %s return code is %r' % (self.thread_name, code))
        except Exception as e:
            self.performance_ticker.tracker.increment_failure()
            self.logger.error('Safety fuse while processing request %r: %r' % (self.message.body, e), exc_info=True)
        finally:
            self.consumer.acknowledge(self.message.delivery_tag)

# 需要导入模块: from synergy.db.dao.unit_of_work_dao import UnitOfWorkDao [as 别名]
# 或者: from synergy.db.dao.unit_of_work_dao.UnitOfWorkDao import update [as 别名]
class AbstractStateMachine(object):
    """ Abstract state machine used to host common logic for the rest of registered State Machines """

    def __init__(self, logger, timetable, name):
        self.name = name
        self.logger = logger
        self.mq_transmitter = MqTransmitter(self.logger)
        self.timetable = timetable
        self.uow_dao = UnitOfWorkDao(self.logger)
        self.job_dao = JobDao(self.logger)

    @property
    def run_on_active_timeperiod(self):
        """
        :return: True if given State Machine allows execution on *live* timeperiod, as opposed to the finished one
        """
        raise NotImplementedError('property run_on_active_timeperiod must be implemented by {0}'
                                  .format(self.__class__.__name__))

    def _log_message(self, level, process_name, timeperiod, msg):
        """ method performs logging into log file and Timetable's tree node"""
        self.timetable.add_log_entry(process_name, timeperiod, msg)
        self.logger.log(level, msg)

    @with_reconnect
    def _insert_uow(self, process_name, timeperiod, start_timeperiod, end_timeperiod, start_id, end_id):
        """creates unit_of_work and inserts it into the DB
            :raise DuplicateKeyError: if unit_of_work with given parameters already exists """
        uow = UnitOfWork()
        uow.process_name = process_name
        uow.timeperiod = timeperiod
        uow.start_id = str(start_id)
        uow.end_id = str(end_id)
        uow.start_timeperiod = start_timeperiod
        uow.end_timeperiod = end_timeperiod
        uow.created_at = datetime.utcnow()
        uow.submitted_at = datetime.utcnow()
        uow.source = context.process_context[process_name].source
        uow.sink = context.process_context[process_name].sink
        uow.state = unit_of_work.STATE_REQUESTED
        uow.unit_of_work_type = unit_of_work.TYPE_MANAGED
        uow.number_of_retries = 0
        uow.arguments = context.process_context[process_name].arguments
        uow.db_id = self.uow_dao.insert(uow)

        msg = 'Created: UOW {0} for {1}@{2}.'.format(uow.db_id, process_name, start_timeperiod)
        self._log_message(INFO, process_name, start_timeperiod, msg)
        return uow

    def _publish_uow(self, uow):
        self.mq_transmitter.publish_managed_uow(uow)
        msg = 'Published: UOW {0} for {1}@{2}.'.format(uow.db_id, uow.process_name, uow.start_timeperiod)
        self._log_message(INFO, uow.process_name, uow.start_timeperiod, msg)

    def insert_and_publish_uow(self, job_record, start_id, end_id):
        """ method creates and publishes a unit_of_work. it also handles DuplicateKeyError and attempts recovery
        :return: tuple (uow, is_duplicate)
        :raise UserWarning: if the recovery from DuplicateKeyError was unsuccessful
        """
        process_name = job_record.process_name
        timeperiod = job_record.timeperiod
        start_timeperiod = self.compute_start_timeperiod(job_record.process_name, job_record.timeperiod)
        end_timeperiod = self.compute_end_timeperiod(job_record.process_name, job_record.timeperiod)

        try:
            is_duplicate = False
            uow = self._insert_uow(process_name, timeperiod, start_timeperiod, end_timeperiod, start_id, end_id)
        except DuplicateKeyError as e:
            is_duplicate = True
            msg = 'Catching up with latest UOW {0}@{1}, because of: {2}' \
                  .format(process_name, start_timeperiod, e)
            self._log_message(WARNING, process_name, start_timeperiod, msg)
            uow = self.uow_dao.recover_from_duplicatekeyerror(e)

        if not uow:
            msg = 'PERSISTENT TIER ERROR! Unable to locate UOW for {0}@{1}' \
                  .format(process_name, start_timeperiod)
            self._log_message(WARNING, process_name, start_timeperiod, msg)
            raise UserWarning(msg)

        if uow.is_canceled:
            # this UOW was marked for re-processing. recycle it
            uow.created_at = datetime.utcnow()      # reset created_at to bypass GC cancellation logic
            uow.submitted_at = datetime.utcnow()    # reset submitted_at to allow 1 hour free of GC resubmitting
            del uow.started_at
            del uow.finished_at
            del uow.number_of_aggregated_documents
            del uow.number_of_processed_documents
            uow.state = unit_of_work.STATE_REQUESTED
            self.uow_dao.update(uow)

        # publish the created/recovered/recycled unit_of_work
        self._publish_uow(uow)
        return uow, is_duplicate

    def notify(self, uow):
        """ method is used by StateMachine's users to provide notifications about change in UOW
        if applicable, method will update job_record state and Timetable tree node state
        :assumptions: uow is in [STATE_NOOP, STATE_CANCELED, STATE_PROCESSED] """
        pass
#.........这里部分代码省略.........

# 需要导入模块: from synergy.db.dao.unit_of_work_dao import UnitOfWorkDao [as 别名]
# 或者: from synergy.db.dao.unit_of_work_dao.UnitOfWorkDao import update [as 别名]
class Timetable(object):
    """ Timetable holds all known process trees, where every node presents a timeperiod-driven job"""

    def __init__(self, logger):
        self.lock = RLock()
        self.logger = logger
        self.uow_dao = UnitOfWorkDao(self.logger)
        self.job_dao = JobDao(self.logger)
        self.reprocess = dict()

        # self.trees contain all of the trees and manages much of their life cycle
        # remember to enlist there all trees the system is working with
        self.trees = self._construct_trees_from_context()

        self._register_callbacks()
        self._register_dependencies()
        self.load_tree()
        self.build_trees()
        self.validate()

    def _construct_trees_from_context(self):
        trees = dict()
        for tree_name, context_entry in context.timetable_context.items():
            tree = MultiLevelTree(process_names=context_entry.enclosed_processes,
                                  tree_name=tree_name,
                                  mx_name=context_entry.mx_name,
                                  mx_page=context_entry.mx_page)
            trees[tree_name] = tree
        return trees

    def _register_dependencies(self):
        """ register dependencies between trees"""
        for tree_name, context_entry in context.timetable_context.items():
            tree = self.trees[tree_name]
            assert isinstance(tree, MultiLevelTree)
            for dependent_on in context_entry.dependent_on:
                dependent_on_tree = self.trees[dependent_on]
                assert isinstance(dependent_on_tree, MultiLevelTree)
                tree.register_dependent_on(dependent_on_tree)

    def _register_callbacks(self):
        """ register logic that reacts on reprocessing request
        and create embryo timetable record request"""

        # reprocessing request
        for tree_name, tree in self.trees.items():
            tree.register_reprocess_callback(self._callback_reprocess)

        # skip request
        for tree_name, tree in self.trees.items():
            tree.register_skip_callback(self._callback_skip)

        # callbacks register
        for tree_name, tree in self.trees.items():
            tree.register_create_callbacks(self._callback_create_job_record)

    # *** Call-back methods ***
    def _find_dependant_trees(self, tree_obj):
        """ returns list of trees that are dependent_on given tree_obj """
        dependant_trees = []
        for tree_name, tree in self.trees.items():
            if tree_obj in tree.dependent_on:
                dependant_trees.append(tree)
        return dependant_trees

    def _find_dependant_tree_nodes(self, node_a):
        dependant_nodes = set()
        for tree_b in self._find_dependant_trees(node_a.tree):
            node_b = node_a.find_counterpart_in(tree_b)
            if node_b is None:
                continue
            dependant_nodes.add(node_b)
        return dependant_nodes

    def _reprocess_single_tree_node(self, tree_node):
        """ is called from tree to answer reprocessing request.
        It is possible that timetable record will be transferred to STATE_IN_PROGRESS with no related unit_of_work"""
        uow_id = tree_node.job_record.related_unit_of_work
        if uow_id is not None:
            tree_node.job_record.state = job.STATE_IN_PROGRESS
            uow = self.uow_dao.get_one(uow_id)
            uow.state = unit_of_work.STATE_INVALID
            uow.number_of_retries = 0
            uow.created_at = datetime.utcnow()
            self.uow_dao.update(uow)
            msg = 'Transferred job record %s for %s in timeperiod %s to %s; Transferred unit_of_work to %s' \
                  % (tree_node.job_record.db_id,
                     tree_node.process_name,
                     tree_node.job_record.timeperiod,
                     tree_node.job_record.state,
                     uow.state)

            if tree_node.process_name not in self.reprocess:
                self.reprocess[tree_node.process_name] = dict()
            self.reprocess[tree_node.process_name][tree_node.timeperiod] = tree_node

        else:
            tree_node.job_record.state = job.STATE_EMBRYO
            msg = 'Transferred job record %s for %s in timeperiod %s to %s;' \
                  % (tree_node.job_record.db_id,
#.........这里部分代码省略.........

# 需要导入模块: from synergy.db.dao.unit_of_work_dao import UnitOfWorkDao [as 别名]
# 或者: from synergy.db.dao.unit_of_work_dao.UnitOfWorkDao import update [as 别名]
class StateMachineFreerun(object):
    """ State Machine to handle freerun jobs/triggers """

    def __init__(self, logger, name=STATE_MACHINE_FREERUN):
        self.name = name
        self.logger = logger
        self.mq_transmitter = MqTransmitter(self.logger)
        self.uow_dao = UnitOfWorkDao(self.logger)
        self.sfe_dao = FreerunProcessDao(self.logger)

    @with_reconnect
    def _log_message(self, level, freerun_entry, msg):
        """ method performs logging into log file and the freerun_entry """
        self.logger.log(level, msg)

        assert isinstance(freerun_entry, FreerunProcessEntry)
        event_log = freerun_entry.event_log
        if len(event_log) > MAX_NUMBER_OF_EVENTS:
            del event_log[-1]
        event_log.insert(0, msg)
        self.sfe_dao.update(freerun_entry)

    @with_reconnect
    def _insert_uow(self, freerun_entry):
        """ creates unit_of_work and inserts it into the DB
            :raise DuplicateKeyError: if unit_of_work with given parameters already exists """
        current_timeperiod = time_helper.actual_timeperiod(QUALIFIER_REAL_TIME)
        process_entry = context.process_context[freerun_entry.process_name]

        uow = UnitOfWork()
        uow.process_name = freerun_entry.schedulable_name
        uow.timeperiod = current_timeperiod
        uow.start_id = 0
        uow.end_id = 0
        uow.start_timeperiod = current_timeperiod
        uow.end_timeperiod = current_timeperiod
        uow.created_at = datetime.utcnow()
        uow.submitted_at = datetime.utcnow()
        uow.source = process_entry.source if hasattr(process_entry, 'source') else None
        uow.sink = process_entry.sink if hasattr(process_entry, 'sink') else None
        uow.state = unit_of_work.STATE_REQUESTED
        uow.unit_of_work_type = unit_of_work.TYPE_FREERUN
        uow.number_of_retries = 0
        uow.arguments = freerun_entry.arguments
        uow.db_id = self.uow_dao.insert(uow)

        msg = 'Created: UOW {0} for {1}@{2}.' \
              .format(uow.db_id, freerun_entry.schedulable_name, current_timeperiod)
        self._log_message(INFO, freerun_entry, msg)
        return uow

    def _publish_uow(self, freerun_entry, uow):
        self.mq_transmitter.publish_freerun_uow(freerun_entry, uow)
        msg = 'Published: UOW {0} for {1}.'.format(uow.db_id, freerun_entry.schedulable_name)
        self._log_message(INFO, freerun_entry, msg)

    def insert_and_publish_uow(self, freerun_entry):
        try:
            uow = self._insert_uow(freerun_entry)
        except DuplicateKeyError as e:
            msg = 'Duplication of UOW found for {0}. Error msg: {1}'.format(freerun_entry.schedulable_name, e)
            self._log_message(WARNING, freerun_entry, msg)
            uow = self.uow_dao.recover_from_duplicatekeyerror(e)

        if uow is not None:
            # publish the created/caught up unit_of_work
            self._publish_uow(freerun_entry, uow)
            freerun_entry.related_unit_of_work = uow.db_id
            self.sfe_dao.update(freerun_entry)
        else:
            msg = 'PERSISTENT TIER ERROR! Unable to locate UOW for {0}' \
                  .format(freerun_entry.schedulable_name)
            self._log_message(WARNING, freerun_entry, msg)

    def _process_state_embryo(self, freerun_entry):
        """ method creates unit_of_work and associates it with the FreerunProcessEntry """
        self.insert_and_publish_uow(freerun_entry)

    def _process_state_in_progress(self, freerun_entry, uow):
        """ method that takes care of processing unit_of_work records in STATE_REQUESTED or STATE_IN_PROGRESS states"""
        self._publish_uow(freerun_entry, uow)

    def _process_terminal_state(self, freerun_entry, uow):
        """ method that takes care of processing unit_of_work records in
            STATE_PROCESSED, STATE_NOOP, STATE_INVALID, STATE_CANCELED states"""
        msg = 'UOW for {0} found in state {1}.'.format(freerun_entry.schedulable_name, uow.state)
        self._log_message(INFO, freerun_entry, msg)
        self.insert_and_publish_uow(freerun_entry)

    def manage_schedulable(self, freerun_entry):
        """ method main duty - is to _avoid_ publishing another unit_of_work, if previous was not yet processed
        In case the Scheduler sees that the unit_of_work is pending it will fire another WorkerMqRequest """

        assert isinstance(freerun_entry, FreerunProcessEntry)
        if freerun_entry.related_unit_of_work is None:
            uow = None
        else:
            uow = self.uow_dao.get_one(freerun_entry.related_unit_of_work)

        try:
#.........这里部分代码省略.........

# 需要导入模块: from synergy.db.dao.unit_of_work_dao import UnitOfWorkDao [as 别名]
# 或者: from synergy.db.dao.unit_of_work_dao.UnitOfWorkDao import update [as 别名]

#.........这里部分代码省略.........

        current_timestamp = compute_release_time(lag_in_minutes=0)
        for _ in range(len(q)):
            entry = q.pop()
            assert isinstance(entry, PriorityEntry)

            if ignore_priority or entry.release_time < current_timestamp:
                self._resubmit_uow(entry.entry)
            else:
                q.put(entry)
                break

    @thread_safe
    def flush(self, ignore_priority=False):
        """ method iterates over each reprocessing queues and re-submits UOW whose waiting time has expired """
        for process_name, q in self.reprocess_uows.items():
            self._flush_queue(q, ignore_priority)

    @thread_safe
    def validate(self):
        """ method iterates over the reprocessing queue and synchronizes state of every UOW with the DB
            should it change via the MX to STATE_CANCELED - remove the UOW from the queue """
        for process_name, q in self.reprocess_uows.items():
            if not q:
                continue

            invalid_entries = list()
            for entry in q.queue:
                assert isinstance(entry, PriorityEntry)
                uow = self.uow_dao.get_one(entry.entry.db_id)
                if uow.is_canceled:
                    invalid_entries.append(entry)

            for entry in invalid_entries:
                q.queue.remove(entry)

        self.logger.info('reprocessing queue validated')

    @thread_safe
    def flush_one(self, process_name, ignore_priority=False):
        """ method iterates over the reprocessing queue for the given process
            and re-submits UOW whose waiting time has expired """
        q = self.reprocess_uows[process_name]
        self._flush_queue(q, ignore_priority)

    def _resubmit_uow(self, uow):
        # re-read UOW from the DB, in case it was STATE_CANCELLED by MX
        uow = self.uow_dao.get_one(uow.db_id)
        if uow.is_canceled:
            self.logger.info('suppressed re-submission of UOW {0} for {1}@{2} in {3};'
                             .format(uow.db_id, uow.process_name, uow.timeperiod, uow.state))
            return

        if uow.is_invalid:
            uow.number_of_retries += 1

        uow.state = unit_of_work.STATE_REQUESTED
        uow.submitted_at = datetime.utcnow()
        self.uow_dao.update(uow)

        self.mq_transmitter.publish_managed_uow(uow)
        self.logger.info('re-submitted UOW {0} for {1}@{2}; attempt {3}'
                         .format(uow.db_id, uow.process_name, uow.timeperiod, uow.number_of_retries))

    def _cancel_uow(self, uow):
        uow.state = unit_of_work.STATE_CANCELED
        self.uow_dao.update(uow)

        self.mq_transmitter.publish_uow_status(uow)
        self.logger.info('canceled UOW {0} for {1}@{2}; attempt {3}; created at {4}'
                         .format(uow.db_id, uow.process_name, uow.timeperiod, uow.number_of_retries, uow.created_at))

    def _run(self):
        try:
            self.logger.info('run {')

            self.logger.debug('step 1: validate existing queue entries')
            self.validate()

            self.logger.debug('step 2: scan reprocessing candidates')
            self.scan_uow_candidates()

            self.logger.debug('step 3: repost after timeout')
            self.flush()

            self.logger.debug('step 4: timetable housekeeping')
            self.timetable.build_trees()

            self.logger.debug('step 5: timetable validation')
            self.timetable.validate()
        except Exception as e:
            self.logger.error('GC run exception: {0}'.format(e))
        finally:
            self.logger.info('}')

    def start(self):
        self.timer.start()

    def stop(self):
        self.timer.cancel()

# 需要导入模块: from synergy.db.dao.unit_of_work_dao import UnitOfWorkDao [as 别名]
# 或者: from synergy.db.dao.unit_of_work_dao.UnitOfWorkDao import update [as 别名]
class GarbageCollectorWorker(AbstractMqWorker):
    """ GC is triggered by an empty message from RabbitMQ. It scans for invalid or stalled unit_of_work
    and re-triggers them. GC is vital for the health of the system.
    Deployment with no running GC is considered invalid """

    def __init__(self, process_name):
        super(GarbageCollectorWorker, self).__init__(process_name)
        self.lock = Lock()
        self.publishers = PublishersPool(self.logger)
        self.uow_dao = UnitOfWorkDao(self.logger)
        self.managed_dao = ManagedProcessDao(self.logger)
        self.managed_entries = dict()

    def __del__(self):
        try:
            self.logger.info('Closing Flopsy Publishers Pool...')
            self.publishers.close()
        except Exception as e:
            self.logger.error('Exception caught while closing Flopsy Publishers Pool: %s' % str(e))

        super(GarbageCollectorWorker, self).__del__()

    @thread_safe
    def _mq_callback(self, message):
        """ method looks for stale or invalid units of work re-runs them if needed"""
        try:
            managed_entries = self.managed_dao.get_all()
            self._update_managed_entries(managed_entries)

            since = settings.settings['synergy_start_timeperiod']
            uow_list = self.uow_dao.get_reprocessing_candidates(since)
            for uow in uow_list:
                if uow.process_name not in self.managed_entries:
                    self.logger.debug('Process %r is not known to the Synergy Scheduler. Skipping its unit_of_work.'
                                      % uow.process_name)
                    continue

                process_entry = self.managed_entries[uow.process_name]
                assert isinstance(process_entry, ManagedProcessEntry)
                if not process_entry.is_on:
                    self.logger.debug('Process %r is inactive at the Synergy Scheduler. Skipping its unit_of_work.'
                                      % uow.process_name)
                    continue

                self._process_single_document(uow)

        except LookupError as e:
            self.logger.info('Normal behaviour. %r' % e)
        except Exception as e:
            self.logger.error('_mq_callback: %s' % str(e), exc_info=True)
        finally:
            self.consumer.acknowledge(message.delivery_tag)

    def _update_managed_entries(self, managed_entries):
        self.managed_entries = {me.process_name: me for me in managed_entries}

    def _process_single_document(self, uow):
        """ actually inspects UOW retrieved from the database"""
        repost = False
        if uow.is_invalid:
            repost = True

        elif uow.is_in_progress or uow.is_requested:
            last_activity = uow.started_at
            if last_activity is None:
                last_activity = uow.created_at

            if datetime.utcnow() - last_activity > timedelta(hours=REPOST_AFTER_HOURS):
                repost = True

        if repost:
            mq_request = SynergyMqTransmission(process_name=uow.process_name,
                                               unit_of_work_id=uow.db_id)

            if datetime.utcnow() - uow.created_at < timedelta(hours=LIFE_SUPPORT_HOURS):
                uow.state = unit_of_work.STATE_REQUESTED
                uow.number_of_retries += 1
                self.uow_dao.update(uow)

                publisher = self.publishers.get(uow.process_name)
                publisher.publish(mq_request.document)
                publisher.release()

                self.logger.info('UOW marked for re-processing: process %s; timeperiod %s; id %s; attempt %d'
                                 % (uow.process_name, uow.timeperiod, uow.db_id, uow.number_of_retries))
                self.performance_ticker.tracker.increment_success()
            else:
                uow.state = unit_of_work.STATE_CANCELED
                self.uow_dao.update(uow)

                publisher = self.publishers.get(QUEUE_UOW_REPORT)
                publisher.publish(mq_request.document)
                publisher.release()

                self.logger.info('UOW transferred to STATE_CANCELED: process %s; timeperiod %s; id %s; attempt %d'
                                 % (uow.process_name, uow.timeperiod, uow.db_id, uow.number_of_retries))

# 需要导入模块: from synergy.db.dao.unit_of_work_dao import UnitOfWorkDao [as 别名]
# 或者: from synergy.db.dao.unit_of_work_dao.UnitOfWorkDao import update [as 别名]
class AbstractUowAwareWorker(AbstractMqWorker):
    """ Abstract class is inherited by all workers/aggregators
    that are aware of unit_of_work and capable of processing it"""

    def __init__(self, process_name):
        super(AbstractUowAwareWorker, self).__init__(process_name)
        self.uow_dao = UnitOfWorkDao(self.logger)
        self.publishers = PublishersPool(self.logger)

    def __del__(self):
        self.publishers.close()
        super(AbstractUowAwareWorker, self).__del__()

    # **************** Abstract Methods ************************
    def _init_performance_ticker(self, logger):
        self.performance_ticker = UowAwareTracker(logger)
        self.performance_ticker.start()

    def _process_uow(self, uow):
        """
        :param uow: unit_of_work to process
        :return: a tuple (number of processed items/documents/etc, desired unit_of_work state) or None
        if None is returned then it is assumed that the return tuple is (0, unit_of_work.STATE_PROCESSED)
        :raise an Exception if the UOW shall be marked as STATE_INVALID
        """
        raise NotImplementedError('_process_uow must be overridden in the child class %s' % self.__class__.__name__)

    def _clean_up(self):
        """ method is called from the *finally* clause and is suppose to clean up after the uow processing """
        pass

    def _mq_callback(self, message):
        try:
            mq_request = SynergyMqTransmission.from_json(message.body)
            uow = self.uow_dao.get_one(mq_request.unit_of_work_id)
            if not uow.is_requested:
                # accept only UOW in STATE_REQUESTED
                self.logger.warn('Skipping unit_of_work: id %s; state %s;' % (str(message.body), uow.state),
                                 exc_info=False)
                self.consumer.acknowledge(message.delivery_tag)
                return
        except Exception:
            self.logger.error('Safety fuse. Can not identify unit_of_work %s' % str(message.body), exc_info=True)
            self.consumer.acknowledge(message.delivery_tag)
            return

        try:
            uow.state = unit_of_work.STATE_IN_PROGRESS
            uow.started_at = datetime.utcnow()
            self.uow_dao.update(uow)
            self.performance_ticker.start_uow(uow)

            result = self._process_uow(uow)
            if result is None or isinstance(result, str):
                number_of_aggregated_objects, target_state = 0, unit_of_work.STATE_PROCESSED
            else:
                number_of_aggregated_objects, target_state = result

            uow.number_of_aggregated_documents = number_of_aggregated_objects
            uow.number_of_processed_documents = self.performance_ticker.per_job
            uow.finished_at = datetime.utcnow()
            uow.state = target_state
            self.uow_dao.update(uow)
            self.performance_ticker.finish_uow()
        except Exception as e:
            fresh_uow = self.uow_dao.get_one(mq_request.unit_of_work_id)
            self.performance_ticker.cancel_uow()
            if fresh_uow.is_canceled:
                self.logger.warn('unit_of_work: id %s was likely marked by MX as SKIPPED. '
                                 'No unit_of_work update is performed.' % str(message.body),
                                 exc_info=False)
            else:
                self.logger.error('Safety fuse while processing unit_of_work %s in timeperiod %s : %r'
                                  % (message.body, uow.timeperiod, e), exc_info=True)
                uow.state = unit_of_work.STATE_INVALID
                self.uow_dao.update(uow)

        finally:
            self.consumer.acknowledge(message.delivery_tag)
            self.consumer.close()
            self._clean_up()

        try:
            publisher = self.publishers.get(QUEUE_UOW_REPORT)
            publisher.publish(mq_request.document)
            publisher.release()
            self.logger.info('Published unit_of_work status report into %s queue' % QUEUE_UOW_REPORT)
        except Exception:
            self.logger.error('Error on unit_of_work status report publishing', exc_info=True)

# 需要导入模块: from synergy.db.dao.unit_of_work_dao import UnitOfWorkDao [as 别名]
# 或者: from synergy.db.dao.unit_of_work_dao.UnitOfWorkDao import update [as 别名]
class FreerunActionHandler(AbstractActionHandler):
    def __init__(self, request, **values):
        super(FreerunActionHandler, self).__init__(request, **values)
        self.process_name = self.request_arguments.get('process_name')
        self.entry_name = self.request_arguments.get('entry_name')
        self.freerun_process_dao = FreerunProcessDao(self.logger)
        self.uow_dao = UnitOfWorkDao(self.logger)
        self.is_request_valid = True if self.process_name and self.entry_name else False

        if self.is_request_valid:
            self.process_name = self.process_name.strip()
            self.entry_name = self.entry_name.strip()
            self.is_requested_state_on = 'is_on' in self.request_arguments and self.request_arguments['is_on']

    @AbstractActionHandler.thread_handler.getter
    def thread_handler(self):
        handler_key = (self.process_name, self.entry_name)
        return self.scheduler.freerun_handlers[handler_key]

    @AbstractActionHandler.process_entry.getter
    def process_entry(self):
        return self.thread_handler.process_entry

    @valid_action_request
    def action_cancel_uow(self):
        uow_id = self.process_entry.related_unit_of_work
        if uow_id is None:
            resp = {'response': 'no related unit_of_work'}
        else:
            uow = self.uow_dao.get_one(uow_id)
            uow.state = unit_of_work.STATE_CANCELED
            self.uow_dao.update(uow)
            resp = {'response': 'updated unit_of_work %r' % uow_id}
        return resp

    @valid_action_request
    def action_get_uow(self):
        uow_id = self.process_entry.related_unit_of_work
        if uow_id is None:
            resp = {'response': 'no related unit_of_work'}
        else:
            resp = self.uow_dao.get_one(uow_id).document
            for key in resp:
                resp[key] = str(resp[key])
        return resp

    @valid_action_request
    def action_get_log(self):
        return {'log': self.process_entry.log}

    @valid_action_request
    def action_update_entry(self):
        if 'insert_button' in self.request_arguments:
            process_entry = FreerunProcessEntry()
            process_entry.process_name = self.process_name
            process_entry.entry_name = self.entry_name

            if self.request_arguments['arguments']:
                arguments = self.request_arguments['arguments'].decode('unicode-escape')
                process_entry.arguments = json.loads(arguments)
            else:
                process_entry.arguments = {}

            process_entry.description = self.request_arguments['description']
            process_entry.is_on = self.is_requested_state_on
            process_entry.trigger_frequency = self.request_arguments['trigger_frequency']
            self.freerun_process_dao.update(process_entry)

            self.scheduler._register_process_entry(process_entry, self.scheduler.fire_freerun_worker)

        elif 'update_button' in self.request_arguments:
            is_interval_changed = self.process_entry.trigger_frequency != self.request_arguments['trigger_frequency']

            if self.request_arguments['arguments']:
                arguments = self.request_arguments['arguments'].decode('unicode-escape')
                self.process_entry.arguments = json.loads(arguments)
            else:
                self.process_entry.arguments = {}

            self.process_entry.description = self.request_arguments['description']
            self.process_entry.is_on = self.is_requested_state_on
            self.process_entry.trigger_frequency = self.request_arguments['trigger_frequency']
            self.freerun_process_dao.update(self.process_entry)

            if is_interval_changed:
                self.action_change_interval()

            if self.process_entry.is_on != self.is_requested_state_on:
                if self.is_requested_state_on:
                    self.action_activate_trigger()
                else:
                    self.action_deactivate_trigger()

        elif 'delete_button' in self.request_arguments:
            handler_key = (self.process_name, self.entry_name)
            self.thread_handler.deactivate()
            self.freerun_process_dao.remove(handler_key)
            del self.scheduler.freerun_handlers[handler_key]

        elif 'cancel_button' in self.request_arguments:
#.........这里部分代码省略.........