def setup_environment():
parser = ArgumentParser(description="Frontera strategy worker.")
parser.add_argument('--config', type=str, required=True,
help='Settings module name, should be accessible by import')
parser.add_argument('--log-level', '-L', type=str, default='INFO',
help="Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL")
parser.add_argument('--strategy', type=str,
help='Crawling strategy class path')
parser.add_argument('--partition-id', type=int,
help="Instance partition id.")
args = parser.parse_args()
settings = Settings(module=args.config)
strategy_classpath = args.strategy if args.strategy else settings.get('CRAWLING_STRATEGY')
if not strategy_classpath:
raise ValueError("Couldn't locate strategy class path. Please supply it either using command line option or "
"settings file.")
strategy_class = load_object(strategy_classpath)
partition_id = args.partition_id if args.partition_id is not None else settings.get('SCORING_PARTITION_ID')
if partition_id >= settings.get('SPIDER_LOG_PARTITIONS') or partition_id < 0:
raise ValueError("Partition id (%d) cannot be less than zero or more than SPIDER_LOG_PARTITIONS." %
partition_id)
settings.set('SCORING_PARTITION_ID', partition_id)
logging_config_path = settings.get("LOGGING_CONFIG")
if logging_config_path and exists(logging_config_path):
fileConfig(logging_config_path)
else:
logging.basicConfig(level=args.log_level)
logger.setLevel(args.log_level)
logger.addHandler(CONSOLE)
return settings, strategy_class
def setup_environment():
parser = ArgumentParser(description="Frontera strategy worker.")
parser.add_argument('--config', type=str, required=True,
help='Settings module name, should be accessible by import')
parser.add_argument('--log-level', '-L', type=str, default='INFO',
help="Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL")
parser.add_argument('--strategy', type=str,
help='Crawling strategy class path')
parser.add_argument('--partition-id', type=int,
help="Instance partition id.")
args = parser.parse_args()
settings = Settings(module=args.config)
strategy_classpath = args.strategy if args.strategy else settings.get('CRAWLING_STRATEGY')
if not strategy_classpath:
raise ValueError("Couldn't locate strategy class path. Please supply it either using command line option or "
"settings file.")
strategy_class = load_object(strategy_classpath)
partition_id = args.partition_id if args.partition_id is not None else settings.get('SCORING_PARTITION_ID')
if partition_id >= settings.get('SPIDER_LOG_PARTITIONS') or partition_id < 0:
raise ValueError("Partition id (%d) cannot be less than zero or more than SPIDER_LOG_PARTITIONS." %
partition_id)
settings.set('SCORING_PARTITION_ID', partition_id)
logging_config_path = settings.get("LOGGING_CONFIG")
if logging_config_path and exists(logging_config_path):
fileConfig(logging_config_path)
else:
logging.basicConfig(level=args.log_level)
logger.setLevel(args.log_level)
logger.addHandler(CONSOLE)
return settings, strategy_class
def main():
parser = ArgumentParser(description="Crawl frontier worker.")
parser.add_argument(
'--config',
type=str,
help='Settings module name, should be accessible by import.')
parser.add_argument(
'--hostname',
type=str,
help='Hostname or IP address to bind. Default is 127.0.0.1')
parser.add_argument(
'--log-level',
'-L',
type=str,
default='INFO',
help=
'Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL. Default is INFO.')
parser.add_argument(
'--port',
type=int,
help=
'Base port number, server will bind to 6 ports starting from base. Default is 5550'
)
args = parser.parse_args()
settings = Settings(module=args.config)
hostname = args.hostname if args.hostname else settings.get("ZMQ_HOSTNAME")
port = args.port if args.port else settings.get("ZMQ_BASE_PORT")
server = Server(hostname, port)
server.logger.setLevel(args.log_level)
server.start()
def main():
"""
Parse arguments, set configuration values, then start the broker
"""
parser = ArgumentParser(description="Crawl frontier worker.")
parser.add_argument(
'--config',
type=str,
help='Settings module name, should be accessible by import.')
parser.add_argument(
'--address',
type=str,
help='Hostname, IP address or Wildcard * to bind. Default is 127.0.0.1'
'. When binding to wildcard it defaults to IPv4.')
parser.add_argument(
'--log-level',
'-L',
type=str,
default='INFO',
help='Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL. Default is'
' INFO.')
parser.add_argument(
'--port',
type=int,
help='Base port number, server will bind to 6 ports starting from base'
'. Default is 5550')
args = parser.parse_args()
settings = Settings(module=args.config)
address = args.address if args.address else settings.get("ZMQ_ADDRESS")
port = args.port if args.port else settings.get("ZMQ_BASE_PORT")
server = Server(address, port)
server.logger.setLevel(args.log_level)
server.start()
def main():
"""
Parse arguments, set configuration values, then start the broker
"""
parser = ArgumentParser(description="Crawl frontier worker.")
parser.add_argument(
'--config', type=str,
help='Settings module name, should be accessible by import.')
parser.add_argument(
'--address', type=str,
help='Hostname, IP address or Wildcard * to bind. Default is 127.0.0.1'
'. When binding to wildcard it defaults to IPv4.')
parser.add_argument(
'--log-level', '-L', type=str, default='INFO',
help='Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL. Default is'
' INFO.')
parser.add_argument(
'--port', type=int,
help='Base port number, server will bind to 6 ports starting from base'
'. Default is 5550')
args = parser.parse_args()
settings = Settings(module=args.config)
address = args.address if args.address else settings.get("ZMQ_ADDRESS")
port = args.port if args.port else settings.get("ZMQ_BASE_PORT")
server = Server(address, port)
server.logger.setLevel(args.log_level)
server.start()
def setup_environment():
parser = ArgumentParser(description="Frontera strategy worker.")
parser.add_argument('--config', type=str, required=True,
help='Settings module name, should be accessible by import')
parser.add_argument('--log-level', '-L', type=str, default='INFO',
help="Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL")
parser.add_argument('--strategy', type=str,
help='Crawling strategy class path')
parser.add_argument('--partition-id', type=int,
help="Instance partition id.")
parser.add_argument('--port', type=int, help="Json Rpc service port to listen.")
parser.add_argument('--args', '-a', nargs='*', type=str, help="Optional arguments for crawling strategy, "
"in a form of key=value separated with space")
parser.add_argument('--add-seeds', action='store_true', help="Run in add seeds mode. Worker finishes after running "
"of strategy add_seeds method")
parser.add_argument('--seeds-url', type=str, help="Seeds url. S3 and native urlopen schemas are currently "
"supported, implies add seeds run mode")
args = parser.parse_args()
settings = Settings(module=args.config)
strategy_classpath = args.strategy if args.strategy else settings.get('STRATEGY')
if not strategy_classpath:
raise ValueError("Couldn't locate strategy class path. Please supply it either using command line option or "
"settings file.")
settings.set('STRATEGY', strategy_classpath)
partition_id = args.partition_id if args.partition_id is not None else settings.get('SCORING_PARTITION_ID')
if partition_id >= settings.get('SPIDER_LOG_PARTITIONS') or partition_id < 0:
raise ValueError("Partition id (%d) cannot be less than zero or more than SPIDER_LOG_PARTITIONS." %
partition_id)
settings.set('SCORING_PARTITION_ID', partition_id)
if args.port:
settings.set('JSONRPC_PORT', args.port)
strategy_args = {}
if args.args:
for arg in args.args:
key, _, value = arg.partition("=")
strategy_args[key] = value if value else None
settings.set("STRATEGY_ARGS", strategy_args)
logging_config_path = settings.get("LOGGING_CONFIG")
if logging_config_path and exists(logging_config_path):
fileConfig(logging_config_path, disable_existing_loggers=False)
else:
logging.basicConfig(level=args.log_level)
logger.setLevel(args.log_level)
logger.addHandler(CONSOLE)
return settings, args.add_seeds, args.seeds_url
def __init__(self, manager):
self._manager = manager
settings = Settings(attributes=manager.settings.attributes)
messagebus = load_object(settings.get('MESSAGE_BUS'))
self.mb = messagebus(settings)
store_content = settings.get('STORE_CONTENT')
self._encoder = Encoder(manager.request_model, send_body=store_content)
self._decoder = Decoder(manager.request_model, manager.response_model)
self.spider_log_producer = self.mb.spider_log().producer()
spider_feed = self.mb.spider_feed()
self.partition_id = settings.get('SPIDER_PARTITION_ID')
self.consumer = spider_feed.consumer(partition_id=self.partition_id)
self._get_timeout = float(settings.get('KAFKA_GET_TIMEOUT'))
self._buffer = OverusedBuffer(self._get_next_requests,
manager.logger.manager.debug)
def __init__(self, manager):
self._manager = manager
settings = Settings(attributes=manager.settings.attributes)
messagebus = load_object(settings.get('MESSAGE_BUS'))
self.mb = messagebus(settings)
store_content = settings.get('STORE_CONTENT')
self._encoder = Encoder(manager.request_model, send_body=store_content)
self._decoder = Decoder(manager.request_model, manager.response_model)
self.spider_log_producer = self.mb.spider_log().producer()
spider_feed = self.mb.spider_feed()
self.partition_id = settings.get('SPIDER_PARTITION_ID')
self.consumer = spider_feed.consumer(partition_id=self.partition_id)
self._get_timeout = float(settings.get('KAFKA_GET_TIMEOUT'))
self._buffer = OverusedBuffer(self._get_next_requests,
manager.logger.manager.debug)
def main():
parser = ArgumentParser(description="Crawl frontier worker.")
parser.add_argument('--config', type=str,
help='Settings module name, should be accessible by import.')
parser.add_argument('--hostname', type=str,
help='Hostname or IP address to bind. Default is 127.0.0.1')
parser.add_argument('--log-level', '-L', type=str, default='INFO',
help='Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL. Default is INFO.')
parser.add_argument('--port', type=int,
help='Base port number, server will bind to 6 ports starting from base. Default is 5550')
args = parser.parse_args()
settings = Settings(module=args.config)
hostname = args.hostname if args.hostname else settings.get("ZMQ_HOSTNAME")
port = args.port if args.port else settings.get("ZMQ_BASE_PORT")
server = Server(hostname, port)
server.logger.setLevel(args.log_level)
server.start()
self.states.update_cache(request)
if __name__ == '__main__':
parser = ArgumentParser(description="Frontera strategy worker.")
parser.add_argument('--config', type=str, required=True,
help='Settings module name, should be accessible by import')
parser.add_argument('--log-level', '-L', type=str, default='INFO',
help="Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL")
parser.add_argument('--strategy', type=str,
help='Crawling strategy class path')
parser.add_argument('--partition-id', type=int,
help="Instance partition id.")
args = parser.parse_args()
settings = Settings(module=args.config)
strategy_classpath = args.strategy if args.strategy else settings.get('CRAWLING_STRATEGY')
if not strategy_classpath:
raise ValueError("Couldn't locate strategy class path. Please supply it either using command line option or "
"settings file.")
strategy_class = load_object(strategy_classpath)
partition_id = args.partition_id if args.partition_id is not None else settings.get('SCORING_PARTITION_ID')
if partition_id >= settings.get('SPIDER_LOG_PARTITIONS') or partition_id < 0:
raise ValueError("Partition id (%d) cannot be less than zero or more than SPIDER_LOG_PARTITIONS." %
partition_id)
settings.set('SCORING_PARTITION_ID', partition_id)
logging_config_path = settings.get("LOGGING_CONFIG")
if logging_config_path and exists(logging_config_path):
fileConfig(logging_config_path)
else:
help='Settings module name, should be accessible by import')
parser.add_argument(
'--log-level',
'-L',
type=str,
default='INFO',
help="Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL")
parser.add_argument('--strategy',
type=str,
help='Crawling strategy class path')
parser.add_argument('--partition-id',
type=int,
help="Instance partition id.")
args = parser.parse_args()
settings = Settings(module=args.config)
strategy_classpath = args.strategy if args.strategy else settings.get(
'CRAWLING_STRATEGY')
if not strategy_classpath:
raise ValueError(
"Couldn't locate strategy class path. Please supply it either using command line option or "
"settings file.")
strategy_class = load_object(strategy_classpath)
partition_id = args.partition_id if args.partition_id is not None else settings.get(
'SCORING_PARTITION_ID')
if partition_id >= settings.get(
'SPIDER_LOG_PARTITIONS') or partition_id < 0:
raise ValueError(
"Partition id (%d) cannot be less than zero or more than SPIDER_LOG_PARTITIONS."
% partition_id)
settings.set('SCORING_PARTITION_ID', partition_id)
def setup_environment():
parser = ArgumentParser(description="Frontera strategy worker.")
parser.add_argument(
'--config',
type=str,
required=True,
help='Settings module name, should be accessible by import')
parser.add_argument(
'--log-level',
'-L',
type=str,
default='INFO',
help="Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL")
parser.add_argument('--strategy',
type=str,
help='Crawling strategy class path')
parser.add_argument('--partition-id',
type=int,
help="Instance partition id.")
parser.add_argument('--port',
type=int,
help="Json Rpc service port to listen.")
parser.add_argument('--args',
'-a',
nargs='*',
type=str,
help="Optional arguments for crawling strategy, "
"in a form of key=value separated with space")
parser.add_argument(
'--add-seeds',
action='store_true',
help="Run in add seeds mode. Worker finishes after running "
"of strategy add_seeds method")
parser.add_argument(
'--seeds-url',
type=str,
help="Seeds url. S3 and native urlopen schemas are currently "
"supported, implies add seeds run mode")
args = parser.parse_args()
settings = Settings(module=args.config)
strategy_classpath = args.strategy if args.strategy else settings.get(
'STRATEGY')
if not strategy_classpath:
raise ValueError(
"Couldn't locate strategy class path. Please supply it either using command line option or "
"settings file.")
settings.set('STRATEGY', strategy_classpath)
partition_id = args.partition_id if args.partition_id is not None else settings.get(
'SCORING_PARTITION_ID')
if partition_id >= settings.get(
'SPIDER_LOG_PARTITIONS') or partition_id < 0:
raise ValueError(
"Partition id (%d) cannot be less than zero or more than SPIDER_LOG_PARTITIONS."
% partition_id)
settings.set('SCORING_PARTITION_ID', partition_id)
if args.port:
settings.set('JSONRPC_PORT', args.port)
strategy_args = {}
if args.args:
for arg in args.args:
key, _, value = arg.partition("=")
strategy_args[key] = value if value else None
settings.set("STRATEGY_ARGS", strategy_args)
logging_config_path = settings.get("LOGGING_CONFIG")
if logging_config_path and exists(logging_config_path):
fileConfig(logging_config_path, disable_existing_loggers=False)
else:
logging.basicConfig(level=args.log_level)
logger.setLevel(args.log_level)
logger.addHandler(CONSOLE)
return settings, args.add_seeds, args.seeds_url
parser.add_argument(
'--log-level',
'-L',
type=str,
default='INFO',
help="Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL.")
parser.add_argument('--port',
type=int,
help="Json Rpc service port to listen.")
args = parser.parse_args()
settings = Settings(module=args.config)
if args.port:
settings.set("JSONRPC_PORT", [args.port])
logging_config_path = settings.get("LOGGING_CONFIG")
if logging_config_path and os.path.exists(logging_config_path):
fileConfig(logging_config_path, disable_existing_loggers=False)
else:
logging.basicConfig(level=args.log_level)
logger.setLevel(args.log_level)
logger.addHandler(CONSOLE)
worker = DBWorker(settings,
args.no_batches,
args.no_incoming,
args.no_scoring,
partitions=args.partitions)
server = WorkerJsonRpcService(worker, settings)
server.start_listening()
worker.run()
def run_add_seeds(settings, seeds_file):
fh = open(seeds_file, "rb")
logger.info("Starting local seeds addition from file %s", seeds_file)
manager = LocalFrontierManager.from_settings(settings)
manager.add_seeds(fh)
manager.stop()
manager.close()
logger.info("Seeds addition finished")
if __name__ == '__main__':
parser = ArgumentParser(description="Frontera local add seeds utility")
parser.add_argument('--config', type=str, required=True,
help='Settings module name, should be accessible by import')
parser.add_argument('--log-level', '-L', type=str, default='INFO',
help="Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL")
parser.add_argument('--seeds-file', type=str, required=True, help="Seeds file path")
args = parser.parse_args()
settings = Settings(module=args.config)
logging_config_path = settings.get("LOGGING_CONFIG")
if logging_config_path and exists(logging_config_path):
fileConfig(logging_config_path, disable_existing_loggers=False)
else:
logging.basicConfig(level=args.log_level)
logger.setLevel(args.log_level)
logger.addHandler(CONSOLE)
run_add_seeds(settings, args.seeds_file)
if __name__ == '__main__':
parser = ArgumentParser(description="Frontera strategy worker.")
parser.add_argument(
'--config',
type=str,
required=True,
help='Settings module name, should be accessible by import')
parser.add_argument(
'--log-level',
'-L',
type=str,
default='INFO',
help="Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL")
parser.add_argument('--strategy',
type=str,
help='Crawling strategy class path')
args = parser.parse_args()
logger.setLevel(args.log_level)
logger.addHandler(CONSOLE)
settings = Settings(module=args.config)
strategy_classpath = args.strategy if args.strategy else settings.get(
'CRAWLING_STRATEGY')
if not strategy_classpath:
raise ValueError(
"Couldn't locate strategy class path. Please supply it either using command line option or "
"settings file.")
strategy_class = load_object(strategy_classpath)
worker = StrategyWorker(settings, strategy_class)
worker.run()
def test_fallsback_to_frontera_default_settings():
settings = Settings()
assert settings.get('MAX_NEXT_REQUESTS') == 64
def test_settings_on_a_python_module_are_loaded():
settings = Settings('tests.scrapy_spider.frontera.settings')
assert settings.get('MAX_REQUESTS') == 5
self.strategy.page_error(request, error)
self.states.update_cache(request)
if __name__ == '__main__':
parser = ArgumentParser(description="Frontera strategy worker.")
parser.add_argument('--config', type=str, required=True,
help='Settings module name, should be accessible by import')
parser.add_argument('--log-level', '-L', type=str, default='INFO',
help="Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL")
parser.add_argument('--strategy', type=str,
help='Crawling strategy class path')
args = parser.parse_args()
settings = Settings(module=args.config)
strategy_classpath = args.strategy if args.strategy else settings.get('CRAWLING_STRATEGY')
if not strategy_classpath:
raise ValueError("Couldn't locate strategy class path. Please supply it either using command line option or "
"settings file.")
strategy_class = load_object(strategy_classpath)
logging_config_path = settings.get("LOGGING_CONFIG")
if logging_config_path and exists(logging_config_path):
fileConfig(logging_config_path)
else:
logging.basicConfig(level=args.log_level)
logger.setLevel(args.log_level)
logger.addHandler(CONSOLE)
worker = StrategyWorker(settings, strategy_class)
worker.run()
def test_settings_passed_as_attributes_can_be_found():
settings = Settings(attributes={'SETTING': 'value'})
assert settings.get('SETTING') == 'value'
type=str,
required=True,
help='Settings module name, should be accessible by import')
parser.add_argument(
'--log-level',
'-L',
type=str,
default='INFO',
help="Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL")
parser.add_argument('--strategy',
type=str,
help='Crawling strategy class path')
args = parser.parse_args()
settings = Settings(module=args.config)
strategy_classpath = args.strategy if args.strategy else settings.get(
'CRAWLING_STRATEGY')
if not strategy_classpath:
raise ValueError(
"Couldn't locate strategy class path. Please supply it either using command line option or "
"settings file.")
strategy_class = load_object(strategy_classpath)
logging_config_path = settings.get("LOGGING_CONFIG")
if logging_config_path and exists(logging_config_path):
fileConfig(logging_config_path)
else:
logging.basicConfig(level=args.log_level)
logger.setLevel(args.log_level)
logger.addHandler(CONSOLE)
worker = StrategyWorker(settings, strategy_class)
worker.run()
self.states.update_cache(links)
self.states.update_cache(response)
def on_request_error(self, request, error):
logger.debug("Page error %s (%s)", request.url, error)
self.states.set_states(request)
self.strategy.page_error(request, error)
self.states.update_cache(request)
if __name__ == '__main__':
parser = ArgumentParser(description="Frontera strategy worker.")
parser.add_argument('--config', type=str, required=True,
help='Settings module name, should be accessible by import')
parser.add_argument('--log-level', '-L', type=str, default='INFO',
help="Log level, for ex. DEBUG, INFO, WARN, ERROR, FATAL")
parser.add_argument('--strategy', type=str,
help='Crawling strategy class path')
args = parser.parse_args()
logger.setLevel(args.log_level)
logger.addHandler(CONSOLE)
settings = Settings(module=args.config)
strategy_classpath = args.strategy if args.strategy else settings.get('CRAWLING_STRATEGY')
if not strategy_classpath:
raise ValueError("Couldn't locate strategy class path. Please supply it either using command line option or "
"settings file.")
strategy_class = load_object(strategy_classpath)
worker = StrategyWorker(settings, strategy_class)
worker.run()
def test_settings_on_a_python_module_are_loaded():
settings = Settings('frontera.tests.scrapy_spider.frontera.settings')
assert settings.get('MAX_REQUESTS') == 5
def test_fallsback_to_frontera_default_settings():
settings = Settings()
assert settings.get('MAX_NEXT_REQUESTS') == 0
def test_settings_passed_as_attributes_can_be_found():
settings = Settings(attributes={'SETTING': 'value'})
assert settings.get('SETTING') == 'value'
