def _parse(self):
while True:
task, body = ParserQueues.WAITING_PARSE.get()
cls = self._rules_updater.get_parse_model(task.platform,
task.feature)
if not cls:
fmt = 'Parse No Match: [P:{platform}][F:{feature}][K:{row_key}]'
TDDCLogging.warning(
fmt.format(platform=task.platform,
feature=task.feature,
row_key=task.row_key))
continue
try:
ret = cls(task, body)
except Exception, e:
TDDCLogging.error(e)
continue
self._storage(task, ret.items)
self._new_task_push(ret.tasks)
fmt = 'Parsed: [{platform}:{row_key}:{feature}][S:{items}][N:{tasks}]'
TDDCLogging.info(
fmt.format(platform=task.platform,
feature=task.feature,
row_key=task.row_key,
items=len(ret.items),
tasks=len(ret.tasks)))
ParserQueues.TASK_STATUS.put(task)
def _auto_create_table(self, connection, table):
for cnt in range(2):
if table not in self._tables:
if cnt == 1:
connection.create_table(table, {k:{} for k in ['source', 'valuable', 'task']})
TDDCLogging.warning('Create New Table(%s) to HBase.' % table)
self._tables = connection.tables()
else:
break
def _dispatch(self):
while True:
event = self._event_queue.get()
callback = self._event_call.get(event.event_type, None)
if callback:
callback(event)
else:
TDDCLogging.warning('Event Exception: %d Not Register.' %
event.event_type)
def _process(self):
while True:
exception = MonitorQueues.EXCEPTION.get()
cls = self._exception_process.get(exception.code)
if not cls:
TDDCLogging.warning(
'No Match Process To Exception: {exp_id}'.format(
exp_id=exception.id))
continue
cls(exception)
def get(self, table_name, row_key, family=None, qualifier=None):
if not self._status:
TDDCLogging.warning(
'[Get Operation Was Failed] HBase Server Is Exception.')
return False, None
get = TGet()
get.row = row_key
if family:
tc = TColumn()
tc.family = family
if qualifier:
tc.qualifier = qualifier
get.columns = [tc]
try:
ret = None
ret = self._client.get(table_name, get)
except Exception, e:
TDDCLogging.error(e)
return False, None
def put(self, table, row_key, items=None):
if not self._status:
TDDCLogging.warning(
'[Put Operation Was Failed] HBase Server Is Exception.')
return False
cvs = []
for family, info in items.items():
if not isinstance(info, dict):
raise TypeError
for k, v in info.items():
if isinstance(v, list) or isinstance(v, dict):
v = json.dumps(v)
cv = TColumnValue(family, k, v)
cvs.append(cv)
tp = TPut(row_key, cvs)
try:
self._client.put(table, tp)
except Exception, e:
TDDCLogging.error(e)
return False
def error_back(self, response):
task, times = response.request.meta['item']
proxy = response.request.meta.get('proxy', None)
if response.type == httperror.HttpError:
status = response.value.response.status
if status >= 500 or status == 408:
fmt = '[%s][%s] Crawled Failed(\033[0m %d \033[1;37;43m| %s ). Will Retry After While.'
TDDCLogging.warning(fmt % (task.platform,
task.url,
status,
proxy))
self.add_task(task, True)
return
elif status == 404:
retry_times = task.retry if task.retry else 3
if times >= retry_times:
exception = CrawlerTaskFailedException(task)
CrawlerQueues.EXCEPTION.put(exception)
CrawlerQueues.TASK_STATUS_REMOVE.put(task)
fmt = '[%s:%s] Crawled Failed(\033[0m 404 \033[1;37;43m| %s ). Not Retry.'
TDDCLogging.warning(fmt % (task.platform,
task.url,
proxy))
return
times += 1
fmt = '[%s:%s] Crawled Failed(\033[0m %d \033[1;37;43m| %s ). Will Retry After While.'
TDDCLogging.warning(fmt % (task.platform,
task.url,
status,
proxy))
self.add_task(task, True, times)
return
elif response.type == internet_err.TimeoutError:
err_msg = 'TimeoutError'
elif response.type in [internet_err.ConnectionRefusedError,
internet_err.TCPTimedOutError]:
err_msg = '%d:%s' % (response.value.osError, response.value.message)
elif response.type == newclient_err.ResponseNeverReceived:
err_msg = 'ResponseNeverReceived'
else:
err_msg = '%s' % (response.value)
if proxy:
proxy = proxy.split('//')[1]
CrawlerQueues.UNUSEFUL_PROXY_FEEDBACK.put([task.platform, proxy])
fmt = '[%s][%s] Crawled Failed(\033[0m %s \033[1;37;43m| %s ). Will Retry After While.'
TDDCLogging.warning(fmt % (task.platform,
task.url,
err_msg,
proxy))
self.add_task(task, True, times)
def _connect(self):
try:
self._current_host_port = random.choice(
self._host_ports_pool).split(':')
self._sock = TSocket.TSocket(host=self._current_host_port[0],
port=self._current_host_port[1])
self._transport = TTransport.TFramedTransport(self._sock)
self._protocol = TCompactProtocol(self._transport)
self._client = THBaseService.Client(self._protocol)
self._transport.open()
except Exception, e:
TDDCLogging.error(e)
current_host_port = ':'.join(self._current_host_port)
self._host_ports_pool.remove(current_host_port)
if len(self._host_ports_pool) > 0:
TDDCLogging.warning(
'HBase Server Exception. Now Is Reconnecting.')
else:
TDDCLogging.warning(
'HBase Server Fatal Error. Please Check It.')
gevent.sleep(30)
self._host_ports_pool = list(self._host_ports)
TDDCLogging.warning('Retry Connecting HHase.')
self._reconnect()
self._models_table[cls.EXCEPTION_TYPE] = cls
def _recv(self):
self._consumer = KafkaHelper.make_consumer(MonitorSite.KAFKA_NODES,
MonitorSite.EXCEPTION_TOPIC,
MonitorSite.EXCEPTION_GROUP)
while True:
partition_records = self._consumer.poll(2000, 16)
if not len(partition_records):
gevent.sleep(1)
continue
for _, records in partition_records.items():
for record in records:
self._record_proc(record)
def _record_proc(self, record):
try:
exception = json.loads(record.value)
except Exception, e:
self._consume_msg_exp('PARSE_TASK_JSON_ERR', record.value, e)
else:
code = exception.get('code')
if not code:
TDDCLogging.warning('This Exception Is Not Type Of `ExceptionModel`.')
return
cls = self._models_table.get(code)
if not cls:
TDDCLogging.warning('This Exception Is No Match Model.')
return
MonitorQueues.EXCEPTION.put(cls(**exception))
