class Consumer(object):
def __init__(self, addr):
self.client = KafkaClient(addr)
self.topic = "steps_data_part4"
self.consumer_group = 's3_consumer'
self.consumer = SimpleConsumer(self.client, self.consumer_group, self.topic)
def consume_message(self):
while True:
timestamp = time.strftime('%Y%m%d%H%M%S')
temp_file_name = "%s_%s_%s.dat" %(self.topic, self.consumer_group, timestamp)
temp_file = open("/home/ubuntu/rankMyStep/kafka/"+temp_file_name,"w")
messages = self.consumer.get_messages(count=1000, block=False)
for msg in messages:
print msg.message.value + "\n"
temp_file.write(msg.message.value + "\n")
self.save_to_s3(temp_file_name)
def save_to_s3(self, file_name):
mybucket = "anurag-raw-data-store"
aws_access_key = os.getenv('AWS_ACCESS_KEY_ID', 'default')
aws_secret_access_key = os.getenv('AWS_SECRET_ACCESS_KEY', 'default')
s3_client = boto3.client('s3')
s3_client.upload_file("/home/ubuntu/rankMyStep/kafka/"+file_name,
mybucket,"rankmysteps/"+file_name)
os.remove("/home/ubuntu/rankMyStep/kafka/"+file_name)
class Consumer(object):
def __init__(self, addr, group, topic):
"""Initialize Consumer with kafka broker IP, group, and topic."""
self.client = KafkaClient(addr)
self.consumer = SimpleConsumer(self.client, group, topic,
max_buffer_size=1310720000)
self.temp_file_path = None
self.temp_file = None
self.hadoop_path = "/insight/artsy/geo"
self.topic = topic
self.group = group
self.block_cnt = 0
def consume_topic(self, output_dir):
"""Consumes a stream of messages from the "post_geo_activity" topic.
Code template from https://github.com/ajmssc/bitcoin-inspector.git
"""
timestamp = time.strftime('%Y%m%d%H%M%S')
# open file for writing
self.temp_file_path = "%s/kafka_%s_%s_%s.dat" % (output_dir,self.topic,self.group,timestamp)
self.temp_file = open(self.temp_file_path,"w")
while True:
try:
# get 1000 messages at a time, non blocking
messages = self.consumer.get_messages(count=1000, block=False)
for message in messages:
self.temp_file.write(message.message.value + "\n")
# file size > 20MB
if self.temp_file.tell() > 20000000:
self.flush_to_hdfs(output_dir)
self.consumer.commit()
except:
# move to tail of kafka topic if consumer is referencing
# unknown offset
self.consumer.seek(0, 2)
def flush_to_hdfs(self, output_dir):
"""Flushes the 20MB file into HDFS."""
self.temp_file.close()
timestamp = time.strftime('%Y%m%d%H%M%S')
hadoop_fullpath = "%s/%s_%s_%s.dat" % (self.hadoop_path, self.group,self.topic, timestamp)
print "Block {}: Flushing data file to HDFS => {}".format(str(self.block_cnt),hadoop_fullpath)
self.block_cnt += 1
os.system("hdfs dfs -put %s %s" % (self.temp_file_path, hadoop_fullpath)) # save from local to hdfs
os.remove(self.temp_file_path) # remove temp local file
timestamp = time.strftime('%Y%m%d%H%M%S')
self.temp_file_path = "%s/kafka_%s_%s_%s.dat" % (output_dir,self.topic,self.group,timestamp)
self.temp_file = open(self.temp_file_path, "w")
class Consumer(object):
def __init__(self, addr, group, topic):
self.client = KafkaClient(addr)
self.consumer = SimpleConsumer(self.client, group, topic, max_buffer_size=1310720000, auto_offset_reset='smallest')
self.temp_file_path = None
self.temp_file = None
self.topic = topic
self.group = group
self.block_cnt = 0
def consume_topic(self):
timestamp = time.strftime('%Y%m%d%H%M%S')
#open file for writing
self.temp_file_path = "/home/ubuntu/datamill/kafka_%s_%s_%s.dat" % (self.topic, self.group, timestamp)
self.temp_file = open(self.temp_file_path,"w")
header = 'experiment_id,job_id,results_file,package_id,package_name,worker_id,config_id,replicate_no,setup_time,run_time,collect_time,hw_cpu_arch,hw_cpu_mhz,hw_gpu_mhz,hw_num_cpus,hw_page_sz,hw_ram_mhz,hw_ram_sz,sw_address_randomization,sw_autogroup,sw_compiler,sw_drop_caches,sw_env_padding,sw_filesystem,sw_freq_scaling,sw_link_order,sw_opt_flag,sw_swap,sw_sys_time'
self.temp_file.write(header)
while True:
try:
messages = self.consumer.get_messages(count=100, block=False)
for message in messages:
self.temp_file.write(message.message.value + "\n")
if self.temp_file.tell() > 20000:
self.save_to_hdfs()
self.consumer.commit()
except:
self.consumer.seek(0, 2)
self.consumer.commit()
def save_to_hdfs(self):
self.temp_file.close()
timestamp = time.strftime('%Y%m%d%H%M%S')
hadoop_path = "/datamill/%s_%s_%s.csv" % (self.group, self.topic, timestamp)
print "Block " + str(self.block_cnt) + ": Saving file to HDFS " + hadoop_path
self.block_cnt += 1
# place blocked messages into history and cached folders on hdfs
os.system("hdfs dfs -put %s %s" % (self.temp_file_path, hadoop_path))
os.remove(self.temp_file_path)
timestamp = time.strftime('%Y%m%d%H%M%S')
self.temp_file_path = "/home/ubuntu/datamill/kafka_%s_%s_%s.dat" % (self.topic, self.group, timestamp)
self.temp_file = open(self.temp_file_path, "w")
class KafkaDatawakeLookaheadSpout(Spout):
group = 'datawake-crawler-out-consumer'.encode()
def __init__(self):
Spout.__init__(self)
self.queue = None
def initialize(self, stormconf, context):
try:
settings = all_settings.get_settings(stormconf['topology.deployment'])
self.topic = settings['crawler-out-topic'].encode()
self.conn_pool = settings['conn_pool'].encode()
self.log('KafkaDatawakeLookaheadSpout initialized with topic =' + self.topic + ' conn_pool=' + self.conn_pool)
self.kafka = KafkaClient(self.conn_pool)
self.consumer = SimpleConsumer(self.kafka, self.group, self.topic, max_buffer_size=None)
self.consumer.seek(0, 2) # move to the tail of the queue
except:
self.log("KafkaDatawakeLookaheadSpout initialize error", level='error')
self.log(traceback.format_exc(), level='error')
raise
def next_tuple(self):
"""
input message:
dict(
id = input['id'],
appid = input['appid'],
url = url,
status_code = response.getcode(),
status_msg = 'Success',
timestamp = response.info()['date'],
links_found = links,
raw_html = html,
attrs = input['attrs']
)
:return: (url, status, headers, flags, body, timestamp, source,context)
"""
offsetAndMessage = self.consumer.get_messages(timeout=None)[0]
message = offsetAndMessage.message.value
crawled = json.loads(message)
safeurl = crawled['url'].encode('utf-8', 'ignore')
self.log("Lookahead spout received id: " + crawled['id'] + " url: " + safeurl)
context = {
'source': 'datawake-lookahead',
'userId': crawled['attrs']['userId'],
'org': crawled['attrs']['org'],
'domain': crawled['attrs']['domain'],
'url': crawled['url']
}
self.emit([crawled['url'], crawled['status_code'], '', '', crawled['raw_html'], crawled['timestamp'], context['source'], context])
def spiderIdle(self, spider):
consumer = SimpleConsumer(self.kafka_conn, "test", "commands")
for msg in consumer.get_messages():
print msg.message.value
if msg.message.value == spider.name + "_stop":
print "stop"
spider.spider_pause()
# spider.close(spider,'ok')
# self.scrapy.engine.close_spider(spider, 'closespider_itemcount')
if msg.message.value == spider.name + "_start":
# self.scrapy.engine.scraper.open_spider(spider)
spider.spider_resume()
class KafkaConsumer:
group = "python-lookahead-consumer"
def __init__(self,conn_pool,topic,group):
self.conn_pool = conn_pool
self.topic = topic
self.group = group
self.kafka = KafkaClient(self.conn_pool)
self.consumer = SimpleConsumer(self.kafka,self.group,self.topic,max_buffer_size=None)
self.consumer.seek(0,2) # move to the tail of the queue
def next(self):
offsetAndMessage = self.consumer.get_messages(timeout=None)[0]
message = offsetAndMessage.message.value
return message
def get_message() :
try :
kconn = KafkaClient(kafka_producer.hosts , timeout = 10)
getter = SimpleConsumer(kconn , 'test_group', kafka_producer.topic)
#getter.seek(0, 0)
while True:
try:
messages = getter.get_messages(200,timeout=3)
if messages:
logging.info('get message from kafka done'+str(decode(messages)))
import time
time.sleep(0.1)
except BaseException ,e:
logging.error(str(e))
except BaseException , e :
logging.error(str(e) + 'get message from kafka failed')
class KafkaDatawakeVisitedSpout(Spout):
group = 'datawake-visited-consumer'.encode()
def __init__(self):
Spout.__init__(self)
self.queue = None
def initialize(self, stormconf, context):
try:
settings = all_settings.get_settings(stormconf['topology.deployment'])
self.topic = settings['visited-topic'].encode()
self.conn_pool = settings['conn_pool'].encode()
self.log('KafkaDatawakeVisitedSpout initialized with topic =' + self.topic + ' conn_pool=' + self.conn_pool)
self.kafka = KafkaClient(self.conn_pool)
self.consumer = SimpleConsumer(self.kafka, self.group, self.topic, max_buffer_size=None)
self.consumer.seek(0, 2) # move to the tail of the queue
except:
self.log("KafkaDatawakeVisitedSpout initialize error", level='error')
self.log(traceback.format_exc(), level='error')
raise
def next_tuple(self):
"""
input: (timestamp,org,domain,user_id,url,html)
:return: (url, status, headers, flags, body, timestamp, source,context)
"""
offsetAndMessage = self.consumer.get_messages(timeout=None)[0]
message = offsetAndMessage.message.value
message = message.decode('utf-8')
message = message.split('\0')
(timestamp, org, domain, userId, url, html) = message
context = {
'source': 'datawake-visited',
'userId': userId,
'org': org,
'domain': domain,
'url': url
}
self.emit([url, '', '', '', html, timestamp, context['source'], context])
class PerfConsumerSync ( threading.Thread ):
running = True
def __init__(self, factory, destination):
self.factory = factory
self.destination = destination
self.consumer = SimpleConsumer(self.factory, "test-group", self.destination)
self.rate = PerfRate()
threading.Thread.__init__ ( self )
def run (self):
while (self.running):
textMessage = self.consumer.get_messages(block=True, timeout=1000000)
if (textMessage != None):
self.rate.increment()
def stop(self):
self.running = False
def start(self):
threading.Thread.start(self)
class KafkaMonitor:
def __init__(self, settings_name, unit_test=False):
'''
@param settings_name: the local settings file name
@param unit_test: whether running unit tests or not
'''
self.settings_name = settings_name
self.wrapper = SettingsWrapper()
self.logger = None
self.unit_test = unit_test
def _import_class(self, cl):
'''
Imports a class from a string
@param name: the module and class name in dot notation
'''
d = cl.rfind(".")
classname = cl[d + 1:len(cl)]
m = __import__(cl[0:d], globals(), locals(), [classname])
return getattr(m, classname)
def _load_plugins(self):
'''
Sets up all plugins, defaults and settings.py
'''
plugins = self.settings['PLUGINS']
self.plugins_dict = {}
for key in plugins:
# skip loading the plugin if its value is None
if plugins[key] is None:
continue
# valid plugin, import and setup
self.logger.debug("Trying to load plugin {cls}".format(cls=key))
the_class = self._import_class(key)
instance = the_class()
instance._set_logger(self.logger)
if not self.unit_test:
instance.setup(self.settings)
the_schema = None
print("self.settings['PLUGIN_DIR'] + instance.schema====",
self.settings['PLUGIN_DIR'] + instance.schema)
with open(self.settings['PLUGIN_DIR'] +
instance.schema) as the_file:
the_schema = json.load(the_file)
mini = {}
mini['instance'] = instance
mini['schema'] = the_schema
self.plugins_dict[plugins[key]] = mini
self.plugins_dict = OrderedDict(
sorted(self.plugins_dict.items(), key=lambda t: t[0]))
def setup(self, level=None, log_file=None, json=None):
'''
Load everything up. Note that any arg here will override both
default and custom settings
@param level: the log level
@param log_file: boolean t/f whether to log to a file, else stdout
@param json: boolean t/f whether to write the logs in json
'''
self.settings = self.wrapper.load(self.settings_name)
my_level = level if level else self.settings['LOG_LEVEL']
# negate because logger wants True for std out
my_output = not log_file if log_file else self.settings['LOG_STDOUT']
my_json = json if json else self.settings['LOG_JSON']
self.logger = LogFactory.get_instance(
json=my_json,
stdout=my_output,
level=my_level,
name=self.settings['LOGGER_NAME'],
dir=self.settings['LOG_DIR'],
file=self.settings['LOG_FILE'],
bytes=self.settings['LOG_MAX_BYTES'],
backups=self.settings['LOG_BACKUPS'])
self.validator = self.extend_with_default(Draft4Validator)
def _setup_stats(self):
'''
Sets up the stats collection
'''
self.stats_dict = {}
redis_conn = redis.Redis(host=self.settings['REDIS_HOST'],
port=self.settings['REDIS_PORT'])
try:
redis_conn.info()
self.logger.debug("Connected to Redis in StatsCollector Setup")
except ConnectionError:
self.logger.warn("Failed to connect to Redis in StatsCollector"
" Setup, no stats will be collected")
return
if self.settings['STATS_TOTAL']:
self._setup_stats_total(redis_conn)
if self.settings['STATS_PLUGINS']:
self._setup_stats_plugins(redis_conn)
def _setup_stats_total(self, redis_conn):
'''
Sets up the total stats collectors
@param redis_conn: the redis connection
'''
self.stats_dict['total'] = {}
self.stats_dict['fail'] = {}
temp_key1 = 'stats:kafka-monitor:total'
temp_key2 = 'stats:kafka-monitor:fail'
for item in self.settings['STATS_TIMES']:
try:
time = getattr(StatsCollector, item)
self.stats_dict['total'][time] = StatsCollector \
.get_rolling_time_window(
redis_conn=redis_conn,
key='{k}:{t}'.format(k=temp_key1, t=time),
window=time,
cycle_time=self.settings['STATS_CYCLE'])
self.stats_dict['fail'][time] = StatsCollector \
.get_rolling_time_window(
redis_conn=redis_conn,
key='{k}:{t}'.format(k=temp_key2, t=time),
window=time,
cycle_time=self.settings['STATS_CYCLE'])
self.logger.debug("Set up total/fail Stats Collector '{i}'"\
.format(i=item))
except AttributeError as e:
self.logger.warning("Unable to find Stats Time '{s}'"\
.format(s=item))
total1 = StatsCollector.get_hll_counter(
redis_conn=redis_conn,
key='{k}:lifetime'.format(k=temp_key1),
cycle_time=self.settings['STATS_CYCLE'],
roll=False)
total2 = StatsCollector.get_hll_counter(
redis_conn=redis_conn,
key='{k}:lifetime'.format(k=temp_key2),
cycle_time=self.settings['STATS_CYCLE'],
roll=False)
self.logger.debug("Set up total/fail Stats Collector 'lifetime'")
self.stats_dict['total']['lifetime'] = total1
self.stats_dict['fail']['lifetime'] = total2
def _setup_stats_plugins(self, redis_conn):
'''
Sets up the plugin stats collectors
@param redis_conn: the redis connection
'''
self.stats_dict['plugins'] = {}
for key in self.plugins_dict:
plugin_name = self.plugins_dict[key]['instance'].__class__.__name__
temp_key = 'stats:kafka-monitor:{p}'.format(p=plugin_name)
self.stats_dict['plugins'][plugin_name] = {}
for item in self.settings['STATS_TIMES']:
try:
time = getattr(StatsCollector, item)
self.stats_dict['plugins'][plugin_name][time] = StatsCollector \
.get_rolling_time_window(
redis_conn=redis_conn,
key='{k}:{t}'.format(k=temp_key, t=time),
window=time,
cycle_time=self.settings['STATS_CYCLE'])
self.logger.debug("Set up {p} plugin Stats Collector '{i}'"\
.format(p=plugin_name, i=item))
except AttributeError:
self.logger.warning("Unable to find Stats Time '{s}'"\
.format(s=item))
total = StatsCollector.get_hll_counter(
redis_conn=redis_conn,
key='{k}:lifetime'.format(k=temp_key),
cycle_time=self.settings['STATS_CYCLE'],
roll=False)
self.logger.debug("Set up {p} plugin Stats Collector 'lifetime'"\
.format(p=plugin_name))
self.stats_dict['plugins'][plugin_name]['lifetime'] = total
def _setup_kafka(self):
'''
Sets up kafka connections
'''
@MethodTimer.timeout(self.settings['KAFKA_CONN_TIMEOUT'], False)
def _hidden_setup():
try:
self.kafka_conn = KafkaClient(self.settings['KAFKA_HOSTS'])
self.kafka_conn.ensure_topic_exists(
self.settings['KAFKA_INCOMING_TOPIC'])
self.consumer = SimpleConsumer(
self.kafka_conn,
self.settings['KAFKA_GROUP'],
self.settings['KAFKA_INCOMING_TOPIC'],
auto_commit=True,
iter_timeout=1.0)
except KafkaUnavailableError as ex:
message = "An exception '{0}' occured. Arguments:\n{1!r}" \
.format(type(ex).__name__, ex.args)
self.logger.error(message)
sys.exit(1)
return True
ret_val = _hidden_setup()
if ret_val:
self.logger.debug("Successfully connected to Kafka")
else:
self.logger.error("Failed to set up Kafka Connection within"
" timeout")
# this is essential to running the kafka monitor
sys.exit(1)
def extend_with_default(self, validator_class):
'''
Method to add default fields to our schema validation
( From the docs )
'''
validate_properties = validator_class.VALIDATORS["properties"]
def set_defaults(validator, properties, instance, schema):
for error in validate_properties(
validator,
properties,
instance,
schema,
):
yield error
for property, subschema in properties.iteritems():
if "default" in subschema:
instance.setdefault(property, subschema["default"])
return validators.extend(
validator_class,
{"properties": set_defaults},
)
def _main_loop(self):
'''
Continuous loop that reads from a kafka topic and tries to validate
incoming messages
'''
self.logger.debug("Processing messages")
old_time = 0
while True:
self._process_messages()
if self.settings['STATS_DUMP'] != 0:
new_time = int(time.time() / self.settings['STATS_DUMP'])
# only log every X seconds
if new_time != old_time:
self._dump_stats()
old_time = new_time
time.sleep(.01)
def _process_messages(self):
try:
for message in self.consumer.get_messages():
if message is None:
self.logger.debug("no message")
break
try:
self._increment_total_stat(message.message.value)
the_dict = json.loads(message.message.value)
print('the_dict', the_dict)
found_plugin = False
print('self.plugins_dict', self.plugins_dict)
for key in self.plugins_dict:
obj = self.plugins_dict[key]
instance = obj['instance']
print('instance==', instance)
schema = obj['schema']
print(
'schema********************************************',
schema)
try:
print('before v = self.validator(schema)')
v = self.validator(schema)
print('after v = self.validator(schema)')
print('the_dict-------', the_dict)
v.validate(the_dict)
found_plugin = True
print('found_plugin====', found_plugin)
self._increment_plugin_stat(
instance.__class__.__name__, the_dict)
print('instance.handle(the_dict)', the_dict)
ret = instance.handle(the_dict)
# break if nothing is returned
if ret is None:
break
except ValidationError:
print(' except ValidationError:======')
pass
if not found_plugin:
extras = {}
extras['parsed'] = True
extras['valid'] = False
extras['data'] = the_dict
self.logger.warn(
"Did not find schema to validate "
"request",
extra=extras)
self._increment_fail_stat(the_dict)
except ValueError:
extras = {}
extras['parsed'] = False
extras['valid'] = False
extras['data'] = message.message.value
self.logger.warning('Unparseable JSON Received',
extra=extras)
self._increment_fail_stat(message.message.value)
except OffsetOutOfRangeError:
# consumer has no idea where they are
self.consumer.seek(0, 2)
self.logger.error("Kafka offset out of range error")
def _increment_total_stat(self, string):
'''
Increments the total stat counters
@param string: the loaded message object for the counter
'''
string = string + str(time.time())
if 'total' in self.stats_dict:
self.logger.debug("Incremented total stats")
for key in self.stats_dict['total']:
if key == 'lifetime':
self.stats_dict['total'][key].increment(string)
else:
self.stats_dict['total'][key].increment()
def _increment_fail_stat(self, item):
'''
Increments the total stat counters
@param item: the loaded message object for HLL counter
'''
if isinstance(item, dict):
item['ts'] = time.time()
elif isinstance(item, str):
item = item + str(time.time())
if 'fail' in self.stats_dict:
self.logger.debug("Incremented fail stats")
for key in self.stats_dict['fail']:
if key == 'lifetime':
self.stats_dict['fail'][key].increment(item)
else:
self.stats_dict['fail'][key].increment()
def _increment_plugin_stat(self, name, item):
'''
Increments the total stat counters
@param name: The formal name of the plugin
@param dict: the loaded message object for HLL counter
'''
item['ts'] = time.time()
if 'plugins' in self.stats_dict:
self.logger.debug("Incremented plugin '{p}' plugin stats"\
.format(p=name))
for key in self.stats_dict['plugins'][name]:
if key == 'lifetime':
self.stats_dict['plugins'][name][key].increment(item)
else:
self.stats_dict['plugins'][name][key].increment()
def _dump_stats(self):
'''
Dumps the stats out
'''
extras = {}
if 'total' in self.stats_dict:
self.logger.debug("Compiling total/fail dump stats")
for key in self.stats_dict['total']:
final = 'total_{t}'.format(t=key)
extras[final] = self.stats_dict['total'][key].value()
for key in self.stats_dict['fail']:
final = 'fail_{t}'.format(t=key)
extras[final] = self.stats_dict['fail'][key].value()
if 'plugins' in self.stats_dict:
self.logger.debug("Compiling plugin dump stats")
for name in self.stats_dict['plugins']:
for key in self.stats_dict['plugins'][name]:
final = 'plugin_{n}_{t}'.format(n=name, t=key)
extras[final] = self.stats_dict['plugins'][name][
key].value()
if not self.logger.json:
self.logger.info('Kafka Monitor Stats Dump:\n{0}'.format(
json.dumps(extras, indent=4, sort_keys=True)))
else:
self.logger.info('Kafka Monitor Stats Dump', extra=extras)
def run(self):
'''
Set up and run
'''
self._setup_kafka()
self._load_plugins()
self._setup_stats()
self._main_loop()
def feed(self, json_item):
'''
Feeds a json item into the Kafka topic
@param json_item: The loaded json object
'''
@MethodTimer.timeout(self.settings['KAFKA_FEED_TIMEOUT'], False)
def _feed(json_item):
try:
self.kafka_conn = KafkaClient(self.settings['KAFKA_HOSTS'])
topic = self.settings['KAFKA_INCOMING_TOPIC']
producer = SimpleProducer(self.kafka_conn)
except KafkaUnavailableError:
self.logger.error("Unable to connect to Kafka")
return False
if not self.logger.json:
self.logger.info('Feeding JSON into {0}\n{1}'.format(
topic, json.dumps(json_item, indent=4)))
else:
self.logger.info('Feeding JSON into {0}\n'.format(topic),
extra={'value': json_item})
self.kafka_conn.ensure_topic_exists(topic)
producer.send_messages(topic, json.dumps(json_item))
return True
result = _feed(json_item)
if result:
self.logger.info("Successfully fed item to Kafka")
else:
self.logger.error("Failed to feed item into Kafka")
class KafkaMonitor:
def __init__(self, settings):
# dynamic import of settings file
# remove the .py from the filename
self.settings = importlib.import_module(settings[:-3])
# only need kafka for both uses
self.kafka_conn = KafkaClient(self.settings.KAFKA_HOSTS)
def get_method(self, key):
if key == 'handle_crawl_request':
return self.handle_crawl_request
elif key == 'handle_action_request':
return self.handle_action_request
raise AttributeError(key)
def setup(self):
self.redis_conn = redis.Redis(host=self.settings.REDIS_HOST,
port=self.settings.REDIS_PORT)
self.kafka_conn.ensure_topic_exists(self.settings.KAFKA_INCOMING_TOPIC)
self.consumer = SimpleConsumer(self.kafka_conn,
self.settings.KAFKA_GROUP,
self.settings.KAFKA_INCOMING_TOPIC,
auto_commit=True,
iter_timeout=1.0)
self.result_method = self.get_method(self.settings.SCHEMA_METHOD)
self.validator = self.extend_with_default(Draft4Validator)
def extend_with_default(self, validator_class):
'''
Method to add default fields to our schema validation
( From the docs )
'''
validate_properties = validator_class.VALIDATORS["properties"]
def set_defaults(validator, properties, instance, schema):
for error in validate_properties(
validator,
properties,
instance,
schema,
):
yield error
for property, subschema in properties.iteritems():
if "default" in subschema:
instance.setdefault(property, subschema["default"])
return validators.extend(
validator_class,
{"properties": set_defaults},
)
def handle_crawl_request(self, dict):
'''
Processes a vaild crawl request
@param dict: a valid dictionary object
'''
# format key
key = "{sid}:queue".format(sid=dict['spiderid'])
val = pickle.dumps(dict, protocol=-1)
# shortcut to shove stuff into the priority queue
self.redis_conn.zadd(key, val, -dict['priority'])
# if timeout crawl, add value to redis
if 'expires' in dict:
key = "timeout:{sid}:{appid}:{crawlid}".format(
sid=dict['spiderid'],
appid=dict['appid'],
crawlid=dict['crawlid'])
self.redis_conn.set(key, dict['expires'])
print 'Added crawl to Redis'
def handle_action_request(self, dict):
'''
Processes a vaild action request
@param dict: The valid dictionary object
'''
# format key
key = "{action}:{spiderid}:{appid}".format(action=dict['action'],
spiderid=dict['spiderid'],
appid=dict['appid'])
if "crawlid" in dict:
key = key + ":" + dict['crawlid']
self.redis_conn.set(key, dict['uuid'])
print 'Added action to Redis'
def _main_loop(self):
'''
Continuous loop that reads from a kafka topic and tries to validate
incoming messages
'''
while True:
start = time.time()
try:
for message in self.consumer.get_messages():
if message is None:
break
try:
the_dict = json.loads(message.message.value)
try:
self.validator(self.schema).validate(the_dict)
self.result_method(the_dict)
except ValidationError as ex:
print "invalid json received"
except ValueError:
print "bad json recieved"
except OffsetOutOfRangeError:
# consumer has no idea where they are
self.consumer.seek(0, 2)
end = time.time()
time.sleep(.01)
def run(self):
'''
Sets up the schema to be validated against
'''
self.setup()
with open(self.settings.SCHEMA) as the_file:
# No try/catch so we can see if there is a json parse error
# on the schemas
self.schema = json.load(the_file)
self._main_loop()
def feed(self, json_item):
'''
Feeds a json item into the Kafka topic
@param json_item: The loaded json object
'''
topic = self.settings.KAFKA_INCOMING_TOPIC
producer = SimpleProducer(self.kafka_conn)
print "=> feeding JSON request into {0}...".format(topic)
print json.dumps(json_item, indent=4)
self.kafka_conn.ensure_topic_exists(topic)
producer.send_messages(topic, json.dumps(json_item))
print "=> done feeding request."
class ZKConsumer(object):
zk_timeout = 30
jitter_seconds = 30
broker_prefix = '/brokers/ids'
def __init__(self,
zk_hosts,
group,
topic,
nodes,
zk_handler=None,
logger=None,
identifier=None,
**consumer_kwargs):
"""Creates a Consumer that tracks state in ZooKeeper,
rebalancing partition ownership as registered consumers change.
NOTE: this class is intended for version 0.8.1 of Kafka, where offsets
are managed by Kafka but there is no rebalancing in the protocol.
"""
if logger is None:
logger = logging.getLogger('kafka.consumer.ZKConsumer')
self.logger = logger
self.identifier = identifier
if KafkaClient is None:
raise RuntimeError(
"Kafka support requires cs.eyrie to be installed with the Kafka extra: install_requires= ['cs.eyrie[Kafka]']"
)
self.zk_handler = zk_handler
self.zk_hosts = zk_hosts
self.broker_hosts = []
self.group = group
self.topic = topic
self.zk = None
self.nodes = nodes
self.client = None
self.consumer = None
self.consumer_kwargs = consumer_kwargs
# This will kick off a cascading sequence to initialize ourselves:
# 1. Connect to ZK and pull list of Kafka brokers
# 2. Register ourselves as a consumer in ZK
# 3. Rebalance partitions across all connected consumers
self.init_zk()
def zk_session_watch(self, state):
self.logger.debug('ZK transitioned to: %s', state)
if state == KazooState.SUSPENDED:
if self.consumer is not None:
self.logger.info('Stopping Kafka consumer')
self.consumer.stop()
self.consumer = None
# Lost connection to ZK; we can't call any methods that would
# try to contact it (i.e., we can't do self.zkp.finish() )
self.zkp = None
elif state == KazooState.CONNECTED:
self.logger.info('Restarting ZK partitioner')
self.zk.handler.spawn(self.init_zkp)
def _zkp_wait(self):
handler = self.zk.handler
while 1:
if self.zkp.failed:
self.logger.warning("Lost or unable to acquire partition")
self.stop()
elif self.zkp.release:
self.zkp.release_set()
elif self.zkp.acquired:
def group_change_proxy(event):
self.logger.warn('Connected consumers changed')
if self.zkp is None:
self.logger.info('Restarting ZK partitioner')
handler.spawn(self.init_zkp)
elif self.zkp is not None and self.zkp.failed:
self.logger.warning(
"Lost or unable to acquire partition")
self.stop()
else:
self.logger.info(
'Scheduling ZK partitioner set release')
rel_greenlet = handler.spawn(self.zkp.release_set)
self.logger.info('Scheduling group re-join')
rel_greenlet.link_value(
lambda greenlet: self.zkp.join_group)
if not self.nodes:
self.logger.info(
'Partitioner aquired; setting child watch')
result = self.zk.get_children_async(self.zkp._group_path)
result.rawlink(group_change_proxy)
# Break out of while loop to begin consuming events
break
elif self.zkp.allocating:
self.zkp.wait_for_acquire()
def init_zkp(self):
if not hasattr(self, 'zkp') or self.zkp is None:
if self.nodes:
self.zkp = StaticZKPartitioner(
self.zk,
self.group,
self.topic,
self.nodes,
partitions_changed_cb=self.init_consumer,
logger=self.logger,
identifier=self.identifier)
else:
self.zkp = ZKPartitioner(
self.zk,
self.group,
self.topic,
time_boundary=self.jitter_seconds,
partitions_changed_cb=self.init_consumer,
logger=self.logger,
identifier=self.identifier)
self._zkp_wait()
def init_zk(self):
# TODO: switch to async
# 1. implement kazoo.interfaces.IHandler in terms of Tornado's IOLoop
self.zk = KazooClient(hosts=self.zk_hosts, handler=self.zk_handler)
self.zk.start()
self.zk.add_listener(self.zk_session_watch)
@self.zk.ChildrenWatch(self.broker_prefix)
def broker_change_proxy(broker_ids):
self.onBrokerChange(broker_ids)
self.init_zkp()
def onBrokerChange(self, broker_ids):
self.broker_hosts = []
for b_id in broker_ids:
b_json, zstat = self.zk.get('/'.join([self.broker_prefix, b_id]))
b_data = json.loads(b_json)
self.broker_hosts.append('{}:{}'.format(b_data['host'],
b_data['port']))
my_partitions = []
if self.consumer is not None:
self.logger.warn('Brokers changed, stopping Kafka consumer.')
my_partitions = self.consumer.offsets.keys()
self.consumer.stop()
self.consumer = None
if self.client is not None:
self.logger.warn('Brokers changed, stopping Kafka client.')
self.client.close()
self.client = None
if my_partitions:
msg = 'Brokers changed, queuing restart of Kafka client / consumer.'
self.logger.warn(msg)
self.zk.handler.spawn(self.init_consumer, my_partitions)
def init_consumer(self, my_partitions):
if self.consumer is None:
self.logger.warn('Starting Kafka client')
self.client = KafkaClient(self.broker_hosts,
client_id=self.zkp._identifier)
else:
if self.consumer is None or \
sorted(my_partitions) != sorted(self.consumer.offsets.keys()):
self.logger.warn(
'Partitions changed, restarting Kafka consumer.')
self.consumer.stop()
else:
self.logger.info(
'Partitions unchanged, not restarting Kafka consumer.')
return
self.consumer = SimpleConsumer(self.client,
self.group,
self.topic,
partitions=my_partitions,
**self.consumer_kwargs)
self.consumer.provide_partition_info()
self.logger.info("Consumer connected to Kafka: %s",
self.consumer.offsets)
def stop(self):
if self.consumer is not None:
self.logger.info('Stopping Kafka consumer')
self.consumer.stop()
self.consumer = None
if self.client is not None:
self.logger.info('Stopping Kafka client')
self.client.close()
self.client = None
if self.zk is not None:
self.logger.info('Stopping ZooKeeper client')
if self.zkp is not None and not self.zkp.failed:
self.zkp.finish()
self.zk.stop()
self.zkp = None
self.zk = None
def commit(self, partitions=None):
"""
Commit offsets for this consumer
partitions: list of partitions to commit, default is to commit
all of them
"""
if self.consumer is None:
return
self.logger.debug('Begin committing offsets for partitions: %s',
partitions if partitions else 'All')
self.consumer.commit(partitions)
self.logger.debug('End committing offsets for partitions: %s',
partitions if partitions else 'All')
def pending(self, partitions=None):
"""
Gets the pending message count
partitions: list of partitions to check for, default is to check all
"""
return self.consumer.pending(partitions)
def provide_partition_info(self):
"""
Indicates that partition info must be returned by the consumer
"""
self.consumer.provide_partition_info()
def seek(self, offset, whence):
"""
Alter the current offset in the consumer, similar to fseek
offset: how much to modify the offset
whence: where to modify it from
0 is relative to the earliest available offset (head)
1 is relative to the current offset
2 is relative to the latest known offset (tail)
"""
self.consumer.seek(offset, whence)
def get_messages(self, count=1, block=True, timeout=0.1):
"""
Fetch the specified number of messages
count: Indicates the maximum number of messages to be fetched
block: If True, the API will block till some messages are fetched.
timeout: If block is True, the function will block for the specified
time (in seconds) until count messages is fetched. If None,
it will block forever.
"""
if self.consumer is None:
return []
else:
try:
messages = self.consumer.get_messages(count, block, timeout)
if not messages and self.zkp.failed:
raise FailedPayloadsError
return messages
except FailedPayloadsError as err:
msg = 'Failed to retrieve payload, restarting consumer'
self.logger.exception(msg)
raise err
def get_message(self, block=True, timeout=0.1, get_partition_info=None):
return self.consumer.get_message(block, timeout, get_partition_info)
def _get_message(self,
block=True,
timeout=0.1,
get_partition_info=None,
update_offset=True):
return self.consumer._get_message(block, timeout, get_partition_info,
update_offset)
def __iter__(self):
for msg in self.consumer:
yield msg
class Consumer(object):
"""Kafka consumer class with functions to consume messages to HDFS.
Messages are blocked into 20MB files and transferred to HDFS
Attributes:
client: string representing IP:port of the kafka broker
consumer: Consumer object specifying the client group, and topic
temp_file_path: location of the 20MB file to be appended to before
transfer to HDFS
temp_file: File object opened from temp_file_path
topic: String representing the topic on Kafka
group: String representing the Kafka consumer group to be associated
with
block_cnt: integer representing the block count for print statements
"""
def __init__(self, addr, group, topic):
"""Initialize Consumer with kafka broker IP, group, and topic."""
self.client = KafkaClient(addr)
self.consumer = SimpleConsumer(self.client, group, topic, max_buffer_size=1310720000)
self.temp_file_path = None
self.temp_file = None
self.hadoop_path = "/user/parking_data/history"
self.topic = topic
self.group = group
self.block_cnt = 0
def consume_topic(self, output_dir):
"""Consumes a stream of messages from the "messages" topic.
Code template from https://github.com/ajmssc/bitcoin-inspector.git
Args:
output_dir: string representing the directory to store the 20MB
before transferring to HDFS
Returns:
None
"""
timestamp = time.strftime("%Y%m%d%H%M%S")
# open file for writing
self.temp_file_path = "%s/kafka_%s_%s_%s.dat" % (output_dir, self.topic, self.group, timestamp)
self.temp_file = open(self.temp_file_path, "w")
# while True:
for ii in range(0, 2):
try:
# get 1000 messages at a time, non blocking
messages = self.consumer.get_messages(count=1000, block=False)
# OffsetAndMessage(offset=43, message=Message(magic=0,
# attributes=0, key=None, value='some message'))
for message in messages:
self.temp_file.write(message.message.value + "\n")
# file size > 20MB
if self.temp_file.tell() > 20000000:
self.flush_to_hdfs(output_dir)
self.consumer.commit()
except:
# move to tail of kafka topic if consumer is referencing
# unknown offset
self.consumer.seek(0, 2)
def flush_to_hdfs(self, output_dir):
"""Flushes the 20MB file into HDFS.
Code template from https://github.com/ajmssc/bitcoin-inspector.git
Flushes the file into HDFS folders
Args:
output_dir: string representing the directory to store the 20MB
before transferring to HDFS
Returns:
None
"""
self.temp_file.close()
timestamp = time.strftime("%Y%m%d%H%M%S")
hadoop_fullpath = "%s/%s_%s_%s.dat" % (self.hadoop_path, self.group, self.topic, timestamp)
print "Block {}: Flushing 20MB file to HDFS => {}".format(str(self.block_cnt), hadoop_fullpath)
self.block_cnt += 1
# place blocked messages into history and cached folders on hdfs
print ("hdfs dfs -put %s %s" % (self.temp_file_path, hadoop_fullpath))
os.system("sudo hdfs dfs -put %s %s" % (self.temp_file_path, hadoop_fullpath))
# os.system("sudo -u hdfs hdfs dfs -put %s %s" % (self.temp_file_path,
# cached_fullpath))
os.remove(self.temp_file_path)
timestamp = time.strftime("%Y%m%d%H%M%S")
self.temp_file_path = "%s/kafka_%s_%s_%s.dat" % (output_dir, self.topic, self.group, timestamp)
self.temp_file = open(self.temp_file_path, "w")
kafka_hosts = arguments.hosts
# getting the path to the file where we will dump the content of topic
path_to_sink = arguments.path
# getting the required number of messages to fetch from kafka
message_count = arguments.count
# instantiating a Kafka client
kafka_client = SimpleClient(hosts=kafka_hosts)
# instantiating a Kafka consumer
kafka_consumer = SimpleConsumer(client=kafka_client,
topic=topic_to_collect_from,
group='simple_consumer_group')
# fetching the messages
messages_data = kafka_consumer.get_messages(count=message_count)
# creating a dictionary
message_dictionary = {}
# running through the
for index, message in enumerate(messages_data):
message_dictionary[index] = {
'value': message.message.value.decode('utf-8'),
'offset': message.offset,
}
# dumping messages
with open(path_to_sink, 'w') as sink_file:
json.dump(message_dictionary, sink_file)
# printing the content of the json file in the terminal
pprint.pprint(message_dictionary)
class KafkaMonitor:
def __init__(self, settings):
# dynamic import of settings file
# remove the .py from the filename
self.settings = importlib.import_module(settings[:-3])
# only need kafka for both uses
self.kafka_conn = KafkaClient(self.settings.KAFKA_HOSTS)
def get_method(self, key):
if key == 'handle_crawl_request':
return self.handle_crawl_request
elif key == 'handle_action_request':
return self.handle_action_request
raise AttributeError(key)
def setup(self):
self.redis_conn = redis.Redis(host=self.settings.REDIS_HOST,
port=self.settings.REDIS_PORT)
self.kafka_conn.ensure_topic_exists(self.settings.KAFKA_INCOMING_TOPIC)
self.consumer = SimpleConsumer(self.kafka_conn,
self.settings.KAFKA_GROUP,
self.settings.KAFKA_INCOMING_TOPIC,
auto_commit=True,
iter_timeout=1.0)
self.result_method = self.get_method(self.settings.SCHEMA_METHOD)
self.validator = self.extend_with_default(Draft4Validator)
def extend_with_default(self, validator_class):
'''
Method to add default fields to our schema validation
( From the docs )
'''
validate_properties = validator_class.VALIDATORS["properties"]
def set_defaults(validator, properties, instance, schema):
for error in validate_properties(
validator, properties, instance, schema,
):
yield error
for property, subschema in properties.iteritems():
if "default" in subschema:
instance.setdefault(property, subschema["default"])
return validators.extend(
validator_class, {"properties": set_defaults},
)
def handle_crawl_request(self, dict):
'''
Processes a vaild crawl request
@param dict: a valid dictionary object
'''
# format key
key = "{sid}:queue".format(sid=dict['spiderid'])
val = pickle.dumps(dict, protocol=-1)
# shortcut to shove stuff into the priority queue
self.redis_conn.zadd(key, val, -dict['priority'])
# if timeout crawl, add value to redis
if 'expires' in dict:
key = "timeout:{sid}:{appid}:{crawlid}".format(
sid=dict['spiderid'],
appid=dict['appid'],
crawlid=dict['crawlid'])
self.redis_conn.set(key, dict['expires'])
def handle_action_request(self, dict):
'''
Processes a vaild action request
@param dict: The valid dictionary object
'''
# format key
key = "{action}:{spiderid}:{appid}".format(
action=dict['action'],
spiderid=dict['spiderid'],
appid=dict['appid'])
if "crawlid" in dict:
key = key + ":" + dict['crawlid']
self.redis_conn.set(key, dict['uuid'])
def _main_loop(self):
'''
Continuous loop that reads from a kafka topic and tries to validate
incoming messages
'''
while True:
start = time.time()
try:
for message in self.consumer.get_messages():
if message is None:
break
try:
the_dict = json.loads(message.message.value)
try:
self.validator(self.schema).validate(the_dict)
self.result_method(the_dict)
except ValidationError as ex:
print "invalid json received"
except ValueError:
print "bad json recieved"
except OffsetOutOfRangeError:
# consumer has no idea where they are
self.consumer.seek(0, 2)
end = time.time()
time.sleep(.01)
def run(self):
'''
Sets up the schema to be validated against
'''
self.setup()
with open(self.settings.SCHEMA) as the_file:
# No try/catch so we can see if there is a json parse error
# on the schemas
self.schema = json.load(the_file)
self._main_loop()
def feed(self, json_item):
'''
Feeds a json item into the Kafka topic
@param json_item: The loaded json object
'''
topic = self.settings.KAFKA_INCOMING_TOPIC
producer = SimpleProducer(self.kafka_conn)
print "=> feeding JSON request into {0}...".format(topic)
print json.dumps(json_item, indent=4)
self.kafka_conn.ensure_topic_exists(topic)
producer.send_messages(topic, json.dumps(json_item))
print "=> done feeding request."
class Consumer(object):
def __init__(self, addr, group, topic):
self.client = KafkaClient(addr)
self.consumer = SimpleConsumer(self.client, group, topic, max_buffer_size=1310720000)
self.temp_file_path = None
self.temp_file = None
self.topic = topic
self.group = group
self.block_cnt = 0
def consume_topic(self, output_dir):
timestamp = time.strftime('%Y%m%d%H%M%S')
#open file for writing
self.temp_file_path = "/home/ubuntu/FantasyFootball/ingestion/kafka_%s_%s_%s.dat" % (self.topic, self.group, timestamp)
self.temp_file = open(self.temp_file_path,"w")
one_entry = False
while True:
try:
messages = self.consumer.get_messages(count=100, block=False)
#OffsetAndMessage(offset=43, message=Message(magic=0,
# attributes=0, key=None, value='some message'))
for message in messages:
one_entry = True
self.tempfile.write(message.message.value + "\n")
if self.tempfile.tell() > 2000:
self.save_to_hdfs(output_dir)
self.consumer.commit()
except:
self.consumer.seek(0, 2)
if one_entry:
self.save_to_hdfs(output_dir, self.topic)
self.consumer.commit()
def save_to_hdfs(self, output_dir):
self.tempfile.close()
timestamp = time.strftime('%Y%m%d%H%M%S')
hadoop_path = "/user/solivero/playerpoints/history/%s_%s_%s.dat" % (self.group, self.topic, timestamp)
cached_path = "/user/solivero/playerpoints/cached/%s_%s_%s.dat" % (self.group, self.topic, timestamp)
print "Block " + str(self.block_cnt) + ": Saving file to HDFS " + hadoop_path
self.block_cnt += 1
# place blocked messages into history and cached folders on hdfs
os.system("sudo -u hdfs hdfs dfs -put %s %s" % (self.temp_file_path,hadoop_path))
os.system("sudo -u hdfs hdfs dfs -put %s %s" % (self.temp_file_path,cached_path))
os.remove(self.temp_file_path)
timestamp = time.strftime('%Y%m%d%H%M%S')
self.temp_file_path = "/home/ubuntu/fantasyfootball/ingestion/kafka_%s_%s_%s.dat" % (self.topic, self.group, timestamp)
self.temp_file = open(self.temp_file_path, "w")
class Consumer(object):
def __init__(self, addr, group, topic):
self.client = KafkaClient(addr)
self.consumer = SimpleConsumer(self.client, group, topic, max_buffer_size=1310720000)
self.temp_file_path = None
self.temp_file = None
self.hadoop_path = "/user/AdReport/%s/history" %(topic)
self.cached_path = "/user/AdReport/%s/cached" % (topic)
self.topic = topic
self.group = group
self.block_cnt = 0
def consume_topic(self, output_dir):
timestamp = time.strftime('%Y%m%d%H%M%S')
#open file for writing
self.temp_file_path = "%s/kafka_%s_%s_%s.dat" % (output_dir,
self.topic,
self.group,
timestamp)
self.temp_file = open(self.temp_file_path,"w")
print ( self.temp_file)
#one_entry = False
while True:
try:
messages = self.consumer.get_messages(count=10, block=False)
#OffsetAndMessage(offset=43, message=Message(magic=0,
# attributes=0, key=None, value='some message'))
for message in messages:
print (message)
#one_entry = True
#print (self.temp_file.tell())
self.temp_file.write(message.message.value + "\n")
if self.temp_file.tell() > 2000000:
self.save_to_hdfs(output_dir)
self.consumer.commit()
except:
self.consumer.seek(0, 2)
#if one_entry:
#print ("sending to hdfs")
#self.save_to_hdfs(output_dir, self.topic)
#self.consumer.commit()
def save_to_hdfs(self, output_dir):
print ("Saving file to hdfs")
self.temp_file.close()
print ("Closed open file")
timestamp = time.strftime('%Y%m%d%H%M%S')
hadoop_fullpath = "%s/%s_%s_%s.dat" % (self.hadoop_path, self.group,
self.topic, timestamp)
cached_fullpath = "%s/%s_%s_%s.dat" % (self.cached_path, self.group,
self.topic, timestamp)
#print ("Block " + str(self.block_cnt) + ": Saving file to HDFS " + hadoop_fullpath)
self.block_cnt += 1
# place blocked messages into history and cached folders on hdfs
os.system("sudo -u ubuntu /usr/local/hadoop/bin/hdfs dfs -put %s %s" % (self.temp_file_path,
hadoop_fullpath))
os.system("sudo -u ubuntu /usr/local/hadoop/bin/hdfs dfs -put %s %s" % (self.temp_file_path,
cached_fullpath))
os.remove(self.temp_file_path)
timestamp = time.strftime('%Y%m%d%H%M%S')
self.temp_file_path = "%s/kafka_%s_%s_%s.dat" % (output_dir,
self.topic,
self.group,
timestamp)
self.temp_file = open(self.temp_file_path, "w")
class Consumer(object):
def __init__(self, addr, group, topic):
self.client = KafkaClient(addr)
self.consumer = SimpleConsumer(self.client,
group,
topic,
max_buffer_size=1310720000)
self.temp_file_path = None
self.temp_file = None
self.hadoop_path = "/user/AdReport/%s/history" % (topic)
self.cached_path = "/user/AdReport/%s/cached" % (topic)
self.topic = topic
self.group = group
self.block_cnt = 0
def consume_topic(self, output_dir):
timestamp = time.strftime('%Y%m%d%H%M%S')
#open file for writing
self.temp_file_path = "%s/kafka_%s_%s_%s.dat" % (
output_dir, self.topic, self.group, timestamp)
self.temp_file = open(self.temp_file_path, "w")
print(self.temp_file)
#one_entry = False
while True:
try:
messages = self.consumer.get_messages(count=10, block=False)
#OffsetAndMessage(offset=43, message=Message(magic=0,
# attributes=0, key=None, value='some message'))
for message in messages:
print(message)
#one_entry = True
#print (self.temp_file.tell())
self.temp_file.write(message.message.value + "\n")
if self.temp_file.tell() > 2000000:
self.save_to_hdfs(output_dir)
self.consumer.commit()
except:
self.consumer.seek(0, 2)
#if one_entry:
#print ("sending to hdfs")
#self.save_to_hdfs(output_dir, self.topic)
#self.consumer.commit()
def save_to_hdfs(self, output_dir):
print("Saving file to hdfs")
self.temp_file.close()
print("Closed open file")
timestamp = time.strftime('%Y%m%d%H%M%S')
hadoop_fullpath = "%s/%s_%s_%s.dat" % (self.hadoop_path, self.group,
self.topic, timestamp)
cached_fullpath = "%s/%s_%s_%s.dat" % (self.cached_path, self.group,
self.topic, timestamp)
#print ("Block " + str(self.block_cnt) + ": Saving file to HDFS " + hadoop_fullpath)
self.block_cnt += 1
# place blocked messages into history and cached folders on hdfs
os.system("sudo -u ubuntu /usr/local/hadoop/bin/hdfs dfs -put %s %s" %
(self.temp_file_path, hadoop_fullpath))
os.system("sudo -u ubuntu /usr/local/hadoop/bin/hdfs dfs -put %s %s" %
(self.temp_file_path, cached_fullpath))
os.remove(self.temp_file_path)
timestamp = time.strftime('%Y%m%d%H%M%S')
self.temp_file_path = "%s/kafka_%s_%s_%s.dat" % (
output_dir, self.topic, self.group, timestamp)
self.temp_file = open(self.temp_file_path, "w")
# stdlib
from collections import defaultdict
# 3p
from kafka.client import KafkaClient
from kafka.consumer import SimpleConsumer
kafka_conn = KafkaClient("kafka:9092")
consumer = SimpleConsumer(kafka_conn,
"sample_check",
"test-topic",
auto_commit=True)
for message in consumer.get_messages(count=10):
print message.offset
consumer.commit()
class KafkaHelper(object):
def __init__(self):
self.client = None
self.producer = None
self.consumer = None
self.consumer_fetch_timeout = None
self.consumer_fetch_size = None
def __enter__(self):
self.get_client()
return self
def __exit__(self, exctype, excvalue, traceback):
self.close_client()
@retry(BrokerResponseError, tries=5, delay=3, backoff=2)
def get_client(self):
if not self.client:
self.client = KafkaClient(settings.KAFKA['host'])
return self.client
def get_producer(self):
"""
:return: SimpleProducer
"""
if not self.producer:
self.get_client()
self.producer = SimpleProducer(self.client)
return self.producer
def get_multiprocess_consumer(
self,
consumer_group,
topic,
fetch_size=settings.KAFKA['message_fetch_batch'],
fetch_timeout=settings.KAFKA['message_fetch_timeout'],
auto_commit_every_n=settings.KAFKA['auto_commit_msg_count'],
**kw):
"""
Return MultiProcessConsumer which consumes partitions for a topic in
parallel using multiple processes
Arguments:
consumer_group: a name for this consumer, used for offset storage and must be unique
topic: the topic to consume
Keyword Arguments:
fetch_size: Indicates the maximum number of messages to be fetched
fetch_timeout: The function will block for the specified
time (in seconds) until count messages is fetched
auto_commit_every_n: How many messages to consume
before a commit
"""
if not self.consumer:
self.consumer_fetch_size = fetch_size
self.consumer_fetch_timeout = fetch_timeout
self.get_client()
partitions = len(self.get_partitions(topic))
self.consumer = MultiProcessConsumer(
self.client,
consumer_group,
topic,
num_procs=partitions,
partitions_per_proc=1,
auto_commit_every_n=auto_commit_every_n,
**kw)
return self.consumer
def get_consumer(
self,
consumer_group,
topic,
fetch_size=settings.KAFKA['message_fetch_batch'],
fetch_timeout=settings.KAFKA['message_fetch_timeout'],
auto_commit_every_n=settings.KAFKA['auto_commit_msg_count'],
**kw):
if not self.consumer:
self.consumer_fetch_size = fetch_size
self.consumer_fetch_timeout = fetch_timeout
self.get_client()
self.consumer = SimpleConsumer(
self.client,
consumer_group,
topic,
auto_commit_every_n=auto_commit_every_n,
auto_offset_reset='smallest',
**kw)
return self.consumer
def close_client(self):
if self.client:
self.client.close()
def send_message(self, topic, msgs, logger=None):
content = [(json.dumps(msg) if type(msg) is dict else msg)
for msg in msgs]
try:
resp = self.producer.send_messages(topic, *content)
return resp
except Exception as e:
if logger:
logger.error(
'An error has occured in KafkaHelper.send_message(), please check errors: %s',
traceback.format_exc())
raise e
def receive_messages(self):
messages = self.consumer.get_messages(
count=self.consumer_fetch_size,
timeout=self.consumer_fetch_timeout)
return messages
def current_offset(self, topic, partition):
offsets, = self.client.send_offset_request(
[OffsetRequest(kafka_bytestring(topic), partition, -1, 1)])
return offsets.offsets[0]
def consumer_offset(self, consumer_name, topic, partition):
offsets, = self.client.send_offset_fetch_request(
consumer_name,
[OffsetRequest(kafka_bytestring(topic), partition, -1, 1)])
return offsets[2]
def get_total_lags(self, consumer_name, topic):
lags = []
lag = 0
partitions = self.get_partitions(topic)
for p in partitions:
offset1 = self.consumer_offset(consumer_name, topic, p)
offset2 = self.current_offset(topic, p)
lag = (offset2 - offset1)
lags.append(lag)
#print offset1,offset2,lag
return sum(lags)
def get_partitions(self, topic):
return self.client.get_partition_ids_for_topic(topic)
class KafkaMonitor:
def __init__(self, settings_name, unit_test=False):
'''
@param settings_name: the local settings file name
@param unit_test: whether running unit tests or not
'''
self.settings_name = settings_name
self.wrapper = SettingsWrapper()
self.logger = None
self.unit_test = unit_test
def _import_class(self, cl):
'''
Imports a class from a string
@param name: the module and class name in dot notation
'''
d = cl.rfind(".")
classname = cl[d+1:len(cl)]
m = __import__(cl[0:d], globals(), locals(), [classname])
return getattr(m, classname)
def _load_plugins(self):
'''
Sets up all plugins, defaults and settings.py
'''
plugins = self.settings['PLUGINS']
self.plugins_dict = {}
for key in plugins:
# skip loading the plugin if its value is None
if plugins[key] is None:
continue
# valid plugin, import and setup
self.logger.debug("Trying to load plugin {cls}".format(cls=key))
the_class = self._import_class(key)
instance = the_class()
instance._set_logger(self.logger)
if not self.unit_test:
instance.setup(self.settings)
the_schema = None
with open(self.settings['PLUGIN_DIR'] + instance.schema) as the_file:
the_schema = json.load(the_file)
mini = {}
mini['instance'] = instance
mini['schema'] = the_schema
self.plugins_dict[plugins[key]] = mini
self.plugins_dict = OrderedDict(sorted(self.plugins_dict.items(),
key=lambda t: t[0]))
def setup(self, level=None, log_file=None, json=None):
'''
Load everything up. Note that any arg here will override both
default and custom settings
@param level: the log level
@param log_file: boolean t/f whether to log to a file, else stdout
@param json: boolean t/f whether to write the logs in json
'''
self.settings = self.wrapper.load(self.settings_name)
my_level = level if level else self.settings['LOG_LEVEL']
# negate because logger wants True for std out
my_output = not log_file if log_file else self.settings['LOG_STDOUT']
my_json = json if json else self.settings['LOG_JSON']
self.logger = LogFactory.get_instance(json=my_json, stdout=my_output,
level=my_level,
name=self.settings['LOGGER_NAME'],
dir=self.settings['LOG_DIR'],
file=self.settings['LOG_FILE'],
bytes=self.settings['LOG_MAX_BYTES'],
backups=self.settings['LOG_BACKUPS'])
self.validator = self.extend_with_default(Draft4Validator)
def _setup_stats(self):
'''
Sets up the stats collection
'''
self.stats_dict = {}
redis_conn = redis.Redis(host=self.settings['REDIS_HOST'],
port=self.settings['REDIS_PORT'])
try:
redis_conn.info()
self.logger.debug("Connected to Redis in StatsCollector Setup")
except ConnectionError:
self.logger.warn("Failed to connect to Redis in StatsCollector"
" Setup, no stats will be collected")
return
if self.settings['STATS_TOTAL']:
self._setup_stats_total(redis_conn)
if self.settings['STATS_PLUGINS']:
self._setup_stats_plugins(redis_conn)
def _setup_stats_total(self, redis_conn):
'''
Sets up the total stats collectors
@param redis_conn: the redis connection
'''
self.stats_dict['total'] = {}
self.stats_dict['fail'] = {}
temp_key1 = 'stats:kafka-monitor:total'
temp_key2 = 'stats:kafka-monitor:fail'
for item in self.settings['STATS_TIMES']:
try:
time = getattr(StatsCollector, item)
self.stats_dict['total'][time] = StatsCollector \
.get_rolling_time_window(
redis_conn=redis_conn,
key='{k}:{t}'.format(k=temp_key1, t=time),
window=time,
cycle_time=self.settings['STATS_CYCLE'])
self.stats_dict['fail'][time] = StatsCollector \
.get_rolling_time_window(
redis_conn=redis_conn,
key='{k}:{t}'.format(k=temp_key2, t=time),
window=time,
cycle_time=self.settings['STATS_CYCLE'])
self.logger.debug("Set up total/fail Stats Collector '{i}'"\
.format(i=item))
except AttributeError as e:
self.logger.warning("Unable to find Stats Time '{s}'"\
.format(s=item))
total1 = StatsCollector.get_hll_counter(redis_conn=redis_conn,
key='{k}:lifetime'.format(k=temp_key1),
cycle_time=self.settings['STATS_CYCLE'],
roll=False)
total2 = StatsCollector.get_hll_counter(redis_conn=redis_conn,
key='{k}:lifetime'.format(k=temp_key2),
cycle_time=self.settings['STATS_CYCLE'],
roll=False)
self.logger.debug("Set up total/fail Stats Collector 'lifetime'")
self.stats_dict['total']['lifetime'] = total1
self.stats_dict['fail']['lifetime'] = total2
def _setup_stats_plugins(self, redis_conn):
'''
Sets up the plugin stats collectors
@param redis_conn: the redis connection
'''
self.stats_dict['plugins'] = {}
for key in self.plugins_dict:
plugin_name = self.plugins_dict[key]['instance'].__class__.__name__
temp_key = 'stats:kafka-monitor:{p}'.format(p=plugin_name)
self.stats_dict['plugins'][plugin_name] = {}
for item in self.settings['STATS_TIMES']:
try:
time = getattr(StatsCollector, item)
self.stats_dict['plugins'][plugin_name][time] = StatsCollector \
.get_rolling_time_window(
redis_conn=redis_conn,
key='{k}:{t}'.format(k=temp_key, t=time),
window=time,
cycle_time=self.settings['STATS_CYCLE'])
self.logger.debug("Set up {p} plugin Stats Collector '{i}'"\
.format(p=plugin_name, i=item))
except AttributeError:
self.logger.warning("Unable to find Stats Time '{s}'"\
.format(s=item))
total = StatsCollector.get_hll_counter(redis_conn=redis_conn,
key='{k}:lifetime'.format(k=temp_key),
cycle_time=self.settings['STATS_CYCLE'],
roll=False)
self.logger.debug("Set up {p} plugin Stats Collector 'lifetime'"\
.format(p=plugin_name))
self.stats_dict['plugins'][plugin_name]['lifetime'] = total
def _setup_kafka(self):
'''
Sets up kafka connections
'''
@MethodTimer.timeout(self.settings['KAFKA_CONN_TIMEOUT'], False)
def _hidden_setup():
try:
self.kafka_conn = KafkaClient(self.settings['KAFKA_HOSTS'])
self.kafka_conn.ensure_topic_exists(
self.settings['KAFKA_INCOMING_TOPIC'])
self.consumer = SimpleConsumer(self.kafka_conn,
self.settings['KAFKA_GROUP'],
self.settings['KAFKA_INCOMING_TOPIC'],
auto_commit=True,
iter_timeout=1.0)
except KafkaUnavailableError as ex:
message = "An exception '{0}' occured. Arguments:\n{1!r}" \
.format(type(ex).__name__, ex.args)
self.logger.error(message)
sys.exit(1)
return True
ret_val = _hidden_setup()
if ret_val:
self.logger.debug("Successfully connected to Kafka")
else:
self.logger.error("Failed to set up Kafka Connection within"
" timeout")
# this is essential to running the kafka monitor
sys.exit(1)
def extend_with_default(self, validator_class):
'''
Method to add default fields to our schema validation
( From the docs )
'''
validate_properties = validator_class.VALIDATORS["properties"]
def set_defaults(validator, properties, instance, schema):
for error in validate_properties(
validator, properties, instance, schema,
):
yield error
for property, subschema in properties.iteritems():
if "default" in subschema:
instance.setdefault(property, subschema["default"])
return validators.extend(
validator_class, {"properties": set_defaults},
)
def _main_loop(self):
'''
Continuous loop that reads from a kafka topic and tries to validate
incoming messages
'''
self.logger.debug("Processing messages")
old_time = 0
while True:
self._process_messages()
if self.settings['STATS_DUMP'] != 0:
new_time = int(time.time() / self.settings['STATS_DUMP'])
# only log every X seconds
if new_time != old_time:
self._dump_stats()
old_time = new_time
time.sleep(.01)
def _process_messages(self):
try:
for message in self.consumer.get_messages():
if message is None:
self.logger.debug("no message")
break
try:
self._increment_total_stat(message.message.value)
the_dict = json.loads(message.message.value)
found_plugin = False
for key in self.plugins_dict:
obj = self.plugins_dict[key]
instance = obj['instance']
schema = obj['schema']
try:
self.validator(schema).validate(the_dict)
found_plugin = True
self._increment_plugin_stat(
instance.__class__.__name__,
the_dict)
ret = instance.handle(the_dict)
# break if nothing is returned
if ret is None:
break
except ValidationError:
pass
if not found_plugin:
extras = {}
extras['parsed'] = True
extras['valid'] = False
extras['data'] = the_dict
self.logger.warn("Did not find schema to validate "
"request", extra=extras)
self._increment_fail_stat(the_dict)
except ValueError:
extras = {}
extras['parsed'] = False
extras['valid'] = False
extras['data'] = message.message.value
self.logger.warning('Unparseable JSON Received',
extra=extras)
self._increment_fail_stat(message.message.value)
except OffsetOutOfRangeError:
# consumer has no idea where they are
self.consumer.seek(0, 2)
self.logger.error("Kafka offset out of range error")
def _increment_total_stat(self, string):
'''
Increments the total stat counters
@param string: the loaded message object for the counter
'''
string = string + str(time.time())
if 'total' in self.stats_dict:
self.logger.debug("Incremented total stats")
for key in self.stats_dict['total']:
if key == 'lifetime':
self.stats_dict['total'][key].increment(string)
else:
self.stats_dict['total'][key].increment()
def _increment_fail_stat(self, item):
'''
Increments the total stat counters
@param item: the loaded message object for HLL counter
'''
if isinstance(item, dict):
item['ts'] = time.time()
elif isinstance(item, str):
item = item + str(time.time())
if 'fail' in self.stats_dict:
self.logger.debug("Incremented fail stats")
for key in self.stats_dict['fail']:
if key == 'lifetime':
self.stats_dict['fail'][key].increment(item)
else:
self.stats_dict['fail'][key].increment()
def _increment_plugin_stat(self, name, item):
'''
Increments the total stat counters
@param name: The formal name of the plugin
@param dict: the loaded message object for HLL counter
'''
item['ts'] = time.time()
if 'plugins' in self.stats_dict:
self.logger.debug("Incremented plugin '{p}' plugin stats"\
.format(p=name))
for key in self.stats_dict['plugins'][name]:
if key == 'lifetime':
self.stats_dict['plugins'][name][key].increment(item)
else:
self.stats_dict['plugins'][name][key].increment()
def _dump_stats(self):
'''
Dumps the stats out
'''
extras = {}
if 'total' in self.stats_dict:
self.logger.debug("Compiling total/fail dump stats")
for key in self.stats_dict['total']:
final = 'total_{t}'.format(t=key)
extras[final] = self.stats_dict['total'][key].value()
for key in self.stats_dict['fail']:
final = 'fail_{t}'.format(t=key)
extras[final] = self.stats_dict['fail'][key].value()
if 'plugins' in self.stats_dict:
self.logger.debug("Compiling plugin dump stats")
for name in self.stats_dict['plugins']:
for key in self.stats_dict['plugins'][name]:
final = 'plugin_{n}_{t}'.format(n=name, t=key)
extras[final] = self.stats_dict['plugins'][name][key].value()
if not self.logger.json:
self.logger.info('Kafka Monitor Stats Dump:\n{0}'.format(
json.dumps(extras, indent=4, sort_keys=True)))
else:
self.logger.info('Kafka Monitor Stats Dump', extra=extras)
def run(self):
'''
Set up and run
'''
self._setup_kafka()
self._load_plugins()
self._setup_stats()
self._main_loop()
def feed(self, json_item):
'''
Feeds a json item into the Kafka topic
@param json_item: The loaded json object
'''
@MethodTimer.timeout(self.settings['KAFKA_FEED_TIMEOUT'], False)
def _feed(json_item):
try:
self.kafka_conn = KafkaClient(self.settings['KAFKA_HOSTS'])
topic = self.settings['KAFKA_INCOMING_TOPIC']
producer = SimpleProducer(self.kafka_conn)
except KafkaUnavailableError:
self.logger.error("Unable to connect to Kafka")
return False
if not self.logger.json:
self.logger.info('Feeding JSON into {0}\n{1}'.format(
topic, json.dumps(json_item, indent=4)))
else:
self.logger.info('Feeding JSON into {0}\n'.format(topic),
extra={'value': json_item})
self.kafka_conn.ensure_topic_exists(topic)
producer.send_messages(topic, json.dumps(json_item))
return True
result = _feed(json_item)
if result:
self.logger.info("Successfully fed item to Kafka")
else:
self.logger.error("Failed to feed item into Kafka")
class ZKConsumer(object):
zk_timeout = 30
jitter_seconds = 30
broker_prefix = '/brokers/ids'
def __init__(
self,
zk_hosts,
group,
topic,
nodes,
zk_handler=None,
logger=None,
identifier=None,
**consumer_kwargs):
"""Creates a Consumer that tracks state in ZooKeeper,
rebalancing partition ownership as registered consumers change.
NOTE: this class is intended for version 0.8.1 of Kafka, where offsets
are managed by Kafka but there is no rebalancing in the protocol.
"""
if logger is None:
logger = logging.getLogger('kafka.consumer.ZKConsumer')
self.logger = logger
self.identifier = identifier
if KafkaClient is None:
raise RuntimeError("Kafka support requires cs.eyrie to be installed with the Kafka extra: install_requires= ['cs.eyrie[Kafka]']")
self.zk_handler = zk_handler
self.zk_hosts = zk_hosts
self.broker_hosts = []
self.group = group
self.topic = topic
self.zk = None
self.nodes = nodes
self.client = None
self.consumer = None
self.consumer_kwargs = consumer_kwargs
# This will kick off a cascading sequence to initialize ourselves:
# 1. Connect to ZK and pull list of Kafka brokers
# 2. Register ourselves as a consumer in ZK
# 3. Rebalance partitions across all connected consumers
self.init_zk()
def zk_session_watch(self, state):
self.logger.debug('ZK transitioned to: %s', state)
if state == KazooState.SUSPENDED:
if self.consumer is not None:
self.logger.info('Stopping Kafka consumer')
self.consumer.stop()
self.consumer = None
# Lost connection to ZK; we can't call any methods that would
# try to contact it (i.e., we can't do self.zkp.finish() )
self.zkp = None
elif state == KazooState.CONNECTED:
self.logger.info('Restarting ZK partitioner')
self.zk.handler.spawn(self.init_zkp)
def _zkp_wait(self):
handler = self.zk.handler
while 1:
if self.zkp.failed:
self.logger.warning("Lost or unable to acquire partition")
self.stop()
elif self.zkp.release:
self.zkp.release_set()
elif self.zkp.acquired:
def group_change_proxy(event):
self.logger.warn('Connected consumers changed')
if self.zkp is None:
self.logger.info('Restarting ZK partitioner')
handler.spawn(self.init_zkp)
elif self.zkp is not None and self.zkp.failed:
self.logger.warning("Lost or unable to acquire partition")
self.stop()
else:
self.logger.info('Scheduling ZK partitioner set release')
rel_greenlet = handler.spawn(self.zkp.release_set)
self.logger.info('Scheduling group re-join')
rel_greenlet.link_value(lambda greenlet: self.zkp.join_group)
if not self.nodes:
self.logger.info('Partitioner aquired; setting child watch')
result = self.zk.get_children_async(self.zkp._group_path)
result.rawlink(group_change_proxy)
# Break out of while loop to begin consuming events
break
elif self.zkp.allocating:
self.zkp.wait_for_acquire()
def init_zkp(self):
if not hasattr(self, 'zkp') or self.zkp is None:
if self.nodes:
self.zkp = StaticZKPartitioner(
self.zk, self.group, self.topic, self.nodes,
partitions_changed_cb=self.init_consumer,
logger=self.logger, identifier=self.identifier)
else:
self.zkp = ZKPartitioner(
self.zk, self.group, self.topic,
time_boundary=self.jitter_seconds,
partitions_changed_cb=self.init_consumer,
logger=self.logger, identifier=self.identifier)
self._zkp_wait()
def init_zk(self):
# TODO: switch to async
# 1. implement kazoo.interfaces.IHandler in terms of Tornado's IOLoop
self.zk = KazooClient(hosts=self.zk_hosts, handler=self.zk_handler)
self.zk.start()
self.zk.add_listener(self.zk_session_watch)
@self.zk.ChildrenWatch(self.broker_prefix)
def broker_change_proxy(broker_ids):
self.onBrokerChange(broker_ids)
self.init_zkp()
def onBrokerChange(self, broker_ids):
self.broker_hosts = []
for b_id in broker_ids:
b_json, zstat = self.zk.get('/'.join([self.broker_prefix, b_id]))
b_data = json.loads(b_json)
self.broker_hosts.append('{}:{}'.format(b_data['host'],
b_data['port']))
my_partitions = []
if self.consumer is not None:
self.logger.warn('Brokers changed, stopping Kafka consumer.')
my_partitions = self.consumer.offsets.keys()
self.consumer.stop()
self.consumer = None
if self.client is not None:
self.logger.warn('Brokers changed, stopping Kafka client.')
self.client.close()
self.client = None
if my_partitions:
msg = 'Brokers changed, queuing restart of Kafka client / consumer.'
self.logger.warn(msg)
self.zk.handler.spawn(self.init_consumer, my_partitions)
def init_consumer(self, my_partitions):
if self.consumer is None:
self.logger.warn('Starting Kafka client')
self.client = KafkaClient(self.broker_hosts,
client_id=self.zkp._identifier)
else:
if self.consumer is None or \
sorted(my_partitions) != sorted(self.consumer.offsets.keys()):
self.logger.warn('Partitions changed, restarting Kafka consumer.')
self.consumer.stop()
else:
self.logger.info('Partitions unchanged, not restarting Kafka consumer.')
return
self.consumer = SimpleConsumer(self.client, self.group, self.topic,
partitions=my_partitions,
**self.consumer_kwargs)
self.consumer.provide_partition_info()
self.logger.info("Consumer connected to Kafka: %s", self.consumer.offsets)
def stop(self):
if self.consumer is not None:
self.logger.info('Stopping Kafka consumer')
self.consumer.stop()
self.consumer = None
if self.client is not None:
self.logger.info('Stopping Kafka client')
self.client.close()
self.client = None
if self.zk is not None:
self.logger.info('Stopping ZooKeeper client')
if self.zkp is not None and not self.zkp.failed:
self.zkp.finish()
self.zk.stop()
self.zkp = None
self.zk = None
def commit(self, partitions=None):
"""
Commit offsets for this consumer
partitions: list of partitions to commit, default is to commit
all of them
"""
if self.consumer is None:
return
self.logger.debug('Begin committing offsets for partitions: %s',
partitions if partitions else 'All')
self.consumer.commit(partitions)
self.logger.debug('End committing offsets for partitions: %s',
partitions if partitions else 'All')
def pending(self, partitions=None):
"""
Gets the pending message count
partitions: list of partitions to check for, default is to check all
"""
return self.consumer.pending(partitions)
def provide_partition_info(self):
"""
Indicates that partition info must be returned by the consumer
"""
self.consumer.provide_partition_info()
def seek(self, offset, whence):
"""
Alter the current offset in the consumer, similar to fseek
offset: how much to modify the offset
whence: where to modify it from
0 is relative to the earliest available offset (head)
1 is relative to the current offset
2 is relative to the latest known offset (tail)
"""
self.consumer.seek(offset, whence)
def get_messages(self, count=1, block=True, timeout=0.1):
"""
Fetch the specified number of messages
count: Indicates the maximum number of messages to be fetched
block: If True, the API will block till some messages are fetched.
timeout: If block is True, the function will block for the specified
time (in seconds) until count messages is fetched. If None,
it will block forever.
"""
if self.consumer is None:
return []
else:
try:
messages = self.consumer.get_messages(count, block, timeout)
if not messages and self.zkp.failed:
raise FailedPayloadsError
return messages
except FailedPayloadsError as err:
msg = 'Failed to retrieve payload, restarting consumer'
self.logger.exception(msg)
raise err
def get_message(self, block=True, timeout=0.1, get_partition_info=None):
return self.consumer.get_message(block, timeout, get_partition_info)
def _get_message(self, block=True, timeout=0.1, get_partition_info=None,
update_offset=True):
return self.consumer._get_message(block, timeout, get_partition_info,
update_offset)
def __iter__(self):
for msg in self.consumer:
yield msg
from kafka.client import KafkaClient
from kafka.common import OffsetRequest
from kafka.util import kafka_bytestring
from kafka.consumer import MultiProcessConsumer
from kafka.producer import SimpleProducer
from kafka.consumer import SimpleConsumer
from kafka.common import BrokerResponseError
cli = KafkaClient("localhost:9092")
consumer = SimpleConsumer(cli, 'test', 'tp_test1', auto_commit_every_n=10)
try:
no_msg_times = 0
while 1:
is_over = False
messages = consumer.get_messages(count=5, timeout=3)
if messages:
for m in messages:
#print m
msg_value = m.message.value
print msg_value
if msg_value == 'over':
is_over = True
else:
print "no msg!"
no_msg_times += 1
if is_over:
print "The show is over! bye..."
break
if no_msg_times >= 5:
class Consumer(object):
def __init__(self, addr, group, topic):
"""Initialize Consumer with kafka broker IP, group, and topic."""
self.client = KafkaClient(addr)
self.consumer = SimpleConsumer(self.client,
group,
topic,
max_buffer_size=1310720000)
self.temp_file_path = None
self.temp_file = None
self.hadoop_path = "/insight/artsy/geo"
self.topic = topic
self.group = group
self.block_cnt = 0
def consume_topic(self, output_dir):
"""Consumes a stream of messages from the "post_geo_activity" topic.
Code template from https://github.com/ajmssc/bitcoin-inspector.git
"""
timestamp = time.strftime('%Y%m%d%H%M%S')
# open file for writing
self.temp_file_path = "%s/kafka_%s_%s_%s.dat" % (
output_dir, self.topic, self.group, timestamp)
self.temp_file = open(self.temp_file_path, "w")
while True:
try:
# get 1000 messages at a time, non blocking
messages = self.consumer.get_messages(count=1000, block=False)
for message in messages:
self.temp_file.write(message.message.value + "\n")
# file size > 20MB
if self.temp_file.tell() > 20000000:
self.flush_to_hdfs(output_dir)
self.consumer.commit()
except:
# move to tail of kafka topic if consumer is referencing
# unknown offset
self.consumer.seek(0, 2)
def flush_to_hdfs(self, output_dir):
"""Flushes the 20MB file into HDFS."""
self.temp_file.close()
timestamp = time.strftime('%Y%m%d%H%M%S')
hadoop_fullpath = "%s/%s_%s_%s.dat" % (self.hadoop_path, self.group,
self.topic, timestamp)
print "Block {}: Flushing data file to HDFS => {}".format(
str(self.block_cnt), hadoop_fullpath)
self.block_cnt += 1
os.system(
"hdfs dfs -put %s %s" %
(self.temp_file_path, hadoop_fullpath)) # save from local to hdfs
os.remove(self.temp_file_path) # remove temp local file
timestamp = time.strftime('%Y%m%d%H%M%S')
self.temp_file_path = "%s/kafka_%s_%s_%s.dat" % (
output_dir, self.topic, self.group, timestamp)
self.temp_file = open(self.temp_file_path, "w")
class Consumer(object):
"""Kafka consumer class with functions to consume messages to HDFS.
Messages are blocked into 20MB files and transferred to HDFS
Attributes:
client: string representing IP:port of the kafka broker
consumer: Consumer object specifying the client group, and topic
temp_file_path: location of the 20MB file to be appended to before
transfer to HDFS
temp_file: File object opened from temp_file_path
topic: String representing the topic on Kafka
group: String representing the Kafka consumer group to be associated
with
block_cnt: integer representing the block count for print statements
"""
def __init__(self, addr, group, topic):
"""Initialize Consumer with kafka broker IP, group, and topic."""
self.client = KafkaClient(addr)
self.consumer = SimpleConsumer(self.client,
group,
topic,
max_buffer_size=1310720000)
self.temp_file_path = None
self.temp_file = None
self.hadoop_path = "/user/parking_data/history"
self.topic = topic
self.group = group
self.block_cnt = 0
def consume_topic(self, output_dir):
"""Consumes a stream of messages from the "messages" topic.
Code template from https://github.com/ajmssc/bitcoin-inspector.git
Args:
output_dir: string representing the directory to store the 20MB
before transferring to HDFS
Returns:
None
"""
timestamp = time.strftime('%Y%m%d%H%M%S')
# open file for writing
self.temp_file_path = "%s/kafka_%s_%s_%s.dat" % (
output_dir, self.topic, self.group, timestamp)
self.temp_file = open(self.temp_file_path, "w")
# while True:
for ii in range(0, 2):
try:
# get 1000 messages at a time, non blocking
messages = self.consumer.get_messages(count=1000, block=False)
# OffsetAndMessage(offset=43, message=Message(magic=0,
# attributes=0, key=None, value='some message'))
for message in messages:
self.temp_file.write(message.message.value + "\n")
# file size > 20MB
if self.temp_file.tell() > 20000000:
self.flush_to_hdfs(output_dir)
self.consumer.commit()
except:
# move to tail of kafka topic if consumer is referencing
# unknown offset
self.consumer.seek(0, 2)
def flush_to_hdfs(self, output_dir):
"""Flushes the 20MB file into HDFS.
Code template from https://github.com/ajmssc/bitcoin-inspector.git
Flushes the file into HDFS folders
Args:
output_dir: string representing the directory to store the 20MB
before transferring to HDFS
Returns:
None
"""
self.temp_file.close()
timestamp = time.strftime('%Y%m%d%H%M%S')
hadoop_fullpath = "%s/%s_%s_%s.dat" % (self.hadoop_path, self.group,
self.topic, timestamp)
print("Block {}: Flushing 20MB file to HDFS => {}".format(
str(self.block_cnt), hadoop_fullpath))
self.block_cnt += 1
# place blocked messages into history and cached folders on hdfs
print("hdfs dfs -put %s %s" % (self.temp_file_path, hadoop_fullpath))
os.system("sudo hdfs dfs -put %s %s" %
(self.temp_file_path, hadoop_fullpath))
# os.system("sudo -u hdfs hdfs dfs -put %s %s" % (self.temp_file_path,
# cached_fullpath))
os.remove(self.temp_file_path)
timestamp = time.strftime('%Y%m%d%H%M%S')
self.temp_file_path = "%s/kafka_%s_%s_%s.dat" % (
output_dir, self.topic, self.group, timestamp)
self.temp_file = open(self.temp_file_path, "w")
class KafkaDatawakeLookaheadSpout(Spout):
group = 'datawake-crawler-out-consumer'.encode()
def __init__(self):
Spout.__init__(self)
self.queue = None
def initialize(self, stormconf, context):
try:
self.settings = all_settings.get_settings(
stormconf['topology.deployment'])
self.topic = self.settings['crawler-out-topic'].encode()
self.conn_pool = self.settings['crawler_conn_pool'].encode()
self.log('KafkaDatawakeLookaheadSpout initialized with topic =' +
self.topic + ' conn_pool=' + self.conn_pool)
self.kafka = KafkaClient(self.conn_pool)
self.kafka.ensure_topic_exists(self.topic)
self.consumer = SimpleConsumer(self.kafka,
self.group,
self.topic,
max_buffer_size=None)
self.consumer.seek(0, 2) # move to the tail of the queue
except:
self.log("KafkaDatawakeLookaheadSpout initialize error",
level='error')
self.log(traceback.format_exc(), level='error')
raise
def next_tuple(self):
"""
input message:
dict(
crawlid = input['crawlid'],
appid = input['appid'],
url = url,
status_code = response.getcode(),
status_msg = 'Success',
timestamp = response.info()['date'],
links_found = links,
body = html,
attrs = input['attrs']
)
:return: (url, status, headers, flags, body, timestamp, source,context)
"""
offsetAndMessage = self.consumer.get_messages(timeout=None)[0]
message = offsetAndMessage.message.value
crawled = json.loads(message)
if crawled['appid'] == self.settings["appid"]:
safeurl = crawled['url'].encode('utf-8', 'ignore')
self.log("Lookahead spout received id: " + crawled['crawlid'] +
" url: " + safeurl)
context = {
'source': 'datawake-lookahead',
'domain': crawled['attrs']['domain']
}
self.emit([
crawled['url'], crawled['status_code'], '', '',
crawled['body'], crawled['timestamp'], context['source'],
context
])
class Consumer(object):
def __init__(self, addr, group, topic):
self.client = KafkaClient(addr)
self.consumer = SimpleConsumer(self.client, group, topic,
max_buffer_size=1310720000)
self.temp_file_path = None
self.temp_file = None
self.topic = topic
self.group = group
self.block_cnt = 0
os.system ( "hdfs dfs -mkdir /data2" )
def consume_topic(self, output_dir):
if not os.path.isdir ( output_dir ): os.makedirs ( output_dir )
timestamp = time.strftime('%Y%m%d%H%M%S')
self.temp_file_path = "%s/kafka_%s_%s_%s.dat" % (output_dir,
self.topic,
self.group,
timestamp)
self.temp_file = open(self.temp_file_path,"w")
while True:
try:
# get 1000 messages at a time, non blocking
messages = self.consumer.get_messages(count=1000, block=False)
# OffsetAndMessage(offset=43, message=Message(magic=0,
# attributes=0, key=None, value='some message'))
for message in messages:
self.temp_file.write(message.message.value + "\n")
# file size > 40MB
if self.temp_file.tell() > 40000000:
self.flush_to_hdfs(output_dir)
self.consumer.commit()
except:
# move to tail of kafka topic if consumer is referencing
# unknown offset
self.consumer.seek(0, 2)
def flush_to_hdfs(self, output_dir):
self.temp_file.close()
timestamp = time.strftime('%Y%m%d%H%M%S')
print "Block {}: Flushing 40MB file to HDFS => /data2".format(str(self.block_cnt))
self.block_cnt += 1
# place blocked messages into history and cached folders on hdfs
os.system("hdfs dfs -copyFromLocal %s %s" % (self.temp_file_path,
"/data2"))
os.remove(self.temp_file_path)
timestamp = time.strftime('%Y%m%d%H%M%S')
self.temp_file_path = "%s/kafka_%s_%s_%s.dat" % (output_dir,
self.topic,
self.group,
timestamp)
self.temp_file = open(self.temp_file_path, "w")
