def main():
producer = KafkaProducer(bootstrap_servers=config.KAFKA_SERVERS,
value_serializer=lambda v: json.dumps(v).encode())
headers = [
'CMTE_ID', 'AMNDT_IND', 'RPT_TP', 'TRANSACTION_PGI', 'IMAGE_NUM',
'TRANSACTION_TP', 'ENTITY_TP', 'NAME', 'CITY', 'STATE', 'ZIP_CODE',
'EMPLOYER', 'OCCUPATION', 'TRANSACTION_DT', 'TRANSACTION_AMT',
'OTHER_ID', 'TRAN_ID', 'FILE_NUM', 'MEMO_CD', 'MEMO_TEXT', 'SUB_ID'
]
with open('/home/ubuntu/2016/by_date/itcont_2016_10151005_20150726.txt'
) as f:
data = f.readlines()
data = [x.split("|") for x in data]
#producer = KafkaProducer(bootstrap_servers = '18.205.181.166:9092',value_serializer = lambda v: json.dumps(v).encode('utf-8'))
for row in data:
row = {h: x for h, x in zip(headers, row)}
#print row
producer.send('data', row)
#print row
producer.flush()
producer = KafkaProducer(retries=5)
class KafkaLoggingHandler(logging.Handler):
def __init__(self, hosts_list, topic, **kwargs):
logging.Handler.__init__(self)
self.kafka_topic_name = topic
self.producer = KafkaProducer(bootstrap_servers=hosts_list)
def emit(self, record):
# drop kafka logging to avoid infinite recursion
if record.name == 'kafka':
return
try:
# use default formatting
msg = self.format(record)
msg = str.encode(msg)
self.producer.send(self.kafka_topic_name, msg)
self.producer.flush()
except (KeyboardInterrupt, SystemExit):
raise
except:
self.handleError(record)
def close(self):
if self.producer is not None:
self.producer.close()
logging.Handler.close(self)
def run(self):
producer = KafkaProducer(bootstrap_servers='DNS from master:9092')
for i in x:
producer.send('my_topic', b'i')
producer.flush()
time.sleep(0.5)
while not self.stop_event.is_set():
s3 = boto3.client('s3') #low-level functional API
my_bucket = resource.Bucket(BUCKET_NAME)
for datum in docs:
self.producer.send('docs', datum)
time.sleep(1)
producer.close()
def main():
producer = KafkaProducer(bootstrap_servers=["localhost:9092"],value_serializer= lambda v: json.dumps(v).encode())
headers = ['CMTE_ID', 'AMNDT_IND', 'RPT_TP', 'TRANSACTION_PGI', 'IMAGE_NUM', \
'TRANSACTION_TP','ENTITY_TP','NAME', 'CITY', 'STATE', 'ZIP_CODE', \
'EMPLOYER', 'OCCUPATION', 'TRANSACTION_DT','TRANSACTION_AMT', \
'OTHER_ID', 'TRAN_ID', 'FILE_NUM', 'MEMO_CD', 'MEMO_TEXT', 'SUB_ID' ]
split_counter = len(glob.glob('/home/ubuntu/manip_data/split_*'))
for j in range(1):
for i in range(split_counter):
with open('/home/ubuntu/manip_data/split_'+file_number(i)) as f:
reader = csv.reader(f, delimiter='|')
for row in reader:
row = {h:x for h,x in zip(headers,row)}
producer.send('datatwo', row)
producer.flush()
producer.close()
def send_avro_record_to_kafka(topic, value, bootstrap_servers,
avro_schema_json):
value_schema = avro.schema.parse(avro_schema_json)
producer = KafkaProducer(bootstrap_servers=bootstrap_servers)
writer = DatumWriter(value_schema)
bytes_writer = io.BytesIO()
encoder = BinaryEncoder(bytes_writer)
writer.write(value, encoder)
try:
producer.send(topic=topic, value=bytes_writer.getvalue())
except Exception as e:
print(
f"Exception while producing record value - {value} to topic - {topic}: {e}"
)
else:
print(
f"Successfully producing record value - {value} to topic - {topic}"
)
producer.flush()
class KafkaEventProducer:
FLUSH_PER = 10000
gobmodel = GOBModel()
def __init__(self, catalogue: str, collection: str, logger):
self.catalogue = catalogue
self.collection = collection
self.logger = logger
self.gob_db_session = None
self.db_session = None
self.gob_db_base = None
self.Event = None
self.producer = None
self.total_cnt = 0
self._init_connections()
self.event_builder = EventDataBuilder(self.gob_db_session,
self.gob_db_base, catalogue,
collection)
def _get_tables_to_reflect(self):
"""Returns tables to reflect:
- events
- object table (e.g. gebieden_buurten)
- relation tables (e.g. rel_gb_brt_gbd_wijk_ligt_in_wijk, ...)
:return:
"""
relations = get_relations_for_collection(self.gobmodel, self.catalogue,
self.collection)
relation_tables = [
self.gobmodel.get_table_name('rel', rel_table)
for rel_table in relations.values()
]
return [
'events',
self.gobmodel.get_table_name(self.catalogue, self.collection)
] + relation_tables
def _init_gob_db_session(self):
"""Inits db session for gob db (to access events)
:return:
"""
engine = create_engine(URL(**GOB_DATABASE_CONFIG),
connect_args={'sslmode': 'require'})
self.gob_db_session = Session(engine)
meta = MetaData()
meta.reflect(engine, only=self._get_tables_to_reflect())
base = automap_base(metadata=meta)
base.prepare()
self.Event = base.classes.events
self.gob_db_base = base
self.logger.info("Initialised events storage")
def _init_local_db_session(self):
"""Inits db session for local (gob_kafka) db
:return:
"""
engine = create_engine(URL(**DATABASE_CONFIG),
connect_args={'sslmode': 'require'})
Base.metadata.bind = engine
self.db_session = Session(engine)
def _init_kafka(self):
self.producer = KafkaProducer(
**KAFKA_CONNECTION_CONFIG,
max_in_flight_requests_per_connection=1,
# With retries, max_in_flight should always be 1 to ensure ordering of batches!
retries=3)
self.logger.info("Initialised Kafka connection")
def _init_connections(self):
self._init_gob_db_session()
self._init_local_db_session()
self._init_kafka()
def _get_last_event(self):
last_event = self.db_session \
.query(LastSentEvent) \
.filter_by(catalogue=self.catalogue, collection=self.collection) \
.first()
return last_event
def _get_last_eventid(self):
last_event = self._get_last_event()
return last_event.last_event if last_event else -1
def _set_last_eventid(self, eventid: int):
last_event = self._get_last_event()
if last_event:
last_event.last_event = eventid
else:
last_event = LastSentEvent(catalogue=self.catalogue,
collection=self.collection,
last_event=eventid)
self.db_session.add(last_event)
self.db_session.commit()
def _get_events(self, min_eventid: int):
return self.gob_db_session \
.query(self.Event) \
.yield_per(10000) \
.filter(and_(self.Event.catalogue == self.catalogue, self.Event.entity == self.collection,
self.Event.eventid > min_eventid)) \
.order_by(self.Event.eventid.asc())
def _add_event(self, event):
header = {
'event_type': event.action,
'event_id': event.eventid,
'tid': event.tid,
'catalog': event.catalogue,
'collection': event.entity,
}
headers = [(k, _to_bytes(str(v)) if v else b'')
for k, v in header.items()]
data = self.event_builder.build_event(event.tid)
self.producer.send(KAFKA_TOPIC,
key=_to_bytes(header['tid']),
value=_to_bytes(json.dumps(data)),
headers=headers)
def _flush(self, last_eventid: int):
self.producer.flush(timeout=120)
self._set_last_eventid(last_eventid)
print(
f"Flushed Kafka events. Total events: {self.total_cnt}. Last event id: {last_eventid}"
)
def produce(self):
last_eventid = self._get_last_eventid()
self.logger.info(f"Start producing. Last event was {last_eventid}")
events = self._get_events(last_eventid)
for event in events:
self._add_event(event)
self.total_cnt += 1
last_eventid = event.eventid
self.gob_db_session.expunge(event)
if self.total_cnt % self.FLUSH_PER == 0:
self._flush(last_eventid)
self._flush(last_eventid)
self.logger.info(f"Produced {self.total_cnt} Kafka events")
from kafka.producer import KafkaProducer
import json
import csv
from time import sleep
producer = KafkaProducer(
bootstrap_servers='localhost:9092',
value_serializer=lambda v: json.dumps(v).encode('utf-8'))
with open('blacklist.csv') as file:
reader = csv.DictReader(file, delimiter=",")
for row in reader:
producer.send(topic='blacklist', value=row)
producer.flush()
sleep(1)
class AIMSDownsamplingTCPServerConsumer:
LOG_FORMAT ="{} UTC_TS\t"\
"{}"
INTERVAL = 60
DELAY = 0
MAX_CONNECTION = 32
def __init__(self, kafka_host, kafka_port, tcp_host, tcp_port, topic,
log_topic):
self.kafka_host = kafka_host
self.kafka_port = kafka_port
self.tcp_host = tcp_host
self.tcp_port = tcp_port
self.topic = topic
self.log_topic = log_topic
self.consumer = KafkaConsumer(
topic,
bootstrap_servers=["{}:{}".format(kafka_host, kafka_port)],
enable_auto_commit=False,
max_poll_records=1024 * 1024,
max_partition_fetch_bytes=1024 * 1024 * 100)
self.producer = KafkaProducer(
bootstrap_servers=["{}:{}".format(kafka_host, kafka_port)])
self.connections = {}
self.sample_end_time = self.get_end_time(time())
self.lastPolled = []
def run(self):
self.log("running")
asyncio.run(self._async_run())
async def _async_run(self):
tcpServer = await asyncio.start_server(self.connection_handler,
self.tcp_host, self.tcp_port)
await asyncio.gather(tcpServer.serve_forever(), self.poll_from_kafka())
async def connection_handler(self, reader, writer):
addr = str(writer.get_extra_info("peername"))
addr = str(writer.get_extra_info("peername"))
# A new connection, but we can accept no more
if addr not in self.connections and \
len(self.connections)>=self.MAX_CONNECTION:
self.refuse_client(addr, writer)
return
# Add connection
self.add_client(addr, writer)
# Read data from connection
remaining_data = b""
try:
while True:
data = await reader.read(1) # 1024*8 bytes
if not data:
break
except BrokenPipeError:
"""
Catches connecton reset by peer when we are sending the batched data,
which is also when we cannot check for reader. The broken connection
on the writer side will ultimately lead to BrokenPipeError on the
reader side. Hence
"""
pass
finally:
self.remove_client(addr)
async def poll_from_kafka(self):
polled = self.consumer.poll(timeout_ms=self.INTERVAL * 1000 / 2)
self.lastPolled = polled
while True:
t = time()
if t >= self.sample_end_time + self.DELAY:
polled = self.consumer.poll(timeout_ms=self.INTERVAL * 1000 /
2)
lastPolled = self.lastPolled
start_time = self.sample_end_time - self.INTERVAL
end_time = self.sample_end_time
self.lastPolled = polled
self.sample_end_time = self.get_end_time(time())
if len(self.connections) != 0:
# run on lastPolled first to hit the cache
parsed_records = self.get_parsed_records(lastPolled) + \
self.get_parsed_records(polled)
parsed_records = list(
filter(
lambda rec: rec["observation_date_time"] is
not None, parsed_records))
ds_records = self.down_sample(parsed_records, start_time,
end_time)
messages = [rec["message"] for rec in ds_records]
for addr in self.connections.keys():
await self.send_or_ignore_message(addr, messages)
await asyncio.sleep(0.1)
def get_parsed_records(self, polled):
# Create cache
if "_get_parsed_records__polled" not in self.__dict__:
self._get_parsed_records__polled = []
if "_get_parsed_records__ret" not in self.__dict__:
self._get_parsed_records__ret = []
# Cache hit
if self._get_parsed_records__polled == polled:
return self._get_parsed_records__ret
# Cache not hit
self._get_parsed_records__polled = polled
self._get_parsed_records__ret = []
records = []
for recordList in polled.values():
records.extend([rec.value for rec in recordList])
for rec in records:
self._get_parsed_records__ret.append(self.parse_hl7(rec))
return self._get_parsed_records__ret
def parse_hl7(self, message):
segments = message.decode(errors="ignore") \
.strip() \
.split(MESSAGE_SEGMENT_END_BYTE)
location = None
date_time = None
observation_types = []
observation_type = None
for seg in segments:
fields = seg.split('|')
if fields[0] == "PV1":
try:
location = fields[3]
except IndexError:
pass
if fields[0] == "OBR":
try:
date_time = mktime(strptime(fields[7], "%Y%m%d%H%M%S"))
except IndexError:
pass
if fields[0] == "OBX":
try:
observation_types.append(fields[13])
except IndexError:
observation_types.append(None)
observation_type_set = set(observation_types)
if len(observation_type_set)==1 and \
"APERIODIC" in observation_type_set:
observation_type = "aperiodic"
if len(observation_type_set)==1 and \
None in observation_type_set:
observation_type = "default"
return {
"assigned_patient_location": location,
"observation_date_time": date_time,
"observation_type": observation_type,
"message": message
}
def down_sample(self, parsed_records, start_time, end_time):
dt = [rec["observation_date_time"] for rec in parsed_records]
records = []
tmp = {}
sorted_records = sorted(parsed_records,
key=lambda rec: rec["observation_date_time"])
for rec in sorted_records:
date_time = rec["observation_date_time"]
location = rec["assigned_patient_location"]
observation_type = rec["observation_type"]
message = rec["message"]
if date_time<start_time or \
date_time>=end_time:
continue
tmp[location] = tmp.get(location, {})
tmp[location][observation_type] = rec
for d in tmp.values():
for rec in d.values():
records.append(rec)
return records
def log(self, msg):
self.producer.send( self.log_topic,
self.LOG_FORMAT.format( datetime.now().timestamp(),
msg
) \
.encode()
)
def get_end_time(self, current_time):
interval = self.INTERVAL
return current_time - current_time % interval + interval
async def send_or_ignore_message(self, addr, messages):
writer = self.connections[addr]
try:
for msg in messages:
writer.write(msg)
await writer.drain()
except ConnectionResetError:
"""
The error is not thrown reliably. If a connection is broken, and
one try to
writer.write(record)
await writer.drain()
This error may not manifest. It is thrown more often when one try
to repeatedly write to and drain a broken connection.
"""
self.remove_client(addr)
def refuse_client(self, addr, writer):
self.log("{} refused".format(addr))
writer.close()
def add_client(self, addr, writer):
if addr not in self.connections:
self.log("{} accepted".format(addr))
self.connections[addr] = writer
else:
self.remove_client(addr)
self.add_client(addr)
def remove_client(self, addr):
if addr in self.connections:
self.log("{} closed".format(addr))
writer = self.connections.pop(addr)
try:
writer.close()
except ConnectionResetError:
pass
def cleanup(self):
self.log("shutdown")
for addr in self.connections.keys():
self.remove_client(addr)
self.producer.flush()
self.producer.close()
class IBUSStreamingDownsamplingConsumer:
LOG_FORMAT ="{} UTC_TS\t"\
"{}"
def __init__(self, kafkaHost, kafkaPort, tcpHost, tcpPort, group_id, topic,
logTopic, interval):
self.kafkaHost = kafkaHost
self.kafkaPort = kafkaPort
self.tcpHost = tcpHost
self.tcpPort = tcpPort
self.group_id = group_id
self.topic = topic
self.logTopic = logTopic
self.interval = int(interval)
self.consumer = KafkaConsumer(
topic,
bootstrap_servers=["{}:{}".format(kafkaHost, kafkaPort)],
group_id=group_id,
enable_auto_commit=False)
self.producer = KafkaProducer(
bootstrap_servers=["{}:{}".format(kafkaHost, kafkaPort)])
self.tcpWriter = None
def getTopicPartitions(self):
self.consumer.topics() #This ensures local cache is updated with
# information about partitions, offsets etc.
pids = self.consumer.partitions_for_topic(self.topic)
tps = [TopicPartition(self.topic, pid) for pid in pids]
return tps
def getTopicPartitionsCommittedPositions(self):
tps = self.getTopicPartitions()
ret = [(tp, self.consumer.committed(tp)) for tp in tps]
return ret
async def tcp_server_handler(self, reader, writer):
addr = str(writer.get_extra_info("socket").getpeername())
if self.tcpWriter is not None:
self.log("refused " + addr)
writer.write(b"Connection limit reached; connection refused.")
writer.close()
return
self.log("accepted " + addr)
self.tcpWriter = writer
t1 = asyncio.create_task(self.poll_from_Kafka(writer))
try:
while True:
data = await reader.read(1) # 1024*16 bytes
if not data:
break
except BrokenPipeError:
"""
Catches connecton reset by peer when we are sending the batched data,
which is also when we cannot check for reader. The broken connection
on the writer side will ultimately lead to BrokenPipeError on the
reader side. Hence
"""
pass
finally:
t1.cancel()
self.log("closed " + addr)
writer.close()
self.tcpWriter = None
async def poll_from_Kafka(self, writer):
while True:
prevPos = self.getTopicPartitionsCommittedPositions()
polled = self.consumer.poll(timeout_ms=1000)
records = [
record.value for recordList in polled.values()
for record in recordList
]
try:
for record in records:
writer.write(record)
await writer.drain()
except ConnectionResetError:
"""
The error is not thrown reliably. If a connection is broken, and
one try to
writer.write(record)
await writer.drain()
This error may not manifest. It is thrown more often when one try
to repeatedly write to and drain a broken connection.
"""
print("Last batch not fully sent, not commited.")
for tp, pos in prevPos:
self.consumer.seek(tp, pos)
break
else:
self.consumer.commit()
await asyncio.sleep(self.interval)
def log(self, msg):
self.producer.send( self.logTopic,
self.LOG_FORMAT.format( datetime.now().timestamp(),
msg
) \
.encode()
)
def cleanup(self):
self.log("shutdown")
self.consumer.close()
self.producer.flush()
self.producer.close()
def run(self):
self.log("running")
asyncio.run(self._async_run())
async def _async_run(self):
tcpServer = await asyncio.start_server(self.tcp_server_handler,
self.tcpHost, self.tcpPort)
await tcpServer.serve_forever()
