diff options
Diffstat (limited to 'python/fatcat_tools/harvest')
| -rw-r--r-- | python/fatcat_tools/harvest/doi_registrars.py | 13 | ||||
| -rw-r--r-- | python/fatcat_tools/harvest/harvest_common.py | 37 | ||||
| -rw-r--r-- | python/fatcat_tools/harvest/oaipmh.py | 13 |
3 files changed, 49 insertions, 14 deletions
diff --git a/python/fatcat_tools/harvest/doi_registrars.py b/python/fatcat_tools/harvest/doi_registrars.py index 3362df35..7e791745 100644 --- a/python/fatcat_tools/harvest/doi_registrars.py +++ b/python/fatcat_tools/harvest/doi_registrars.py @@ -56,11 +56,7 @@ class HarvestCrossrefWorker: self.is_update_filter = is_update_filter self.kafka_config = { 'bootstrap.servers': kafka_hosts, - 'delivery.report.only.error': True, 'message.max.bytes': 20000000, # ~20 MBytes; broker is ~50 MBytes - 'default.topic.config': { - 'request.required.acks': 'all', - }, } self.state = HarvestState(start_date, end_date) @@ -97,7 +93,14 @@ class HarvestCrossrefWorker: # TODO: should it be sys.exit(-1)? raise KafkaException(err) - producer = Producer(self.kafka_config) + producer_conf = self.kafka_config.copy() + producer_conf.update({ + 'delivery.report.only.error': True, + 'default.topic.config': { + 'request.required.acks': -1, # all brokers must confirm + }, + }) + producer = Producer(producer_conf) date_str = date.isoformat() params = self.params(date_str) diff --git a/python/fatcat_tools/harvest/harvest_common.py b/python/fatcat_tools/harvest/harvest_common.py index aa7a69f5..78830a1c 100644 --- a/python/fatcat_tools/harvest/harvest_common.py +++ b/python/fatcat_tools/harvest/harvest_common.py @@ -130,30 +130,56 @@ class HarvestState: if err: raise KafkaException(err) print("Commiting status to Kafka: {}".format(kafka_topic)) - producer = Producer(kafka_config) - producer.produce(kafka_topic, state_json, on_delivery=fail_fast) + producer_conf = kafka_config.copy() + producer_conf.update({ + 'delivery.report.only.error': True, + 'default.topic.config': { + 'request.required.acks': -1, # all brokers must confirm + }, + }) + producer = Producer(producer_conf) + producer.produce( + kafka_topic, + state_json, + on_delivery=fail_fast) producer.flush() return state_json def initialize_from_kafka(self, kafka_topic, kafka_config): """ kafka_topic should have type str + + TODO: this method does not fail if client can't connect to host. """ if not kafka_topic: return print("Fetching state from kafka topic: {}".format(kafka_topic)) + def fail_fast(err, msg): + if err: + raise KafkaException(err) conf = kafka_config.copy() conf.update({ + 'group.id': 'dummy_init_group', # should never be commited + 'enable.auto.commit': False, 'auto.offset.reset': 'earliest', 'session.timeout.ms': 10000, - 'group.id': kafka_topic + "-init", }) consumer = Consumer(conf) + + # this watermark fetch is mostly to ensure we are connected to broker and + # fail fast if not, but we also confirm that we read to end below. + hwm = consumer.get_watermark_offsets( + TopicPartition(kafka_topic, 0), + timeout=5.0, + cached=False) + if not hwm: + raise Exception("Kafka consumer timeout, or topic {} doesn't exist".format(kafka_topic)) + consumer.assign([TopicPartition(kafka_topic, 0, 0)]) c = 0 while True: - msg = consumer.poll(timeout=1.0) + msg = consumer.poll(timeout=2.0) if not msg: break if msg.error(): @@ -162,4 +188,7 @@ class HarvestState: self.update(msg.value().decode('utf-8')) c += 1 consumer.close() + + # verify that we got at least to HWM + assert c >= hwm[1] print("... got {} state update messages, done".format(c)) diff --git a/python/fatcat_tools/harvest/oaipmh.py b/python/fatcat_tools/harvest/oaipmh.py index 3e3bea03..f908ba83 100644 --- a/python/fatcat_tools/harvest/oaipmh.py +++ b/python/fatcat_tools/harvest/oaipmh.py @@ -39,10 +39,7 @@ class HarvestOaiPmhWorker: self.state_topic = state_topic self.kafka_config = { 'bootstrap.servers': kafka_hosts, - 'delivery.report.only.error': True, 'message.max.bytes': 20000000, # ~20 MBytes; broker is ~50 MBytes - 'default.topic.config': - {'request.required.acks': 'all'}, } self.loop_sleep = 60*60 # how long to wait, in seconds, between date checks @@ -62,7 +59,14 @@ class HarvestOaiPmhWorker: # TODO: should it be sys.exit(-1)? raise KafkaException(err) - producer = Producer(self.kafka_config) + producer_conf = self.kafka_config.copy() + producer_conf.update({ + 'delivery.report.only.error': True, + 'default.topic.config': { + 'request.required.acks': -1, # all brokers must confirm + }, + }) + producer = Producer(producer_conf) api = sickle.Sickle(self.endpoint_url) date_str = date.isoformat() @@ -88,7 +92,6 @@ class HarvestOaiPmhWorker: item.raw.encode('utf-8'), key=item.header.identifier.encode('utf-8'), on_delivery=fail_fast) - producer.poll(0) producer.flush() def run(self, continuous=False): |