diff --git a/swh/journal/publisher.py b/swh/journal/publisher.py index ff0e0f4..9774c44 100644 --- a/swh/journal/publisher.py +++ b/swh/journal/publisher.py @@ -1,194 +1,198 @@ # Copyright (C) 2016-2019 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from collections import defaultdict import logging from kafka import KafkaProducer, KafkaConsumer from swh.storage import get_storage from swh.storage.algos import snapshot from .serializers import kafka_to_key, key_to_kafka +logger = logging.getLogger(__name__) + class JournalPublisher: """The journal publisher is a layer in charge of: - consuming messages from topics (1 topic per object_type) - reify the object ids read from those topics (using the storage) - producing those reified objects to output topics (1 topic per object type) The main entry point for this class is the 'poll' method. """ def __init__(self, config): self.config = config self._prepare_storage(config) self._prepare_journal(config) self.max_messages = self.config['max_messages'] def _prepare_journal(self, config): """Prepare the consumer and subscriber instances for the publisher to actually be able to discuss with the journal. """ # yes, the temporary topics contain values that are actually _keys_ self.consumer = KafkaConsumer( bootstrap_servers=config['brokers'], value_deserializer=kafka_to_key, auto_offset_reset='earliest', enable_auto_commit=False, group_id=config['consumer_id'], ) self.producer = KafkaProducer( bootstrap_servers=config['brokers'], key_serializer=key_to_kafka, value_serializer=key_to_kafka, client_id=config['publisher_id'], ) - logging.debug('Subscribing to object types event: %s' % ( + logger.info('Subscribing to object types event: %s' % ( config['object_types'], )) self.consumer.subscribe( topics=['%s.%s' % (config['temporary_prefix'], object_type) for object_type in config['object_types']], ) def _prepare_storage(self, config): """Prepare the storage instance needed for the publisher to be able to discuss with the storage to retrieve the objects. """ self.storage = get_storage(**config['storage']) def poll(self, max_messages=None): """Process a batch of messages from the consumer's topics. Use the storage to reify those ids. Produces back those reified objects to the production topics. This method polls a given amount of message then stops. The number of messages to consume is either provided or configured as fallback. The following method is expected to be called from within a loop. """ messages = defaultdict(list) if max_messages is None: max_messages = self.max_messages for num, message in enumerate(self.consumer): object_type = message.topic.split('.')[-1] - logging.debug('num: %s, object_type: %s, message: %s' % ( + logger.debug('num: %s, object_type: %s, message: %s' % ( num, object_type, message)) messages[object_type].append(message.value) if num + 1 >= self.max_messages: break + logger.debug('number of messages: %s', num) + new_objects = self.process_objects(messages) self.produce_messages(new_objects) self.consumer.commit() def process_objects(self, messages): """Given a dict of messages {object type: [object id]}, reify those ids to swh object from the storage and returns a corresponding dict. Args: messages (dict): Dict of {object_type: [id-as-bytes]} Returns: Dict of {object_type: [tuple]}. object_type (str): content, revision, release tuple (bytes, dict): object id as bytes, object as swh dict. """ processors = { 'content': self.process_contents, 'revision': self.process_revisions, 'release': self.process_releases, 'snapshot': self.process_snapshots, 'origin': self.process_origins, 'origin_visit': self.process_origin_visits, } return { key: processors[key](value) for key, value in messages.items() } def produce_messages(self, messages): """Produce new swh object to the producer topic. Args: messages ([dict]): Dict of {object_type: [tuple]}. object_type (str): content, revision, release tuple (bytes, dict): object id as bytes, object as swh dict. """ for object_type, objects in messages.items(): topic = '%s.%s' % (self.config['final_prefix'], object_type) for key, object in objects: - logging.debug('topic: %s, key: %s, value: %s' % ( + logger.debug('topic: %s, key: %s, value: %s' % ( topic, key, object)) self.producer.send(topic, key=key, value=object) self.producer.flush() def process_contents(self, content_objs): - logging.debug('contents: %s' % content_objs) + logger.debug('contents: %s' % content_objs) metadata = self.storage.content_get_metadata( (c[b'sha1'] for c in content_objs)) return [(content['sha1'], content) for content in metadata] def process_revisions(self, revision_objs): - logging.debug('revisions: %s' % revision_objs) + logger.debug('revisions: %s' % revision_objs) metadata = self.storage.revision_get((r[b'id'] for r in revision_objs)) return [(revision['id'], revision) for revision in metadata if revision] def process_releases(self, release_objs): - logging.debug('releases: %s' % release_objs) + logger.debug('releases: %s' % release_objs) metadata = self.storage.release_get((r[b'id'] for r in release_objs)) return [(release['id'], release) for release in metadata] def process_origins(self, origin_objs): - logging.debug('origins: %s' % origin_objs) + logger.debug('origins: %s' % origin_objs) r = [] for o in origin_objs: origin = {'url': o[b'url'], 'type': o[b'type']} r.append((origin, origin)) return r def process_origin_visits(self, origin_visits): - logging.debug('origin_visits: %s' % origin_visits) + logger.debug('origin_visits: %s' % origin_visits) metadata = [] for ov in origin_visits: origin_visit = self.storage.origin_visit_get_by( ov[b'origin'], ov[b'visit']) if origin_visit: pk = ov[b'origin'], ov[b'visit'] origin_visit['date'] = str(origin_visit['date']) metadata.append((pk, origin_visit)) return metadata def process_snapshots(self, snapshot_objs): - logging.debug('snapshots: %s' % snapshot_objs) + logger.debug('snapshots: %s' % snapshot_objs) metadata = [] for snap in snapshot_objs: full_obj = snapshot.snapshot_get_all_branches( self.storage, snap[b'id']) metadata.append((full_obj['id'], full_obj)) return metadata if __name__ == '__main__': print('Please use the "swh-journal publisher run" command')