first-draft kafka workers (changelog, release_update)

3 files changed, 197 insertions, 0 deletions

diff --git a/python/fatcat/changelog_workers.py b/python/fatcat/changelog_workers.py
new file mode 100644
index 00000000..5f8621cf
--- /dev/null
+++ b/python/fatcat/changelog_workers.py
@@ -0,0 +1,120 @@
+
+import json
+import time
+from itertools import islice
+from fatcat.worker_common import FatcatWorker
+from pykafka.common import OffsetType
+
+
+class FatcatChangelogWorker(FatcatWorker):
+    """
+    Periodically polls the fatcat API looking for new changelogs. When they are
+    found, fetch them and push (as JSON) into a Kafka topic.
+    """
+
+    def __init__(self, api_host_url, kafka_hosts, produce_topic, poll_interval=10.0, offset=None):
+        # TODO: should be offset=0
+        super().__init__(kafka_hosts=kafka_hosts,
+                         produce_topic=produce_topic,
+                         api_host_url=api_host_url)
+        self.poll_interval = poll_interval
+        self.offset = offset    # the fatcat changelog offset, not the kafka offset
+
+    def most_recent_message(self, topic):
+        """
+        Tries to fetch the most recent message from a given topic.
+        This only makes sense for single partition topics, though could be
+        extended with "last N" behavior.
+
+        Following "Consuming the last N messages from a topic"
+        from https://pykafka.readthedocs.io/en/latest/usage.html#consumer-patterns
+        """
+        consumer = topic.get_simple_consumer(
+            auto_offset_reset=OffsetType.LATEST,
+            reset_offset_on_start=True)
+        offsets = [(p, op.last_offset_consumed - 1)
+                    for p, op in consumer._partitions.items()]
+        offsets = [(p, (o if o > -1 else -2)) for p, o in offsets]
+        if -2 in [o for p, o in offsets]:
+            return None
+        else:
+            consumer.reset_offsets(offsets)
+            msg = islice(consumer, 1)
+            if msg:
+                return list(msg)[0].value
+            else:
+                return None
+
+    def run(self):
+        topic = self.kafka.topics[self.produce_topic]
+        # On start, try to consume the most recent from the topic, and using
+        # that as the starting offset. Note that this is a single-partition
+        # topic
+        if self.offset is None:
+            print("Checking for most recent changelog offset...")
+            msg = self.most_recent_message(topic)
+            if msg:
+                self.offset = json.loads(msg.decode('utf-8'))['index']
+            else:
+                self.offset = 1
+
+        with topic.get_sync_producer() as producer:
+            while True: 
+                latest = int(self.api.get_changelog(limit=1)[0].index)
+                if latest > self.offset:
+                    print("Fetching changelogs from {} through {}".format(
+                        self.offset+1, latest))
+                for i in range(self.offset+1, latest+1):
+                    cle = self.api.get_changelog_entry(i)
+                    obj = self.api.api_client.sanitize_for_serialization(cle)
+                    producer.produce(
+                        message=json.dumps(obj).encode('utf-8'),
+                        partition_key=None,
+                        timestamp=None,
+                        #XXX: timestamp=cle.timestamp,
+                    )
+                    self.offset = i
+                print("Sleeping {} seconds...".format(self.poll_interval))
+                time.sleep(self.poll_interval)
+
+
+class FatcatEntityUpdatesWorker(FatcatWorker):
+    """
+    Consumes from the changelog topic and publishes expanded entities (fetched
+    from API) to update topics.
+
+    For now, only release updates are published.
+    """
+
+    def __init__(self, api_host_url, kafka_hosts, consume_topic, release_topic):
+        super().__init__(kafka_hosts=kafka_hosts,
+                         consume_topic=consume_topic,
+                         api_host_url=api_host_url)
+        self.release_topic = release_topic
+        self.consumer_group = "entity-updates"
+
+    def run(self):
+        changelog_topic = self.kafka.topics[self.consume_topic]
+        release_topic = self.kafka.topics[self.release_topic]
+
+        consumer = changelog_topic.get_simple_consumer(
+            consumer_group=self.consumer_group,
+            auto_offset_reset=OffsetType.LATEST,
+            reset_offset_on_start=False,
+        )
+
+        with release_topic.get_sync_producer() as producer:
+            for msg in consumer:
+                cle = json.loads(msg.value.decode('utf-8'))
+                #print(cle)
+                release_edits = cle['editgroup']['edits']['releases']
+                for re in release_edits:
+                    ident = re['ident']
+                    release = self.api.get_release(ident, expand="files,container")
+                    release_dict = self.api.api_client.sanitize_for_serialization(release)
+                    producer.produce(
+                        message=json.dumps(release_dict).encode('utf-8'),
+                        partition_key=ident.encode('utf-8'),
+                        timestamp=None,
+                    )
+
diff --git a/python/fatcat/worker_common.py b/python/fatcat/worker_common.py
new file mode 100644
index 00000000..77ea2c15
--- /dev/null
+++ b/python/fatcat/worker_common.py
@@ -0,0 +1,25 @@
+
+import re
+import sys
+import csv
+import json
+import itertools
+import fatcat_client
+from pykafka import KafkaClient
+from fatcat_client.rest import ApiException
+
+
+class FatcatWorker:
+    """
+    Common code for for Kafka producers and consumers.
+    """
+
+    def __init__(self, kafka_hosts, produce_topic=None, consume_topic=None, api_host_url=None):
+        if api_host_url:
+            conf = fatcat_client.Configuration()
+            conf.host = api_host_url
+            self.api = fatcat_client.DefaultApi(fatcat_client.ApiClient(conf))
+        self.kafka = KafkaClient(hosts=kafka_hosts, broker_version="1.0.0")
+        self.produce_topic = produce_topic
+        self.consume_topic = consume_topic
+
diff --git a/python/fatcat_worker.py b/python/fatcat_worker.py
new file mode 100755
index 00000000..cc11beca
--- /dev/null
+++ b/python/fatcat_worker.py
@@ -0,0 +1,52 @@
+#!/usr/bin/env python3
+
+import sys
+import argparse
+from fatcat.changelog_workers import FatcatChangelogWorker, FatcatEntityUpdatesWorker
+
+def run_changelog_worker(args):
+    topic = "fatcat-{}.changelog".format(args.env)
+    worker = FatcatChangelogWorker(args.api_host_url, args.kafka_hosts, topic,
+        args.poll_interval)
+    worker.run()
+
+def run_entity_updates_worker(args):
+    changelog_topic = "fatcat-{}.changelog".format(args.env)
+    release_topic = "fatcat-{}.release-updates".format(args.env)
+    worker = FatcatEntityUpdatesWorker(args.api_host_url, args.kafka_hosts,
+        changelog_topic, release_topic)
+    worker.run()
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--debug',
+        action='store_true',
+        help="enable debug logging")
+    parser.add_argument('--api-host-url',
+        default="http://localhost:9411/v0",
+        help="fatcat API host/port to use")
+    parser.add_argument('--kafka-hosts',
+        default="localhost:9092",
+        help="list of Kafka brokers (host/port) to use")
+    parser.add_argument('--env',
+        default="qa",
+        help="Kafka topic namespace to use (eg, prod, qa)")
+    subparsers = parser.add_subparsers()
+
+    sub_changelog = subparsers.add_parser('changelog')
+    sub_changelog.set_defaults(func=run_changelog_worker)
+    sub_changelog.add_argument('--poll-interval',
+        help="how long to wait between polling (seconds)",
+        default=10.0, type=float)
+
+    sub_entity_updates = subparsers.add_parser('entity-updates')
+    sub_entity_updates.set_defaults(func=run_entity_updates_worker)
+
+    args = parser.parse_args()
+    if not args.__dict__.get("func"):
+        print("tell me what to do!")
+        sys.exit(-1)
+    args.func(args)
+
+if __name__ == '__main__':
+    main()