aboutsummaryrefslogtreecommitdiffstats
path: root/python/fatcat_tools/workers/changelog.py
blob: 8b1ba5e930a1eb723dcf067e290b3601575f7744 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220

import json
import time
from confluent_kafka import Consumer, Producer, KafkaException

from .worker_common import FatcatWorker, most_recent_message


class ChangelogWorker(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, kafka_hosts, produce_topic, poll_interval=10.0, offset=None):
        super().__init__(kafka_hosts=kafka_hosts,
                         produce_topic=produce_topic,
                         api=api)
        self.poll_interval = poll_interval
        self.offset = offset    # the fatcat changelog offset, not the kafka offset

    def run(self):

        # 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 = most_recent_message(self.produce_topic, self.kafka_config)
            if msg:
                self.offset = json.loads(msg.decode('utf-8'))['index']
            else:
                self.offset = 0
            print("Most recent changelog index in Kafka seems to be {}".format(self.offset))

        def fail_fast(err, msg):
            if err is not None:
                print("Kafka producer delivery error: {}".format(err))
                print("Bailing out...")
                # TODO: should it be sys.exit(-1)?
                raise KafkaException(err)

        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)

        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(
                    self.produce_topic,
                    json.dumps(obj).encode('utf-8'),
                    key=str(i),
                    on_delivery=fail_fast,
                    #NOTE timestamp could be timestamp=cle.timestamp (?)
                )
                self.offset = i
            producer.flush()
            print("Sleeping {} seconds...".format(self.poll_interval))
            time.sleep(self.poll_interval)


class EntityUpdatesWorker(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, kafka_hosts, consume_topic, release_topic, file_topic, container_topic, poll_interval=5.0):
        super().__init__(kafka_hosts=kafka_hosts,
                         consume_topic=consume_topic,
                         api=api)
        self.release_topic = release_topic
        self.file_topic = file_topic
        self.container_topic = container_topic
        self.poll_interval = poll_interval
        self.consumer_group = "entity-updates"

    def run(self):

        def fail_fast(err, msg):
            if err is not None:
                print("Kafka producer delivery error: {}".format(err))
                print("Bailing out...")
                # TODO: should it be sys.exit(-1)?
                raise KafkaException(err)

        def on_commit(err, partitions):
            if err is not None:
                print("Kafka consumer commit error: {}".format(err))
                print("Bailing out...")
                # TODO: should it be sys.exit(-1)?
                raise KafkaException(err)
            for p in partitions:
                # check for partition-specific commit errors
                print(p)
                if p.error:
                    print("Kafka consumer commit error: {}".format(p.error))
                    print("Bailing out...")
                    # TODO: should it be sys.exit(-1)?
                    raise KafkaException(err)
            print("Kafka consumer commit successful")
            pass

        def on_rebalance(consumer, partitions):
            for p in partitions:
                if p.error:
                    raise KafkaException(p.error)
            print("Kafka partitions rebalanced: {} / {}".format(
                consumer, partitions))

        consumer_conf = self.kafka_config.copy()
        consumer_conf.update({
            'group.id': self.consumer_group,
            'on_commit': fail_fast,
            # messages don't have offset marked as stored until pushed to
            # elastic, but we do auto-commit stored offsets to broker
            'enable.auto.commit': True,
            'enable.auto.offset.store': False,
            # user code timeout; if no poll after this long, assume user code
            # hung and rebalance (default: 5min)
            'max.poll.interval.ms': 180000,
            'default.topic.config': {
                'auto.offset.reset': 'latest',
            },
        })
        consumer = Consumer(consumer_conf)

        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)

        consumer.subscribe([self.consume_topic],
            on_assign=on_rebalance,
            on_revoke=on_rebalance,
        )
        print("Kafka consuming {}".format(self.consume_topic))

        while True:
            msg = consumer.poll(self.poll_interval)
            if not msg:
                print("nothing new from kafka (poll_interval: {} sec)".format(self.poll_interval))
                continue
            if msg.error():
                raise KafkaException(msg.error())

            cle = json.loads(msg.value().decode('utf-8'))
            #print(cle)
            print("processing changelog index {}".format(cle['index']))
            release_ids = []
            file_ids = []
            container_ids = []
            work_ids = []
            release_edits = cle['editgroup']['edits']['releases']
            for re in release_edits:
                release_ids.append(re['ident'])
            file_edits = cle['editgroup']['edits']['files']
            for e in file_edits:
                file_ids.append(e['ident'])
            container_edits = cle['editgroup']['edits']['containers']
            for e in container_edits:
                container_ids.append(e['ident'])
            work_edits = cle['editgroup']['edits']['works']
            for e in work_edits:
                work_ids.append(e['ident'])

            # TODO: do these fetches in parallel using a thread pool?
            for ident in set(file_ids):
                file_entity = self.api.get_file(ident, expand=None)
                # update release when a file changes
                # TODO: fetch old revision as well, and only update
                # releases for which list changed
                release_ids.extend(file_entity.release_ids or [])
                file_dict = self.api.api_client.sanitize_for_serialization(file_entity)
                producer.produce(
                    self.file_topic,
                    json.dumps(file_dict).encode('utf-8'),
                    key=ident.encode('utf-8'),
                    on_delivery=fail_fast,
                )
            for ident in set(container_ids):
                container = self.api.get_container(ident)
                container_dict = self.api.api_client.sanitize_for_serialization(container)
                producer.produce(
                    self.container_topic,
                    json.dumps(container_dict).encode('utf-8'),
                    key=ident.encode('utf-8'),
                    on_delivery=fail_fast,
                )
            for ident in set(release_ids):
                release = self.api.get_release(ident, expand="files,filesets,webcaptures,container")
                work_ids.append(release.work_id)
                release_dict = self.api.api_client.sanitize_for_serialization(release)
                producer.produce(
                    self.release_topic,
                    json.dumps(release_dict).encode('utf-8'),
                    key=ident.encode('utf-8'),
                    on_delivery=fail_fast,
                )
            producer.flush()
            # TODO: actually update works
            consumer.store_offsets(message=msg)