Esempio n. 1
0
 def __init__(self,
              kafka_brokers,
              security_protocol="",
              sasl_mechanism="",
              sasl_plain_username="",
              sasl_plain_password="",
              ssl_context="",
              timeout=5000):
     self.kafka_brokers = kafka_brokers
     self.security_protocol = security_protocol
     self.sasl_mechanism = sasl_mechanism
     self.sasl_plain_username = sasl_plain_username
     self.sasl_plain_password = sasl_plain_password
     self.ssl_context = ssl_context
     self.timeout = timeout
     if security_protocol:
         self.client = KafkaClient(bootstrap_servers=kafka_brokers,
                                   security_protocol=security_protocol,
                                   sasl_mechanism=sasl_mechanism,
                                   sasl_plain_username=sasl_plain_username,
                                   sasl_plain_password=sasl_plain_password,
                                   ssl_context=ssl_context,
                                   timeout=timeout)
     else:
         self.client = KafkaClient(bootstrap_servers=kafka_brokers,
                                   timeout=timeout)
     self.lag_topics_found = []
     self.lag_total = 0
Esempio n. 2
0
 def __init__(self, **configs):
     self.config = copy.copy(self.DEFAULT_CONFIG)
     self.config.update(configs)
     self._client = KafkaClient(**self.config)
     self._coordinator_id = None
     self.group_id = configs['group_id']
     self.topic = configs['topic']
Esempio n. 3
0
def test_poll(mocker):
    mocker.patch.object(KafkaClient, '_bootstrap')
    metadata = mocker.patch.object(KafkaClient, '_maybe_refresh_metadata')
    _poll = mocker.patch.object(KafkaClient, '_poll')
    cli = KafkaClient()
    tasks = mocker.patch.object(cli._delayed_tasks, 'next_at')

    # metadata timeout wins
    metadata.return_value = 1000
    tasks.return_value = 2
    cli.poll()
    _poll.assert_called_with(1.0, sleep=True)

    # user timeout wins
    cli.poll(250)
    _poll.assert_called_with(0.25, sleep=True)

    # tasks timeout wins
    tasks.return_value = 0
    cli.poll(250)
    _poll.assert_called_with(0, sleep=True)

    # default is request_timeout_ms
    metadata.return_value = 1000000
    tasks.return_value = 10000
    cli.poll()
    _poll.assert_called_with(cli.config['request_timeout_ms'] / 1000.0,
                             sleep=True)
Esempio n. 4
0
    def __init__(self, **configs):
        log.debug("Starting KafkaAdminClient with configuration: %s", configs)
        extra_configs = set(configs).difference(self.DEFAULT_CONFIG)
        if extra_configs:
            raise KafkaConfigurationError(
                "Unrecognized configs: {}".format(extra_configs))

        self.config = copy.copy(self.DEFAULT_CONFIG)
        self.config.update(configs)

        # Configure metrics
        metrics_tags = {'client-id': self.config['client_id']}
        metric_config = MetricConfig(
            samples=self.config['metrics_num_samples'],
            time_window_ms=self.config['metrics_sample_window_ms'],
            tags=metrics_tags)
        reporters = [
            reporter() for reporter in self.config['metric_reporters']
        ]
        self._metrics = Metrics(metric_config, reporters)

        self._client = KafkaClient(metrics=self._metrics,
                                   metric_group_prefix='admin',
                                   **self.config)

        # Get auto-discovered version from client if necessary
        if self.config['api_version'] is None:
            self.config['api_version'] = self._client.config['api_version']

        self._closed = False
        self._refresh_controller_id()
        log.debug("KafkaAdminClient started.")
Esempio n. 5
0
 def _create_kafka_client(self):
     kafka_conn_str = self.instance.get('kafka_connect_str')
     if not isinstance(kafka_conn_str, (string_types, list)):
         raise ConfigurationError('kafka_connect_str should be string or list of strings')
     kafka_version = self.instance.get('kafka_client_api_version')
     if isinstance(kafka_version, str):
         kafka_version = tuple(map(int, kafka_version.split(".")))
     kafka_client = KafkaClient(
         bootstrap_servers=kafka_conn_str,
         client_id='dd-agent',
         request_timeout_ms=self.init_config.get('kafka_timeout', DEFAULT_KAFKA_TIMEOUT) * 1000,
         # if `kafka_client_api_version` is not set, then kafka-python automatically probes the cluster for broker
         # version during the bootstrapping process. Note that probing randomly picks a broker to probe, so in a
         # mixed-version cluster probing returns a non-deterministic result.
         api_version=kafka_version,
         # While we check for SSL params, if not present they will default to the kafka-python values for plaintext
         # connections
         security_protocol=self.instance.get('security_protocol', 'PLAINTEXT'),
         sasl_mechanism=self.instance.get('sasl_mechanism'),
         sasl_plain_username=self.instance.get('sasl_plain_username'),
         sasl_plain_password=self.instance.get('sasl_plain_password'),
         sasl_kerberos_service_name=self.instance.get('sasl_kerberos_service_name', 'kafka'),
         sasl_kerberos_domain_name=self.instance.get('sasl_kerberos_domain_name'),
         ssl_cafile=self.instance.get('ssl_cafile'),
         ssl_check_hostname=self.instance.get('ssl_check_hostname', True),
         ssl_certfile=self.instance.get('ssl_certfile'),
         ssl_keyfile=self.instance.get('ssl_keyfile'),
         ssl_crlfile=self.instance.get('ssl_crlfile'),
         ssl_password=self.instance.get('ssl_password'),
     )
     # Force initial population of the local cluster metadata cache
     kafka_client.poll(future=kafka_client.cluster.request_update())
     if kafka_client.cluster.topics(exclude_internal_topics=False) is None:
         raise RuntimeError("Local cluster metadata cache did not populate.")
     return kafka_client
Esempio n. 6
0
def test_send(conn):
    cli = KafkaClient()

    # Send to unknown node => raises AssertionError
    try:
        cli.send(2, None)
        assert False, 'Exception not raised'
    except AssertionError:
        pass

    # Send to disconnected node => NodeNotReady
    conn.state = ConnectionStates.DISCONNECTED
    f = cli.send(0, None)
    assert f.failed()
    assert isinstance(f.exception, Errors.NodeNotReadyError)

    conn.state = ConnectionStates.CONNECTED
    cli._maybe_connect(0)
    # ProduceRequest w/ 0 required_acks -> no response
    request = ProduceRequest[0](0, 0, [])
    ret = cli.send(0, request)
    assert conn.send.called_with(request, expect_response=False)
    assert isinstance(ret, Future)

    request = MetadataRequest[0]([])
    cli.send(0, request)
    assert conn.send.called_with(request, expect_response=True)
def cli(mocker, conn):
    mocker.patch('kafka.cluster.dns_lookup',
                 return_value=[(socket.AF_INET, None, None, None, ('localhost',
                                                                   9092))])
    client = KafkaClient(api_version=(0, 9))
    client.poll(future=client.cluster.request_update())
    return client
def test_poll(mocker):
    metadata = mocker.patch.object(KafkaClient, '_maybe_refresh_metadata')
    _poll = mocker.patch.object(KafkaClient, '_poll')
    ifrs = mocker.patch.object(KafkaClient, 'in_flight_request_count')
    ifrs.return_value = 1
    cli = KafkaClient(api_version=(0, 9))

    # metadata timeout wins
    metadata.return_value = 1000
    cli.poll()
    _poll.assert_called_with(1.0)

    # user timeout wins
    cli.poll(250)
    _poll.assert_called_with(0.25)

    # default is request_timeout_ms
    metadata.return_value = 1000000
    cli.poll()
    _poll.assert_called_with(cli.config['request_timeout_ms'] / 1000.0)

    # If no in-flight-requests, drop timeout to retry_backoff_ms
    ifrs.return_value = 0
    cli.poll()
    _poll.assert_called_with(cli.config['retry_backoff_ms'] / 1000.0)
 def __init__(self, **configs):
     self.zk_client = None
     self.zk_configuration = None
     self.zookeeper_sleep_time = 5
     self.zookeeper_max_retries = 5
     self.kafka_sleep_time = 5
     self.kafka_max_retries = 5
     self.client = KafkaClient(**configs)
     self.refresh()
Esempio n. 10
0
def test_bootstrap_servers(mocker, bootstrap, expected_hosts):
    mocker.patch.object(KafkaClient, '_bootstrap')
    if bootstrap is None:
        KafkaClient()
    else:
        KafkaClient(bootstrap_servers=bootstrap)

    # host order is randomized internally, so resort before testing
    (hosts, ), _ = KafkaClient._bootstrap.call_args  # pylint: disable=no-member
    assert sorted(hosts) == sorted(expected_hosts)
Esempio n. 11
0
def test_poll(mocker):
    mocker.patch.object(KafkaClient, '_bootstrap')
    metadata = mocker.patch.object(KafkaClient, '_maybe_refresh_metadata')
    _poll = mocker.patch.object(KafkaClient, '_poll')
    cli = KafkaClient(api_version=(0, 9))
    tasks = mocker.patch.object(cli._delayed_tasks, 'next_at')

    # metadata timeout wins
    metadata.return_value = 1000
    tasks.return_value = 2
    cli.poll()
    _poll.assert_called_with(1.0, sleep=True)

    # user timeout wins
    cli.poll(250)
    _poll.assert_called_with(0.25, sleep=True)

    # tasks timeout wins
    tasks.return_value = 0
    cli.poll(250)
    _poll.assert_called_with(0, sleep=True)

    # default is request_timeout_ms
    metadata.return_value = 1000000
    tasks.return_value = 10000
    cli.poll()
    _poll.assert_called_with(cli.config['request_timeout_ms'] / 1000.0,
                             sleep=True)
Esempio n. 12
0
def test_send(conn):
    cli = KafkaClient()

    # Send to unknown node => raises AssertionError
    try:
        cli.send(2, None)
        assert False, 'Exception not raised'
    except AssertionError:
        pass

    # Send to disconnected node => NodeNotReady
    conn.state = ConnectionStates.DISCONNECTED
    f = cli.send(0, None)
    assert f.failed()
    assert isinstance(f.exception, Errors.NodeNotReadyError)

    conn.state = ConnectionStates.CONNECTED
    cli._maybe_connect(0)
    # ProduceRequest w/ 0 required_acks -> no response
    request = ProduceRequest[0](0, 0, [])
    ret = cli.send(0, request)
    assert conn.send.called_with(request, expect_response=False)
    assert isinstance(ret, Future)

    request = MetadataRequest[0]([])
    cli.send(0, request)
    assert conn.send.called_with(request, expect_response=True)
Esempio n. 13
0
def test_finish_connect(conn):
    cli = KafkaClient()
    try:
        # Node not in metadata, raises AssertionError
        cli._initiate_connect(2)
    except AssertionError:
        pass
    else:
        assert False, 'Exception not raised'

    assert 0 not in cli._conns
    cli._initiate_connect(0)

    conn.connect.return_value = ConnectionStates.CONNECTING
    state = cli._finish_connect(0)
    assert 0 in cli._connecting
    assert state is ConnectionStates.CONNECTING

    conn.connect.return_value = ConnectionStates.CONNECTED
    state = cli._finish_connect(0)
    assert 0 not in cli._connecting
    assert state is ConnectionStates.CONNECTED

    # Failure to connect should trigger metadata update
    assert not cli.cluster._need_update
    cli._connecting.add(0)
    conn.connect.return_value = ConnectionStates.DISCONNECTED
    state = cli._finish_connect(0)
    assert 0 not in cli._connecting
    assert state is ConnectionStates.DISCONNECTED
    assert cli.cluster._need_update
Esempio n. 14
0
    def __init__(self, **configs):
        # Only check for extra config keys in top-level class
        extra_configs = set(configs).difference(self.DEFAULT_CONFIG)
        if extra_configs:
            raise KafkaConfigurationError("Unrecognized configs: %s" %
                                          extra_configs)

        self.config = copy.copy(self.DEFAULT_CONFIG)
        self.config.update(configs)

        self._client = KafkaClient(**self.config)
Esempio n. 15
0
    def __init__(self, *topics, **configs):
        self.config = copy.copy(self.DEFAULT_CONFIG)
        for key in self.config:
            if key in configs:
                self.config[key] = configs.pop(key)

        # Only check for extra config keys in top-level class
        assert not configs, 'Unrecognized configs: %s' % configs

        deprecated = {'smallest': 'earliest', 'largest': 'latest'}
        if self.config['auto_offset_reset'] in deprecated:
            new_config = deprecated[self.config['auto_offset_reset']]
            log.warning('use auto_offset_reset=%s (%s is deprecated)',
                        new_config, self.config['auto_offset_reset'])
            self.config['auto_offset_reset'] = new_config

        metrics_tags = {'client-id': self.config['client_id']}
        metric_config = MetricConfig(samples=self.config['metrics_num_samples'],
                                     time_window_ms=self.config['metrics_sample_window_ms'],
                                     tags=metrics_tags)
        reporters = [reporter() for reporter in self.config['metric_reporters']]
        self._metrics = Metrics(metric_config, reporters)
        # TODO _metrics likely needs to be passed to KafkaClient, etc.

        # api_version was previously a str. Accept old format for now
        if isinstance(self.config['api_version'], str):
            str_version = self.config['api_version']
            if str_version == 'auto':
                self.config['api_version'] = None
            else:
                self.config['api_version'] = tuple(map(int, str_version.split('.')))
            log.warning('use api_version=%s [tuple] -- "%s" as str is deprecated',
                        str(self.config['api_version']), str_version)

        self._client = KafkaClient(metrics=self._metrics, **self.config)

        # Get auto-discovered version from client if necessary
        if self.config['api_version'] is None:
            self.config['api_version'] = self._client.config['api_version']

        self._subscription = SubscriptionState(self.config['auto_offset_reset'])
        self._fetcher = Fetcher(
            self._client, self._subscription, self._metrics, **self.config)
        self._coordinator = ConsumerCoordinator(
            self._client, self._subscription, self._metrics,
            assignors=self.config['partition_assignment_strategy'],
            **self.config)
        self._closed = False
        self._iterator = None
        self._consumer_timeout = float('inf')

        if topics:
            self._subscription.subscribe(topics=topics)
            self._client.set_topics(topics)
Esempio n. 16
0
    def __init__(self, *topics, **configs):
        self.config = copy.copy(self.DEFAULT_CONFIG)
        for key in self.config:
            if key in configs:
                self.config[key] = configs.pop(key)

        # Only check for extra config keys in top-level class
        assert not configs, 'Unrecognized configs: %s' % configs

        deprecated = {'smallest': 'earliest', 'largest': 'latest' }
        if self.config['auto_offset_reset'] in deprecated:
            new_config = deprecated[self.config['auto_offset_reset']]
            log.warning('use auto_offset_reset=%s (%s is deprecated)',
                        new_config, self.config['auto_offset_reset'])
            self.config['auto_offset_reset'] = new_config

        metrics_tags = {'client-id': self.config['client_id']}
        metric_config = MetricConfig(samples=self.config['metrics_num_samples'],
                                     time_window_ms=self.config['metrics_sample_window_ms'],
                                     tags=metrics_tags)
        reporters = [reporter() for reporter in self.config['metric_reporters']]
        reporters.append(DictReporter('kafka.consumer'))
        self._metrics = Metrics(metric_config, reporters)
        metric_group_prefix = 'consumer'
        # TODO _metrics likely needs to be passed to KafkaClient, etc.

        self._client = KafkaClient(**self.config)

        # Check Broker Version if not set explicitly
        if self.config['api_version'] == 'auto':
            self.config['api_version'] = self._client.check_version()
        assert self.config['api_version'] in ('0.9', '0.8.2', '0.8.1', '0.8.0'), 'Unrecognized api version'

        # Convert api_version config to tuple for easy comparisons
        self.config['api_version'] = tuple(
            map(int, self.config['api_version'].split('.')))

        self._subscription = SubscriptionState(self.config['auto_offset_reset'])
        self._fetcher = Fetcher(
            self._client, self._subscription, self._metrics, metric_group_prefix, **self.config)
        self._coordinator = ConsumerCoordinator(
            self._client, self._subscription, self._metrics, metric_group_prefix,
            assignors=self.config['partition_assignment_strategy'],
            **self.config)
        self._closed = False
        self._iterator = None
        self._consumer_timeout = float('inf')

        if topics:
            self._subscription.subscribe(topics=topics)
            self._client.set_topics(topics)
Esempio n. 17
0
def test_maybe_refresh_metadata_ttl(mocker):
    mocker.patch.object(KafkaClient, '_bootstrap')
    _poll = mocker.patch.object(KafkaClient, '_poll')

    cli = KafkaClient(request_timeout_ms=9999999, retry_backoff_ms=2222)

    tasks = mocker.patch.object(cli._delayed_tasks, 'next_at')
    tasks.return_value = 9999999

    ttl = mocker.patch.object(cli.cluster, 'ttl')
    ttl.return_value = 1234

    cli.poll(timeout_ms=9999999, sleep=True)
    _poll.assert_called_with(1.234, sleep=True)
Esempio n. 18
0
def test_initiate_connect(conn):
    cli = KafkaClient()
    try:
        # Node not in metadata, raises AssertionError
        cli._initiate_connect(2)
    except AssertionError:
        pass
    else:
        assert False, 'Exception not raised'

    assert 0 not in cli._conns
    state = cli._initiate_connect(0)
    assert cli._conns[0] is conn
    assert state is conn.state
Esempio n. 19
0
def test_maybe_refresh_metadata_ttl(mocker):
    mocker.patch.object(KafkaClient, '_bootstrap')
    _poll = mocker.patch.object(KafkaClient, '_poll')

    cli = KafkaClient(request_timeout_ms=9999999, retry_backoff_ms=2222)

    tasks = mocker.patch.object(cli._delayed_tasks, 'next_at')
    tasks.return_value = 9999999

    ttl = mocker.patch.object(cli.cluster, 'ttl')
    ttl.return_value = 1234

    cli.poll(timeout_ms=9999999, sleep=True)
    _poll.assert_called_with(1.234, sleep=True)
Esempio n. 20
0
def test_initiate_connect(conn):
    cli = KafkaClient()
    try:
        # Node not in metadata, raises AssertionError
        cli._initiate_connect(2)
    except AssertionError:
        pass
    else:
        assert False, 'Exception not raised'

    assert 0 not in cli._conns
    state = cli._initiate_connect(0)
    assert cli._conns[0] is conn
    assert state is conn.state
Esempio n. 21
0
def test_finish_connect(conn):
    cli = KafkaClient()
    try:
        # Node not in metadata, raises AssertionError
        cli._initiate_connect(2)
    except AssertionError:
        pass
    else:
        assert False, 'Exception not raised'

    assert 0 not in cli._conns
    cli._initiate_connect(0)

    conn.connect.return_value = ConnectionStates.CONNECTING
    state = cli._finish_connect(0)
    assert 0 in cli._connecting
    assert state is ConnectionStates.CONNECTING

    conn.connect.return_value = ConnectionStates.CONNECTED
    state = cli._finish_connect(0)
    assert 0 not in cli._connecting
    assert state is ConnectionStates.CONNECTED

    # Failure to connect should trigger metadata update
    assert not cli.cluster._need_update
    cli._connecting.add(0)
    conn.connect.return_value = ConnectionStates.DISCONNECTED
    state = cli._finish_connect(0)
    assert 0 not in cli._connecting
    assert state is ConnectionStates.DISCONNECTED
    assert cli.cluster._need_update
Esempio n. 22
0
def test_init(conn):
    cli = KafkaClient()
    coordinator = ConsumerCoordinator(cli, SubscriptionState())

    # metadata update on init 
    assert cli.cluster._need_update is True
    assert WeakMethod(coordinator._handle_metadata_update) in cli.cluster._listeners
Esempio n. 23
0
def test_autocommit_enable_api_version(conn, api_version):
    coordinator = ConsumerCoordinator(
        KafkaClient(), SubscriptionState(), api_version=api_version)
    if api_version < (0, 8, 1):
        assert coordinator._auto_commit_task is None
    else:
        assert coordinator._auto_commit_task is not None
Esempio n. 24
0
 def __init__(self, **configs):
     self.config = copy.copy(self.DEFAULT_CONFIG)
     self.config.update(configs)
     self._client = KafkaClient(**self.config)
     self._coordinator_id = None
     self.group_id = configs['group_id']
     self.topic = configs['topic']
Esempio n. 25
0
    def __init__(self, **configs):
        log.debug("Starting KafkaAdminClient with configuration: %s", configs)
        extra_configs = set(configs).difference(self.DEFAULT_CONFIG)
        if extra_configs:
            raise KafkaConfigurationError("Unrecognized configs: {}".format(extra_configs))

        self.config = copy.copy(self.DEFAULT_CONFIG)
        self.config.update(configs)

        # Configure metrics
        metrics_tags = {'client-id': self.config['client_id']}
        metric_config = MetricConfig(samples=self.config['metrics_num_samples'],
                                     time_window_ms=self.config['metrics_sample_window_ms'],
                                     tags=metrics_tags)
        reporters = [reporter() for reporter in self.config['metric_reporters']]
        self._metrics = Metrics(metric_config, reporters)

        self._client = KafkaClient(metrics=self._metrics,
                                   metric_group_prefix='admin',
                                   **self.config)

        # Get auto-discovered version from client if necessary
        if self.config['api_version'] is None:
            self.config['api_version'] = self._client.config['api_version']

        self._closed = False
        self._refresh_controller_id()
        log.debug("KafkaAdminClient started.")
Esempio n. 26
0
def test_maybe_connect(conn):
    cli = KafkaClient()
    try:
        # Node not in metadata, raises AssertionError
        cli._maybe_connect(2)
    except AssertionError:
        pass
    else:
        assert False, 'Exception not raised'

    # New node_id creates a conn object
    assert 0 not in cli._conns
    conn.state = ConnectionStates.DISCONNECTED
    conn.connect.side_effect = lambda: conn._set_conn_state(ConnectionStates.CONNECTING)
    assert cli._maybe_connect(0) is False
    assert cli._conns[0] is conn
def test_maybe_auto_commit_offsets_sync(mocker, api_version, group_id, enable,
                                        error, has_auto_commit, commit_offsets,
                                        warn, exc):
    mock_warn = mocker.patch('kafka.coordinator.consumer.log.warning')
    mock_exc = mocker.patch('kafka.coordinator.consumer.log.exception')
    client = KafkaClient(api_version=api_version)
    coordinator = ConsumerCoordinator(client, SubscriptionState(),
                                      Metrics(),
                                      api_version=api_version,
                                      session_timeout_ms=30000,
                                      max_poll_interval_ms=30000,
                                      enable_auto_commit=enable,
                                      group_id=group_id)
    commit_sync = mocker.patch.object(coordinator, 'commit_offsets_sync',
                                      side_effect=error)
    if has_auto_commit:
        assert coordinator.next_auto_commit_deadline is not None
    else:
        assert coordinator.next_auto_commit_deadline is None

    assert coordinator._maybe_auto_commit_offsets_sync() is None

    if has_auto_commit:
        assert coordinator.next_auto_commit_deadline is not None

    assert commit_sync.call_count == (1 if commit_offsets else 0)
    assert mock_warn.call_count == (1 if warn else 0)
    assert mock_exc.call_count == (1 if exc else 0)
Esempio n. 28
0
def test_maybe_auto_commit_offsets_sync(mocker, api_version, group_id, enable,
                                        error, has_auto_commit, commit_offsets,
                                        warn, exc):
    mock_warn = mocker.patch('kafka.coordinator.consumer.log.warning')
    mock_exc = mocker.patch('kafka.coordinator.consumer.log.exception')
    coordinator = ConsumerCoordinator(KafkaClient(),
                                      SubscriptionState(),
                                      Metrics(),
                                      'consumer',
                                      api_version=api_version,
                                      enable_auto_commit=enable,
                                      group_id=group_id)
    commit_sync = mocker.patch.object(coordinator,
                                      'commit_offsets_sync',
                                      side_effect=error)
    if has_auto_commit:
        assert coordinator._auto_commit_task is not None
        coordinator._auto_commit_task.enable()
        assert coordinator._auto_commit_task._enabled is True
    else:
        assert coordinator._auto_commit_task is None

    assert coordinator._maybe_auto_commit_offsets_sync() is None

    if has_auto_commit:
        assert coordinator._auto_commit_task is not None
        assert coordinator._auto_commit_task._enabled is False

    assert commit_sync.call_count == (1 if commit_offsets else 0)
    assert mock_warn.call_count == (1 if warn else 0)
    assert mock_exc.call_count == (1 if exc else 0)
Esempio n. 29
0
def send_task(logger, zk_cli):
    # get next task
    task = tasks.next_task(zk_cli, TASK_DEFINITION_PATH)

    topic = JobSourceTopic()
    # get postfix
    val, stats = zk_cli.get(KAFKA_POSTFIX_PATH)
    producer = topic.init_producer()
    producer.change_postfix(val)

    kfk_cli = KafkaClient(bootstrap_servers=settings.KAFKA_SERVER)
    # current topics (error_code, topic, is_internal, partitions)
    is_fetch_metadata_success = False
    try:
        topics_res = kafka_utils.get_metadata(kfk_cli, 10000)
        if topics_res is not None:
            topics = (x[1] for x in topics_res.topics)
            logger.debug("Current topics in kafka: %s" % str(list(topics)))
            is_fetch_metadata_success = True
    except Exception:
        pass
    if not is_fetch_metadata_success:
        logger.warn("Failed to fetch metadata from kafka")

    producer.produce(task)
    logger.info('sent to Topic %s, feed=%2d. %s' %
                (producer._topic_name, task['order'], task['name']))
    return task
Esempio n. 30
0
 def get_clients(self, cnt=1, client_id=None):
     if client_id is None:
         client_id = 'client'
     return tuple(
         KafkaClient(client_id='%s_%s' % (client_id, random_string(4)),
                     bootstrap_servers=self.bootstrap_server())
         for x in range(cnt))
Esempio n. 31
0
def test_maybe_connect(conn):
    cli = KafkaClient()
    try:
        # Node not in metadata, raises AssertionError
        cli._maybe_connect(2)
    except AssertionError:
        pass
    else:
        assert False, 'Exception not raised'

    # New node_id creates a conn object
    assert 0 not in cli._conns
    conn.state = ConnectionStates.DISCONNECTED
    conn.connect.side_effect = lambda: conn._set_conn_state(ConnectionStates.CONNECTING)
    assert cli._maybe_connect(0) is False
    assert cli._conns[0] is conn
Esempio n. 32
0
 def init_kafka_client(self):
     try:
         self.kafka_client = KafkaClient(
             api_version_auto_timeout_ms=self.api_version_auto_timeout_ms,
             bootstrap_servers=self.config["bootstrap_uri"],
             client_id=self.config["client_id"],
             security_protocol=self.config["security_protocol"],
             ssl_cafile=self.config["ssl_cafile"],
             ssl_certfile=self.config["ssl_certfile"],
             ssl_keyfile=self.config["ssl_keyfile"],
         )
         return True
     except (NodeNotReadyError, NoBrokersAvailable):
         self.log.warning("No Brokers available yet, retrying init_kafka_client()")
         time.sleep(2.0)
     return False
Esempio n. 33
0
def test_bootstrap_failure(conn):
    conn.state = ConnectionStates.DISCONNECTED
    cli = KafkaClient()
    conn.assert_called_once_with('localhost', 9092, **cli.config)
    conn.connect.assert_called_with()
    conn.close.assert_called_with()
    assert cli._bootstrap_fails == 1
    assert cli.cluster.brokers() == set()
def test_bootstrap(mocker, conn):
    conn.state = ConnectionStates.CONNECTED
    cli = KafkaClient(api_version=(0, 9))
    mocker.patch.object(cli, '_selector')
    future = cli.cluster.request_update()
    cli.poll(future=future)

    assert future.succeeded()
    args, kwargs = conn.call_args
    assert args == ('localhost', 9092, socket.AF_UNSPEC)
    kwargs.pop('state_change_callback')
    kwargs.pop('node_id')
    assert kwargs == cli.config
    conn.send.assert_called_once_with(MetadataRequest[0]([]), blocking=False)
    assert cli._bootstrap_fails == 0
    assert cli.cluster.brokers() == set([BrokerMetadata(0, 'foo', 12, None),
                                         BrokerMetadata(1, 'bar', 34, None)])
Esempio n. 35
0
    def __init__(self, *topics, **configs):
        self.config = copy.copy(self.DEFAULT_CONFIG)
        for key in self.config:
            if key in configs:
                self.config[key] = configs.pop(key)

        # Only check for extra config keys in top-level class
        assert not configs, 'Unrecognized configs: %s' % configs

        deprecated = {'smallest': 'earliest', 'largest': 'latest'}
        if self.config['auto_offset_reset'] in deprecated:
            new_config = deprecated[self.config['auto_offset_reset']]
            log.warning('use auto_offset_reset=%s (%s is deprecated)',
                        new_config, self.config['auto_offset_reset'])
            self.config['auto_offset_reset'] = new_config

        self._client = KafkaClient(**self.config)

        # Check Broker Version if not set explicitly
        if self.config['api_version'] == 'auto':
            self.config['api_version'] = self._client.check_version()
        assert self.config['api_version'] in (
            '0.9', '0.8.2', '0.8.1', '0.8.0'), 'Unrecognized api version'

        # Convert api_version config to tuple for easy comparisons
        self.config['api_version'] = tuple(
            map(int, self.config['api_version'].split('.')))

        self._subscription = SubscriptionState(
            self.config['auto_offset_reset'])
        self._fetcher = Fetcher(self._client, self._subscription,
                                **self.config)
        self._coordinator = ConsumerCoordinator(
            self._client,
            self._subscription,
            assignors=self.config['partition_assignment_strategy'],
            **self.config)
        self._closed = False
        self._iterator = None
        self._consumer_timeout = float('inf')

        #self.metrics = None
        if topics:
            self._subscription.subscribe(topics=topics)
            self._client.set_topics(topics)
Esempio n. 36
0
def test_bootstrap_success(conn):
    conn.state = ConnectionStates.CONNECTED
    cli = KafkaClient()
    conn.assert_called_once_with('localhost', 9092, **cli.config)
    conn.connect.assert_called_with()
    conn.send.assert_called_once_with(MetadataRequest([]))
    assert cli._bootstrap_fails == 0
    assert cli.cluster.brokers() == set([BrokerMetadata(0, 'foo', 12),
                                         BrokerMetadata(1, 'bar', 34)])
Esempio n. 37
0
 def _determine_kafka_version(cls, init_config, instance):
     """Return the Kafka cluster version as a tuple."""
     kafka_version = instance.get('kafka_client_api_version')
     if isinstance(kafka_version, str):
         kafka_version = tuple(map(int, kafka_version.split(".")))
     if kafka_version is None:  # if unspecified by the user, we have to probe the cluster
         kafka_connect_str = instance.get(
             'kafka_connect_str')  # TODO call validation method
         kafka_client = KafkaClient(
             bootstrap_servers=kafka_connect_str,
             client_id='dd-agent',
             request_timeout_ms=init_config.get(
                 'kafka_timeout', DEFAULT_KAFKA_TIMEOUT) * 1000,
             # if `kafka_client_api_version` is not set, then kafka-python automatically probes the cluster for
             # broker version during the bootstrapping process. Note that this returns the first version found, so in
             # a mixed-version cluster this will be a non-deterministic result.
             api_version=kafka_version,
             # While we check for SASL/SSL params, if not present they will default to the kafka-python values for
             # plaintext connections
             security_protocol=instance.get('security_protocol',
                                            'PLAINTEXT'),
             sasl_mechanism=instance.get('sasl_mechanism'),
             sasl_plain_username=instance.get('sasl_plain_username'),
             sasl_plain_password=instance.get('sasl_plain_password'),
             sasl_kerberos_service_name=instance.get(
                 'sasl_kerberos_service_name', 'kafka'),
             sasl_kerberos_domain_name=instance.get(
                 'sasl_kerberos_domain_name'),
             ssl_cafile=instance.get('ssl_cafile'),
             ssl_check_hostname=instance.get('ssl_check_hostname', True),
             ssl_certfile=instance.get('ssl_certfile'),
             ssl_keyfile=instance.get('ssl_keyfile'),
             ssl_crlfile=instance.get('ssl_crlfile'),
             ssl_password=instance.get('ssl_password'),
         )
         # version probing happens automatically as part of KafkaClient's __init__()
         kafka_version = kafka_client.config['api_version']
         # Currently, this client is only used for probing, so we need to close it to avoid stale connections on
         # older Kafka brokers. We can't re-use in new code path because KafkaAdminClient doesn't currently support
         # passing in an existing client.
         # TODO this could be re-used by the legacy version of the check to make maintenance easier... ie, we don't
         # have multiple sections of code instantiating clients
         kafka_client.close()
     return kafka_version
Esempio n. 38
0
    def __init__(self, **configs):
        log.debug("Starting Kafka administration interface")
        extra_configs = set(configs).difference(self.DEFAULT_CONFIG)
        if extra_configs:
            raise KafkaConfigurationError("Unrecognized configs: %s" %
                                          extra_configs)

        self.config = copy.copy(self.DEFAULT_CONFIG)
        self.config.update(configs)

        # api_version was previously a str. accept old format for now
        if isinstance(self.config['api_version'], str):
            deprecated = self.config['api_version']
            if deprecated == 'auto':
                self.config['api_version'] = None
            else:
                self.config['api_version'] = tuple(
                    map(int, deprecated.split('.')))
            log.warning(
                'use api_version=%s [tuple] -- "%s" as str is deprecated',
                str(self.config['api_version']), deprecated)

        # Configure metrics
        metrics_tags = {'client-id': self.config['client_id']}
        metric_config = MetricConfig(
            samples=self.config['metrics_num_samples'],
            time_window_ms=self.config['metrics_sample_window_ms'],
            tags=metrics_tags)
        reporters = [
            reporter() for reporter in self.config['metric_reporters']
        ]
        self._metrics = Metrics(metric_config, reporters)

        self._client = KafkaClient(metrics=self._metrics,
                                   metric_group_prefix='admin',
                                   **self.config)

        # Get auto-discovered version from client if necessary
        if self.config['api_version'] is None:
            self.config['api_version'] = self._client.config['api_version']

        self._closed = False
        self._refresh_controller_id()
        log.debug('Kafka administration interface started')
Esempio n. 39
0
def test_maybe_refresh_metadata_backoff(mocker):
    mocker.patch.object(KafkaClient, '_bootstrap')
    _poll = mocker.patch.object(KafkaClient, '_poll')

    cli = KafkaClient(request_timeout_ms=9999999, retry_backoff_ms=2222)

    tasks = mocker.patch.object(cli._delayed_tasks, 'next_at')
    tasks.return_value = 9999999

    ttl = mocker.patch.object(cli.cluster, 'ttl')
    ttl.return_value = 0

    now = time.time()
    t = mocker.patch('time.time')
    t.return_value = now
    cli._last_no_node_available_ms = now * 1000

    cli.poll(timeout_ms=9999999, sleep=True)
    _poll.assert_called_with(2.222, sleep=True)
Esempio n. 40
0
    def __init__(self, *topics, **configs):
        self.config = copy.copy(self.DEFAULT_CONFIG)
        for key in self.config:
            if key in configs:
                self.config[key] = configs.pop(key)

        # Only check for extra config keys in top-level class
        assert not configs, 'Unrecognized configs: %s' % configs

        deprecated = {'smallest': 'earliest', 'largest': 'latest'}
        if self.config['auto_offset_reset'] in deprecated:
            new_config = deprecated[self.config['auto_offset_reset']]
            log.warning('use auto_offset_reset=%s (%s is deprecated)',
                        new_config, self.config['auto_offset_reset'])
            self.config['auto_offset_reset'] = new_config

        metrics_tags = {'client-id': self.config['client_id']}
        metric_config = MetricConfig(samples=self.config['metrics_num_samples'],
                                     time_window_ms=self.config['metrics_sample_window_ms'],
                                     tags=metrics_tags)
        reporters = [reporter() for reporter in self.config['metric_reporters']]
        self._metrics = Metrics(metric_config, reporters)
        metric_group_prefix = 'consumer'
        # TODO _metrics likely needs to be passed to KafkaClient, etc.

        # api_version was previously a str. accept old format for now
        if isinstance(self.config['api_version'], str):
            str_version = self.config['api_version']
            if str_version == 'auto':
                self.config['api_version'] = None
            else:
                self.config['api_version'] = tuple(map(int, str_version.split('.')))
            log.warning('use api_version=%s (%s is deprecated)',
                        str(self.config['api_version']), str_version)

        self._client = KafkaClient(**self.config)

        # Get auto-discovered version from client if necessary
        if self.config['api_version'] is None:
            self.config['api_version'] = self._client.config['api_version']

        self._subscription = SubscriptionState(self.config['auto_offset_reset'])
        self._fetcher = Fetcher(
            self._client, self._subscription, self._metrics, metric_group_prefix, **self.config)
        self._coordinator = ConsumerCoordinator(
            self._client, self._subscription, self._metrics, metric_group_prefix,
            assignors=self.config['partition_assignment_strategy'],
            **self.config)
        self._closed = False
        self._iterator = None
        self._consumer_timeout = float('inf')

        if topics:
            self._subscription.subscribe(topics=topics)
            self._client.set_topics(topics)
Esempio n. 41
0
def test_maybe_refresh_metadata_update(mocker):
    mocker.patch.object(KafkaClient, '_bootstrap')
    _poll = mocker.patch.object(KafkaClient, '_poll')

    cli = KafkaClient(request_timeout_ms=9999999, retry_backoff_ms=2222)

    tasks = mocker.patch.object(cli._delayed_tasks, 'next_at')
    tasks.return_value = 9999999

    ttl = mocker.patch.object(cli.cluster, 'ttl')
    ttl.return_value = 0

    mocker.patch.object(cli, 'least_loaded_node', return_value='foobar')
    mocker.patch.object(cli, '_can_send_request', return_value=True)
    send = mocker.patch.object(cli, 'send')

    cli.poll(timeout_ms=9999999, sleep=True)
    _poll.assert_called_with(0, sleep=True)
    assert cli._metadata_refresh_in_progress
    request = MetadataRequest[0]([])
    send.assert_called_with('foobar', request)
Esempio n. 42
0
def test_maybe_connect(conn):
    cli = KafkaClient()
    try:
        # Node not in metadata, raises AssertionError
        cli._maybe_connect(2)
    except AssertionError:
        pass
    else:
        assert False, 'Exception not raised'

    assert 0 not in cli._conns
    conn.state = ConnectionStates.DISCONNECTED
    conn.connect.side_effect = lambda: ConnectionStates.CONNECTING
    assert cli._maybe_connect(0) is False
    assert cli._conns[0] is conn
    assert 0 in cli._connecting

    conn.state = ConnectionStates.CONNECTING
    conn.connect.side_effect = lambda: ConnectionStates.CONNECTED
    assert cli._maybe_connect(0) is True
    assert 0 not in cli._connecting

    # Failure to connect should trigger metadata update
    assert cli.cluster._need_update is False
    cli._connecting.add(0)
    conn.state = ConnectionStates.CONNECTING
    conn.connect.side_effect = lambda: ConnectionStates.DISCONNECTED
    assert cli._maybe_connect(0) is False
    assert 0 not in cli._connecting
    assert cli.cluster._need_update is True
Esempio n. 43
0
def test_maybe_refresh_metadata_failure(mocker):
    mocker.patch.object(KafkaClient, '_bootstrap')
    _poll = mocker.patch.object(KafkaClient, '_poll')

    cli = KafkaClient(request_timeout_ms=9999999, retry_backoff_ms=2222)

    tasks = mocker.patch.object(cli._delayed_tasks, 'next_at')
    tasks.return_value = 9999999

    ttl = mocker.patch.object(cli.cluster, 'ttl')
    ttl.return_value = 0

    mocker.patch.object(cli, 'least_loaded_node', return_value='foobar')

    now = time.time()
    t = mocker.patch('time.time')
    t.return_value = now

    cli.poll(timeout_ms=9999999, sleep=True)
    _poll.assert_called_with(0, sleep=True)
    assert cli._last_no_node_available_ms == now * 1000
    assert not cli._metadata_refresh_in_progress
Esempio n. 44
0
    def __init__(self, *topics, **configs):
        self.config = copy.copy(self.DEFAULT_CONFIG)
        for key in self.config:
            if key in configs:
                self.config[key] = configs.pop(key)

        # Only check for extra config keys in top-level class
        assert not configs, 'Unrecognized configs: %s' % configs

        deprecated = {'smallest': 'earliest', 'largest': 'latest' }
        if self.config['auto_offset_reset'] in deprecated:
            new_config = deprecated[self.config['auto_offset_reset']]
            log.warning('use auto_offset_reset=%s (%s is deprecated)',
                        new_config, self.config['auto_offset_reset'])
            self.config['auto_offset_reset'] = new_config

        metrics_tags = {'client-id': self.config['client_id']}
        metric_config = MetricConfig(samples=self.config['metrics_num_samples'],
                                     time_window_ms=self.config['metrics_sample_window_ms'],
                                     tags=metrics_tags)
        reporters = [reporter() for reporter in self.config['metric_reporters']]
        reporters.append(DictReporter('kafka.consumer'))
        self._metrics = Metrics(metric_config, reporters)
        metric_group_prefix = 'consumer'
        # TODO _metrics likely needs to be passed to KafkaClient, etc.

        self._client = KafkaClient(**self.config)

        # Check Broker Version if not set explicitly
        if self.config['api_version'] == 'auto':
            self.config['api_version'] = self._client.check_version()
        assert self.config['api_version'] in ('0.10', '0.9', '0.8.2', '0.8.1', '0.8.0'), 'Unrecognized api version'

        # Convert api_version config to tuple for easy comparisons
        self.config['api_version'] = tuple(
            map(int, self.config['api_version'].split('.')))

        self._subscription = SubscriptionState(self.config['auto_offset_reset'])
        self._fetcher = Fetcher(
            self._client, self._subscription, self._metrics, metric_group_prefix, **self.config)
        self._coordinator = ConsumerCoordinator(
            self._client, self._subscription, self._metrics, metric_group_prefix,
            assignors=self.config['partition_assignment_strategy'],
            **self.config)
        self._closed = False
        self._iterator = None
        self._consumer_timeout = float('inf')

        if topics:
            self._subscription.subscribe(topics=topics)
            self._client.set_topics(topics)
Esempio n. 45
0
def test_set_topics(mocker):
    request_update = mocker.patch.object(ClusterMetadata, 'request_update')
    request_update.side_effect = lambda: Future()
    cli = KafkaClient(api_version=(0, 10))

    # replace 'empty' with 'non empty'
    request_update.reset_mock()
    fut = cli.set_topics(['t1', 't2'])
    assert not fut.is_done
    request_update.assert_called_with()

    # replace 'non empty' with 'same'
    request_update.reset_mock()
    fut = cli.set_topics(['t1', 't2'])
    assert fut.is_done
    assert fut.value == set(['t1', 't2'])
    request_update.assert_not_called()

    # replace 'non empty' with 'empty'
    request_update.reset_mock()
    fut = cli.set_topics([])
    assert fut.is_done
    assert fut.value == set()
    request_update.assert_not_called()
Esempio n. 46
0
def test_is_disconnected(conn):
    cli = KafkaClient()

    # False if not connected yet
    conn.state = ConnectionStates.DISCONNECTED
    assert not cli.is_disconnected(0)

    cli._maybe_connect(0)
    assert cli.is_disconnected(0)

    conn.state = ConnectionStates.CONNECTING
    assert not cli.is_disconnected(0)

    conn.state = ConnectionStates.CONNECTED
    assert not cli.is_disconnected(0)
Esempio n. 47
0
def test_send(conn):
    cli = KafkaClient()
    try:
        cli.send(2, None)
    except Errors.NodeNotReadyError:
        pass
    else:
        assert False, 'NodeNotReadyError not raised'

    cli._initiate_connect(0)
    # ProduceRequest w/ 0 required_acks -> no response
    request = ProduceRequest(0, 0, [])
    ret = cli.send(0, request)
    assert conn.send.called_with(request, expect_response=False)
    assert isinstance(ret, Future)

    request = MetadataRequest([])
    cli.send(0, request)
    assert conn.send.called_with(request, expect_response=True)
Esempio n. 48
0
def test_can_connect(conn):
    cli = KafkaClient()

    # Node is not in broker metadata - cant connect
    assert not cli._can_connect(2)

    # Node is in broker metadata but not in _conns
    assert 0 not in cli._conns
    assert cli._can_connect(0)

    # Node is connected, can't reconnect
    cli._initiate_connect(0)
    assert not cli._can_connect(0)

    # Node is disconnected, can connect
    cli._conns[0].state = ConnectionStates.DISCONNECTED
    assert cli._can_connect(0)

    # Node is disconnected, but blacked out
    conn.blacked_out.return_value = True
    assert not cli._can_connect(0)
Esempio n. 49
0
    def __init__(self, *topics, **configs):
        self.config = copy.copy(self.DEFAULT_CONFIG)
        for key in self.config:
            if key in configs:
                self.config[key] = configs.pop(key)

        # Only check for extra config keys in top-level class
        assert not configs, 'Unrecognized configs: %s' % configs

        deprecated = {'smallest': 'earliest', 'largest': 'latest' }
        if self.config['auto_offset_reset'] in deprecated:
            new_config = deprecated[self.config['auto_offset_reset']]
            log.warning('use auto_offset_reset=%s (%s is deprecated)',
                        new_config, self.config['auto_offset_reset'])
            self.config['auto_offset_reset'] = new_config

        self._client = KafkaClient(**self.config)

        # Check Broker Version if not set explicitly
        if self.config['api_version'] == 'auto':
            self.config['api_version'] = self._client.check_version()
        assert self.config['api_version'] in ('0.9', '0.8.2', '0.8.1', '0.8.0'), 'Unrecognized api version'

        # Convert api_version config to tuple for easy comparisons
        self.config['api_version'] = tuple(
            map(int, self.config['api_version'].split('.')))

        self._subscription = SubscriptionState(self.config['auto_offset_reset'])
        self._fetcher = Fetcher(
            self._client, self._subscription, **self.config)
        self._coordinator = ConsumerCoordinator(
            self._client, self._subscription,
            assignors=self.config['partition_assignment_strategy'],
            **self.config)
        self._closed = False
        self._iterator = None
        self._consumer_timeout = float('inf')

        #self.metrics = None
        if topics:
            self._subscription.subscribe(topics=topics)
            self._client.set_topics(topics)
Esempio n. 50
0
def test_poll(mocker):
    mocker.patch.object(KafkaClient, '_bootstrap')
    metadata = mocker.patch.object(KafkaClient, '_maybe_refresh_metadata')
    _poll = mocker.patch.object(KafkaClient, '_poll')
    cli = KafkaClient(api_version=(0, 9))

    # metadata timeout wins
    metadata.return_value = 1000
    cli.poll()
    _poll.assert_called_with(1.0)

    # user timeout wins
    cli.poll(250)
    _poll.assert_called_with(0.25)

    # default is request_timeout_ms
    metadata.return_value = 1000000
    cli.poll()
    _poll.assert_called_with(cli.config['request_timeout_ms'] / 1000.0)
Esempio n. 51
0
    def __init__(self, *topics, **configs):
        self.config = copy.copy(self.DEFAULT_CONFIG)
        for key in self.config:
            if key in configs:
                self.config[key] = configs.pop(key)

        # Only check for extra config keys in top-level class
        assert not configs, "Unrecognized configs: %s" % configs

        deprecated = {"smallest": "earliest", "largest": "latest"}
        if self.config["auto_offset_reset"] in deprecated:
            new_config = deprecated[self.config["auto_offset_reset"]]
            log.warning("use auto_offset_reset=%s (%s is deprecated)", new_config, self.config["auto_offset_reset"])
            self.config["auto_offset_reset"] = new_config

        self._client = KafkaClient(**self.config)

        # Check Broker Version if not set explicitly
        if self.config["api_version"] == "auto":
            self.config["api_version"] = self._client.check_version()
        assert self.config["api_version"] in ("0.9", "0.8.2", "0.8.1", "0.8.0")

        # Convert api_version config to tuple for easy comparisons
        self.config["api_version"] = tuple(map(int, self.config["api_version"].split(".")))

        self._subscription = SubscriptionState(self.config["auto_offset_reset"])
        self._fetcher = Fetcher(self._client, self._subscription, **self.config)
        self._coordinator = ConsumerCoordinator(
            self._client, self._subscription, assignors=self.config["partition_assignment_strategy"], **self.config
        )
        self._closed = False
        self._iterator = None
        self._consumer_timeout = float("inf")

        # self.metrics = None
        if topics:
            self._subscription.subscribe(topics=topics)
            self._client.set_topics(topics)
Esempio n. 52
0
def test_is_ready(mocker, conn):
    cli = KafkaClient()
    cli._maybe_connect(0)
    cli._maybe_connect(1)

    # metadata refresh blocks ready nodes
    assert cli.is_ready(0)
    assert cli.is_ready(1)
    cli._metadata_refresh_in_progress = True
    assert not cli.is_ready(0)
    assert not cli.is_ready(1)

    # requesting metadata update also blocks ready nodes
    cli._metadata_refresh_in_progress = False
    assert cli.is_ready(0)
    assert cli.is_ready(1)
    cli.cluster.request_update()
    cli.cluster.config['retry_backoff_ms'] = 0
    assert not cli._metadata_refresh_in_progress
    assert not cli.is_ready(0)
    assert not cli.is_ready(1)
    cli.cluster._need_update = False

    # if connection can't send more, not ready
    assert cli.is_ready(0)
    conn.can_send_more.return_value = False
    assert not cli.is_ready(0)
    conn.can_send_more.return_value = True

    # disconnected nodes, not ready
    assert cli.is_ready(0)
    conn.state = ConnectionStates.DISCONNECTED
    assert not cli.is_ready(0)
Esempio n. 53
0
def test_ready(mocker, conn):
    cli = KafkaClient()
    maybe_connect = mocker.patch.object(cli, '_maybe_connect')
    node_id = 1
    cli.ready(node_id)
    maybe_connect.assert_called_with(node_id)
Esempio n. 54
0
class OffsetsFetcherAsync(object):

    DEFAULT_CONFIG = {
        'session_timeout_ms': 30000,
        'heartbeat_interval_ms': 3000,
        'retry_backoff_ms': 100,
        'api_version': (0, 9),
        'metric_group_prefix': ''
    }

    def __init__(self, **configs):
        self.config = copy.copy(self.DEFAULT_CONFIG)
        self.config.update(configs)
        self._client = KafkaClient(**self.config)
        self._coordinator_id = None
        self.group_id = configs['group_id']
        self.topic = configs['topic']

    def _ensure_coordinator_known(self):
        """Block until the coordinator for this group is known
        (and we have an active connection -- java client uses unsent queue).
        """
        while self._coordinator_unknown():

            # Prior to 0.8.2 there was no group coordinator
            # so we will just pick a node at random and treat
            # it as the "coordinator"
            if self.config['api_version'] < (0, 8, 2):
                self._coordinator_id = self._client.least_loaded_node()
                self._client.ready(self._coordinator_id)
                continue

            future = self._send_group_coordinator_request()
            self._client.poll(future=future)

            if future.failed():
                if isinstance(future.exception,
                              Errors.GroupCoordinatorNotAvailableError):
                    continue
                elif future.retriable():
                    metadata_update = self._client.cluster.request_update()
                    self._client.poll(future=metadata_update)
                else:
                    raise future.exception  # pylint: disable-msg=raising-bad-type

    def _coordinator_unknown(self):
        """Check if we know who the coordinator is and have an active connection

        Side-effect: reset _coordinator_id to None if connection failed

        Returns:
            bool: True if the coordinator is unknown
        """
        if self._coordinator_id is None:
            return True

        if self._client.is_disconnected(self._coordinator_id):
            self._coordinator_dead()
            return True

        return False

    def _coordinator_dead(self, error=None):
        """Mark the current coordinator as dead."""
        if self._coordinator_id is not None:
            log.warning("Marking the coordinator dead (node %s) for group %s: %s.",
                        self._coordinator_id, self.group_id, error)
            self._coordinator_id = None

    def _send_group_coordinator_request(self):
        """Discover the current coordinator for the group.

        Returns:
            Future: resolves to the node id of the coordinator
        """
        node_id = self._client.least_loaded_node()
        if node_id is None:
            return Future().failure(Errors.NoBrokersAvailable())

        log.debug("Sending group coordinator request for group %s to broker %s",
                  self.group_id, node_id)
        request = GroupCoordinatorRequest[0](self.group_id)
        future = Future()
        _f = self._client.send(node_id, request)
        _f.add_callback(self._handle_group_coordinator_response, future)
        _f.add_errback(self._failed_request, node_id, request, future)
        return future

    def _handle_group_coordinator_response(self, future, response):
        log.debug("Received group coordinator response %s", response)
        if not self._coordinator_unknown():
            # We already found the coordinator, so ignore the request
            log.debug("Coordinator already known -- ignoring metadata response")
            future.success(self._coordinator_id)
            return

        error_type = Errors.for_code(response.error_code)
        if error_type is Errors.NoError:
            ok = self._client.cluster.add_group_coordinator(self.group_id, response)
            if not ok:
                # This could happen if coordinator metadata is different
                # than broker metadata
                future.failure(Errors.IllegalStateError())
                return

            self._coordinator_id = response.coordinator_id
            log.info("Discovered coordinator %s for group %s",
                     self._coordinator_id, self.group_id)
            self._client.ready(self._coordinator_id)
            future.success(self._coordinator_id)
        elif error_type is Errors.GroupCoordinatorNotAvailableError:
            log.debug("Group Coordinator Not Available; retry")
            future.failure(error_type())
        elif error_type is Errors.GroupAuthorizationFailedError:
            error = error_type(self.group_id)
            log.error("Group Coordinator Request failed: %s", error)
            future.failure(error)
        else:
            error = error_type()
            log.error("Unrecognized failure in Group Coordinator Request: %s",
                      error)
            future.failure(error)

    def _failed_request(self, node_id, request, future, error):
        log.error('Error sending %s to node %s [%s]',
                  request.__class__.__name__, node_id, error)
        # Marking coordinator dead
        # unless the error is caused by internal client pipelining
        if not isinstance(error, (Errors.NodeNotReadyError,
                                  Errors.TooManyInFlightRequests)):
            self._coordinator_dead()
        future.failure(error)

    def offsets(self, partitions, timestamp):
        """Fetch a single offset before the given timestamp for the set of partitions.

        Blocks until offset is obtained, or a non-retriable exception is raised

        Arguments:
            partitions (iterable of TopicPartition) The partition that needs fetching offset.
            timestamp (int): timestamp for fetching offset. -1 for the latest
                available, -2 for the earliest available. Otherwise timestamp
                is treated as epoch seconds.

        Returns:
            dict: TopicPartition and message offsets
        """
        retries = 3
        while retries > 0:
            offsets = {}
            for future in self._send_offset_request(partitions, timestamp):
                self._client.poll(future=future)

                if future.succeeded():
                    for tp, offset in future.value:
                        offsets[tp] = offset
                    continue

                if not future.retriable():
                    raise future.exception  # pylint: disable-msg=raising-bad-type

                if future.exception.invalid_metadata:
                    refresh_future = self._client.cluster.request_update()
                    self._client.poll(future=refresh_future)
                    log.warning("Got exception %s and kept the loop", future.exception)
            if offsets:
                return offsets
            retries -= 1
            log.warning("Retrying the offsets fetch loop (%d retries left)", retries)
        log.error("Unsuccessful offsets retrieval")
        return {}

    def _send_offset_request(self, partitions, timestamp):
        """Fetch a single offset before the given timestamp for the partition.

        Arguments:
            partitions iterable of TopicPartition: partitions that needs fetching offset
            timestamp (int): timestamp for fetching offset

        Returns:
            list of Future: resolves to the corresponding offset
        """
        topic = partitions[0].topic
        nodes_per_partitions = {}
        for partition in partitions:
            node_id = self._client.cluster.leader_for_partition(partition)
            if node_id is None:
                log.debug("Partition %s is unknown for fetching offset,"
                          " wait for metadata refresh", partition)
                return [Future().failure(Errors.StaleMetadata(partition))]
            elif node_id == -1:
                log.debug("Leader for partition %s unavailable for fetching offset,"
                          " wait for metadata refresh", partition)
                return [Future().failure(Errors.LeaderNotAvailableError(partition))]
            nodes_per_partitions.setdefault(node_id, []).append(partition)

        # Client returns a future that only fails on network issues
        # so create a separate future and attach a callback to update it
        # based on response error codes

        futures = []
        for node_id, partitions in six.iteritems(nodes_per_partitions):
            request = OffsetRequest[0](
                -1, [(topic, [(partition.partition, timestamp, 1) for partition in partitions])]
            )
            future_request = Future()
            _f = self._client.send(node_id, request)
            _f.add_callback(self._handle_offset_response, partitions, future_request)

            def errback(e):
                log.error("Offset request errback error %s", e)
                future_request.failure(e)
            _f.add_errback(errback)
            futures.append(future_request)

        return futures

    def _handle_offset_response(self, partitions, future, response):
        """Callback for the response of the list offset call above.

        Arguments:
            partition (TopicPartition): The partition that was fetched
            future (Future): the future to update based on response
            response (OffsetResponse): response from the server

        Raises:
            AssertionError: if response does not match partition
        """
        topic, partition_info = response.topics[0]
        assert len(response.topics) == 1, (
            'OffsetResponse should only be for a single topic')
        partition_ids = set([part.partition for part in partitions])
        result = []
        for pi in partition_info:
            part, error_code, offsets = pi
            assert topic == partitions[0].topic and part in partition_ids, (
                'OffsetResponse partition does not match OffsetRequest partition')
            error_type = Errors.for_code(error_code)
            if error_type is Errors.NoError:
                assert len(offsets) == 1, 'Expected OffsetResponse with one offset'
                log.debug("Fetched offset %s for partition %d", offsets[0], part)
                result.append((TopicPartition(topic, part), offsets[0]))
            elif error_type in (Errors.NotLeaderForPartitionError,
                                Errors.UnknownTopicOrPartitionError):
                log.debug("Attempt to fetch offsets for partition %s failed due"
                          " to obsolete leadership information, retrying.",
                          str(partitions))
                future.failure(error_type(partitions))
            else:
                log.warning("Attempt to fetch offsets for partition %s failed due to:"
                            " %s", partitions, error_type)
                future.failure(error_type(partitions))
        future.success(result)

    def fetch_committed_offsets(self, partitions):
        """Fetch the current committed offsets for specified partitions

        Arguments:
            partitions (list of TopicPartition): partitions to fetch

        Returns:
            dict: {TopicPartition: OffsetAndMetadata}
        """
        if not partitions:
            return {}

        while True:
            self._ensure_coordinator_known()

            # contact coordinator to fetch committed offsets
            future = self._send_offset_fetch_request(partitions)
            self._client.poll(future=future)

            if future.succeeded():
                return future.value

            if not future.retriable():
                raise future.exception  # pylint: disable-msg=raising-bad-type

            time.sleep(self.config['retry_backoff_ms'] / 1000.0)

    def _send_offset_fetch_request(self, partitions):
        """Fetch the committed offsets for a set of partitions.

        This is a non-blocking call. The returned future can be polled to get
        the actual offsets returned from the broker.

        Arguments:
            partitions (list of TopicPartition): the partitions to fetch

        Returns:
            Future: resolves to dict of offsets: {TopicPartition: int}
        """
        assert self.config['api_version'] >= (0, 8, 1), 'Unsupported Broker API'
        assert all(map(lambda k: isinstance(k, TopicPartition), partitions))
        if not partitions:
            return Future().success({})

        elif self._coordinator_unknown():
            return Future().failure(Errors.GroupCoordinatorNotAvailableError)

        node_id = self._coordinator_id

        # Verify node is ready
        if not self._client.ready(node_id):
            log.debug("Node %s not ready -- failing offset fetch request",
                      node_id)
            return Future().failure(Errors.NodeNotReadyError)

        log.debug("Group %s fetching committed offsets for partitions: %s",
                  self.group_id, partitions)
        # construct the request
        topic_partitions = collections.defaultdict(set)
        for tp in partitions:
            topic_partitions[tp.topic].add(tp.partition)

        if self.config['api_version'] >= (0, 8, 2):
            request = OffsetFetchRequest[1](
                self.group_id,
                list(topic_partitions.items())
            )
        else:
            request = OffsetFetchRequest[0](
                self.group_id,
                list(topic_partitions.items())
            )

        # send the request with a callback
        future = Future()
        _f = self._client.send(node_id, request)
        _f.add_callback(self._handle_offset_fetch_response, future)
        _f.add_errback(self._failed_request, node_id, request, future)
        return future

    def _handle_offset_fetch_response(self, future, response):
        offsets = {}
        for topic, partitions in response.topics:
            for partition, offset, metadata, error_code in partitions:
                tp = TopicPartition(topic, partition)
                error_type = Errors.for_code(error_code)
                if error_type is not Errors.NoError:
                    error = error_type()
                    log.debug("Group %s failed to fetch offset for partition"
                              " %s: %s", self.group_id, tp, error)
                    if error_type is Errors.GroupLoadInProgressError:
                        # just retry
                        future.failure(error)
                    elif error_type is Errors.NotCoordinatorForGroupError:
                        # re-discover the coordinator and retry
                        self._coordinator_dead()
                        future.failure(error)
                    elif error_type in (Errors.UnknownMemberIdError,
                                        Errors.IllegalGenerationError):
                        future.failure(error)
                    elif error_type is Errors.UnknownTopicOrPartitionError:
                        log.warning("OffsetFetchRequest -- unknown topic %s"
                                    " (have you committed any offsets yet?)",
                                    topic)
                        continue
                    else:
                        log.error("Unknown error fetching offsets for %s: %s",
                                  tp, error)
                        future.failure(error)
                    return
                elif offset >= 0:
                    # record the position with the offset
                    # (-1 indicates no committed offset to fetch)
                    offsets[tp] = OffsetAndMetadata(offset, metadata)
                else:
                    log.debug("Group %s has no committed offset for partition"
                              " %s", self.group_id, tp)
        future.success(offsets)

    def get(self):
        topic_partitions = self._client.cluster.partitions_for_topic(self.topic)
        if not topic_partitions:
            future = self._client.cluster.request_update()
            log.info("No partitions available, performing metadata update.")
            self._client.poll(future=future)
            return {}
        partitions = [TopicPartition(self.topic, partition_id) for partition_id in topic_partitions]
        offsets = self.offsets(partitions, -1)
        committed = self.fetch_committed_offsets(partitions)
        lags = {}
        for tp, offset in six.iteritems(offsets):
            commit_offset = committed[tp] if tp in committed else 0
            numerical = commit_offset if isinstance(commit_offset, int) else commit_offset.offset
            lag = offset - numerical
            pid = tp.partition if isinstance(tp, TopicPartition) else tp
            log.debug("Lag for %s (%s): %s, %s, %s", self.topic, pid, offset, commit_offset, lag)
            lags[pid] = lag
        return lags
Esempio n. 55
0
def test_conn_state_change(mocker, conn):
    cli = KafkaClient()
    sel = mocker.patch.object(cli, '_selector')

    node_id = 0
    conn.state = ConnectionStates.CONNECTING
    cli._conn_state_change(node_id, conn)
    assert node_id in cli._connecting
    sel.register.assert_called_with(conn._sock, selectors.EVENT_WRITE)

    conn.state = ConnectionStates.CONNECTED
    cli._conn_state_change(node_id, conn)
    assert node_id not in cli._connecting
    sel.unregister.assert_called_with(conn._sock)
    sel.register.assert_called_with(conn._sock, selectors.EVENT_READ, conn)

    # Failure to connect should trigger metadata update
    assert cli.cluster._need_update is False
    conn.state = ConnectionStates.DISCONNECTING
    cli._conn_state_change(node_id, conn)
    assert node_id not in cli._connecting
    assert cli.cluster._need_update is True
    sel.unregister.assert_called_with(conn._sock)

    conn.state = ConnectionStates.CONNECTING
    cli._conn_state_change(node_id, conn)
    assert node_id in cli._connecting
    conn.state = ConnectionStates.DISCONNECTING
    cli._conn_state_change(node_id, conn)
    assert node_id not in cli._connecting
Esempio n. 56
0
class KafkaConsumer(six.Iterator):
    """Consume records from a Kafka cluster.

    The consumer will transparently handle the failure of servers in the Kafka
    cluster, and adapt as topic-partitions are created or migrate between
    brokers. It also interacts with the assigned kafka Group Coordinator node
    to allow multiple consumers to load balance consumption of topics (requires
    kafka >= 0.9.0.0).

    Arguments:
        *topics (str): optional list of topics to subscribe to. If not set,
            call subscribe() or assign() before consuming records.

    Keyword Arguments:
        bootstrap_servers: 'host[:port]' string (or list of 'host[:port]'
            strings) that the consumer should contact to bootstrap initial
            cluster metadata. This does not have to be the full node list.
            It just needs to have at least one broker that will respond to a
            Metadata API Request. Default port is 9092. If no servers are
            specified, will default to localhost:9092.
        client_id (str): a name for this client. This string is passed in
            each request to servers and can be used to identify specific
            server-side log entries that correspond to this client. Also
            submitted to GroupCoordinator for logging with respect to
            consumer group administration. Default: 'kafka-python-{version}'
        group_id (str or None): name of the consumer group to join for dynamic
            partition assignment (if enabled), and to use for fetching and
            committing offsets. If None, auto-partition assignment (via
            group coordinator) and offset commits are disabled.
            Default: 'kafka-python-default-group'
        key_deserializer (callable): Any callable that takes a
            raw message key and returns a deserialized key.
        value_deserializer (callable): Any callable that takes a
            raw message value and returns a deserialized value.
        fetch_min_bytes (int): Minimum amount of data the server should
            return for a fetch request, otherwise wait up to
            fetch_max_wait_ms for more data to accumulate. Default: 1.
        fetch_max_wait_ms (int): The maximum amount of time in milliseconds
            the server will block before answering the fetch request if
            there isn't sufficient data to immediately satisfy the
            requirement given by fetch_min_bytes. Default: 500.
        max_partition_fetch_bytes (int): The maximum amount of data
            per-partition the server will return. The maximum total memory
            used for a request = #partitions * max_partition_fetch_bytes.
            This size must be at least as large as the maximum message size
            the server allows or else it is possible for the producer to
            send messages larger than the consumer can fetch. If that
            happens, the consumer can get stuck trying to fetch a large
            message on a certain partition. Default: 1048576.
        request_timeout_ms (int): Client request timeout in milliseconds.
            Default: 40000.
        retry_backoff_ms (int): Milliseconds to backoff when retrying on
            errors. Default: 100.
        reconnect_backoff_ms (int): The amount of time in milliseconds to
            wait before attempting to reconnect to a given host.
            Default: 50.
        max_in_flight_requests_per_connection (int): Requests are pipelined
            to kafka brokers up to this number of maximum requests per
            broker connection. Default: 5.
        auto_offset_reset (str): A policy for resetting offsets on
            OffsetOutOfRange errors: 'earliest' will move to the oldest
            available message, 'latest' will move to the most recent. Any
            other value will raise the exception. Default: 'latest'.
        enable_auto_commit (bool): If true the consumer's offset will be
            periodically committed in the background. Default: True.
        auto_commit_interval_ms (int): milliseconds between automatic
            offset commits, if enable_auto_commit is True. Default: 5000.
        default_offset_commit_callback (callable): called as
            callback(offsets, response) response will be either an Exception
            or a OffsetCommitResponse struct. This callback can be used to
            trigger custom actions when a commit request completes.
        check_crcs (bool): Automatically check the CRC32 of the records
            consumed. This ensures no on-the-wire or on-disk corruption to
            the messages occurred. This check adds some overhead, so it may
            be disabled in cases seeking extreme performance. Default: True
        metadata_max_age_ms (int): The period of time in milliseconds after
            which we force a refresh of metadata even if we haven't seen any
            partition leadership changes to proactively discover any new
            brokers or partitions. Default: 300000
        partition_assignment_strategy (list): List of objects to use to
            distribute partition ownership amongst consumer instances when
            group management is used.
            Default: [RangePartitionAssignor, RoundRobinPartitionAssignor]
        heartbeat_interval_ms (int): The expected time in milliseconds
            between heartbeats to the consumer coordinator when using
            Kafka's group management feature. Heartbeats are used to ensure
            that the consumer's session stays active and to facilitate
            rebalancing when new consumers join or leave the group. The
            value must be set lower than session_timeout_ms, but typically
            should be set no higher than 1/3 of that value. It can be
            adjusted even lower to control the expected time for normal
            rebalances. Default: 3000
        session_timeout_ms (int): The timeout used to detect failures when
            using Kafka's group managementment facilities. Default: 30000
        max_poll_records (int): The maximum number of records returned in a
            single call to poll().
        receive_buffer_bytes (int): The size of the TCP receive buffer
            (SO_RCVBUF) to use when reading data. Default: None (relies on
            system defaults). The java client defaults to 32768.
        send_buffer_bytes (int): The size of the TCP send buffer
            (SO_SNDBUF) to use when sending data. Default: None (relies on
            system defaults). The java client defaults to 131072.
        socket_options (list): List of tuple-arguments to socket.setsockopt
            to apply to broker connection sockets. Default:
            [(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)]
        consumer_timeout_ms (int): number of milliseconds to block during
            message iteration before raising StopIteration (i.e., ending the
            iterator). Default block forever [float('inf')].
        skip_double_compressed_messages (bool): A bug in KafkaProducer <= 1.2.4
            caused some messages to be corrupted via double-compression.
            By default, the fetcher will return these messages as a compressed
            blob of bytes with a single offset, i.e. how the message was
            actually published to the cluster. If you prefer to have the
            fetcher automatically detect corrupt messages and skip them,
            set this option to True. Default: False.
        security_protocol (str): Protocol used to communicate with brokers.
            Valid values are: PLAINTEXT, SSL. Default: PLAINTEXT.
        ssl_context (ssl.SSLContext): pre-configured SSLContext for wrapping
            socket connections. If provided, all other ssl_* configurations
            will be ignored. Default: None.
        ssl_check_hostname (bool): flag to configure whether ssl handshake
            should verify that the certificate matches the brokers hostname.
            default: true.
        ssl_cafile (str): optional filename of ca file to use in certificate
            verification. default: none.
        ssl_certfile (str): optional filename of file in pem format containing
            the client certificate, as well as any ca certificates needed to
            establish the certificate's authenticity. default: none.
        ssl_keyfile (str): optional filename containing the client private key.
            default: none.
        ssl_password (str): optional password to be used when loading the
            certificate chain. default: None.
        ssl_crlfile (str): optional filename containing the CRL to check for
            certificate expiration. By default, no CRL check is done. When
            providing a file, only the leaf certificate will be checked against
            this CRL. The CRL can only be checked with Python 3.4+ or 2.7.9+.
            default: none.
        api_version (tuple): specify which kafka API version to use.
            If set to None, the client will attempt to infer the broker version
            by probing various APIs. Default: None
            Examples:
                (0, 9) enables full group coordination features with automatic
                    partition assignment and rebalancing,
                (0, 8, 2) enables kafka-storage offset commits with manual
                    partition assignment only,
                (0, 8, 1) enables zookeeper-storage offset commits with manual
                    partition assignment only,
                (0, 8, 0) enables basic functionality but requires manual
                    partition assignment and offset management.
            For a full list of supported versions, see KafkaClient.API_VERSIONS
        api_version_auto_timeout_ms (int): number of milliseconds to throw a
            timeout exception from the constructor when checking the broker
            api version. Only applies if api_version set to 'auto'
        metric_reporters (list): A list of classes to use as metrics reporters.
            Implementing the AbstractMetricsReporter interface allows plugging
            in classes that will be notified of new metric creation. Default: []
        metrics_num_samples (int): The number of samples maintained to compute
            metrics. Default: 2
        metrics_sample_window_ms (int): The maximum age in milliseconds of
            samples used to compute metrics. Default: 30000
        selector (selectors.BaseSelector): Provide a specific selector
            implementation to use for I/O multiplexing.
            Default: selectors.DefaultSelector
        exclude_internal_topics (bool): Whether records from internal topics
            (such as offsets) should be exposed to the consumer. If set to True
            the only way to receive records from an internal topic is
            subscribing to it. Requires 0.10+ Default: True
        sasl_mechanism (str): string picking sasl mechanism when security_protocol
            is SASL_PLAINTEXT or SASL_SSL. Currently only PLAIN is supported.
            Default: None
        sasl_plain_username (str): username for sasl PLAIN authentication.
            Default: None
        sasl_plain_password (str): password for sasl PLAIN authentication.
            Default: None

    Note:
        Configuration parameters are described in more detail at
        https://kafka.apache.org/0100/configuration.html#newconsumerconfigs
    """
    DEFAULT_CONFIG = {
        'bootstrap_servers': 'localhost',
        'client_id': 'kafka-python-' + __version__,
        'group_id': 'kafka-python-default-group',
        'key_deserializer': None,
        'value_deserializer': None,
        'fetch_max_wait_ms': 500,
        'fetch_min_bytes': 1,
        'max_partition_fetch_bytes': 1 * 1024 * 1024,
        'request_timeout_ms': 40 * 1000,
        'retry_backoff_ms': 100,
        'reconnect_backoff_ms': 50,
        'max_in_flight_requests_per_connection': 5,
        'auto_offset_reset': 'latest',
        'enable_auto_commit': True,
        'auto_commit_interval_ms': 5000,
        'default_offset_commit_callback': lambda offsets, response: True,
        'check_crcs': True,
        'metadata_max_age_ms': 5 * 60 * 1000,
        'partition_assignment_strategy': (RangePartitionAssignor, RoundRobinPartitionAssignor),
        'heartbeat_interval_ms': 3000,
        'session_timeout_ms': 30000,
        'max_poll_records': sys.maxsize,
        'receive_buffer_bytes': None,
        'send_buffer_bytes': None,
        'socket_options': [(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)],
        'consumer_timeout_ms': float('inf'),
        'skip_double_compressed_messages': False,
        'security_protocol': 'PLAINTEXT',
        'ssl_context': None,
        'ssl_check_hostname': True,
        'ssl_cafile': None,
        'ssl_certfile': None,
        'ssl_keyfile': None,
        'ssl_crlfile': None,
        'ssl_password': None,
        'api_version': None,
        'api_version_auto_timeout_ms': 2000,
        'connections_max_idle_ms': 9 * 60 * 1000, # not implemented yet
        'metric_reporters': [],
        'metrics_num_samples': 2,
        'metrics_sample_window_ms': 30000,
        'metric_group_prefix': 'consumer',
        'selector': selectors.DefaultSelector,
        'exclude_internal_topics': True,
        'sasl_mechanism': None,
        'sasl_plain_username': None,
        'sasl_plain_password': None,
    }

    def __init__(self, *topics, **configs):
        self.config = copy.copy(self.DEFAULT_CONFIG)
        for key in self.config:
            if key in configs:
                self.config[key] = configs.pop(key)

        # Only check for extra config keys in top-level class
        assert not configs, 'Unrecognized configs: %s' % configs

        deprecated = {'smallest': 'earliest', 'largest': 'latest'}
        if self.config['auto_offset_reset'] in deprecated:
            new_config = deprecated[self.config['auto_offset_reset']]
            log.warning('use auto_offset_reset=%s (%s is deprecated)',
                        new_config, self.config['auto_offset_reset'])
            self.config['auto_offset_reset'] = new_config

        metrics_tags = {'client-id': self.config['client_id']}
        metric_config = MetricConfig(samples=self.config['metrics_num_samples'],
                                     time_window_ms=self.config['metrics_sample_window_ms'],
                                     tags=metrics_tags)
        reporters = [reporter() for reporter in self.config['metric_reporters']]
        self._metrics = Metrics(metric_config, reporters)
        # TODO _metrics likely needs to be passed to KafkaClient, etc.

        # api_version was previously a str. accept old format for now
        if isinstance(self.config['api_version'], str):
            str_version = self.config['api_version']
            if str_version == 'auto':
                self.config['api_version'] = None
            else:
                self.config['api_version'] = tuple(map(int, str_version.split('.')))
            log.warning('use api_version=%s [tuple] -- "%s" as str is deprecated',
                        str(self.config['api_version']), str_version)

        self._client = KafkaClient(metrics=self._metrics, **self.config)

        # Get auto-discovered version from client if necessary
        if self.config['api_version'] is None:
            self.config['api_version'] = self._client.config['api_version']

        self._subscription = SubscriptionState(self.config['auto_offset_reset'])
        self._fetcher = Fetcher(
            self._client, self._subscription, self._metrics, **self.config)
        self._coordinator = ConsumerCoordinator(
            self._client, self._subscription, self._metrics,
            assignors=self.config['partition_assignment_strategy'],
            **self.config)
        self._closed = False
        self._iterator = None
        self._consumer_timeout = float('inf')

        if topics:
            self._subscription.subscribe(topics=topics)
            self._client.set_topics(topics)

    def assign(self, partitions):
        """Manually assign a list of TopicPartitions to this consumer.

        Arguments:
            partitions (list of TopicPartition): assignment for this instance.

        Raises:
            IllegalStateError: if consumer has already called subscribe()

        Warning:
            It is not possible to use both manual partition assignment with
            assign() and group assignment with subscribe().

        Note:
            This interface does not support incremental assignment and will
            replace the previous assignment (if there was one).

        Note:
            Manual topic assignment through this method does not use the
            consumer's group management functionality. As such, there will be
            no rebalance operation triggered when group membership or cluster
            and topic metadata change.
        """
        self._subscription.assign_from_user(partitions)
        self._client.set_topics([tp.topic for tp in partitions])

    def assignment(self):
        """Get the TopicPartitions currently assigned to this consumer.

        If partitions were directly assigned using assign(), then this will
        simply return the same partitions that were previously assigned.
        If topics were subscribed using subscribe(), then this will give the
        set of topic partitions currently assigned to the consumer (which may
        be none if the assignment hasn't happened yet, or if the partitions are
        in the process of being reassigned).

        Returns:
            set: {TopicPartition, ...}
        """
        return self._subscription.assigned_partitions()

    def close(self):
        """Close the consumer, waiting indefinitely for any needed cleanup."""
        if self._closed:
            return
        log.debug("Closing the KafkaConsumer.")
        self._closed = True
        self._coordinator.close()
        self._metrics.close()
        self._client.close()
        try:
            self.config['key_deserializer'].close()
        except AttributeError:
            pass
        try:
            self.config['value_deserializer'].close()
        except AttributeError:
            pass
        log.debug("The KafkaConsumer has closed.")

    def commit_async(self, offsets=None, callback=None):
        """Commit offsets to kafka asynchronously, optionally firing callback

        This commits offsets only to Kafka. The offsets committed using this API
        will be used on the first fetch after every rebalance and also on
        startup. As such, if you need to store offsets in anything other than
        Kafka, this API should not be used. To avoid re-processing the last
        message read if a consumer is restarted, the committed offset should be
        the next message your application should consume, i.e.: last_offset + 1.

        This is an asynchronous call and will not block. Any errors encountered
        are either passed to the callback (if provided) or discarded.

        Arguments:
            offsets (dict, optional): {TopicPartition: OffsetAndMetadata} dict
                to commit with the configured group_id. Defaults to current
                consumed offsets for all subscribed partitions.
            callback (callable, optional): called as callback(offsets, response)
                with response as either an Exception or a OffsetCommitResponse
                struct. This callback can be used to trigger custom actions when
                a commit request completes.

        Returns:
            kafka.future.Future
        """
        assert self.config['api_version'] >= (0, 8, 1), 'Requires >= Kafka 0.8.1'
        assert self.config['group_id'] is not None, 'Requires group_id'
        if offsets is None:
            offsets = self._subscription.all_consumed_offsets()
        log.debug("Committing offsets: %s", offsets)
        future = self._coordinator.commit_offsets_async(
            offsets, callback=callback)
        return future

    def commit(self, offsets=None):
        """Commit offsets to kafka, blocking until success or error

        This commits offsets only to Kafka. The offsets committed using this API
        will be used on the first fetch after every rebalance and also on
        startup. As such, if you need to store offsets in anything other than
        Kafka, this API should not be used. To avoid re-processing the last
        message read if a consumer is restarted, the committed offset should be
        the next message your application should consume, i.e.: last_offset + 1.

        Blocks until either the commit succeeds or an unrecoverable error is
        encountered (in which case it is thrown to the caller).

        Currently only supports kafka-topic offset storage (not zookeeper)

        Arguments:
            offsets (dict, optional): {TopicPartition: OffsetAndMetadata} dict
                to commit with the configured group_id. Defaults to current
                consumed offsets for all subscribed partitions.
        """
        assert self.config['api_version'] >= (0, 8, 1), 'Requires >= Kafka 0.8.1'
        assert self.config['group_id'] is not None, 'Requires group_id'
        if offsets is None:
            offsets = self._subscription.all_consumed_offsets()
        self._coordinator.commit_offsets_sync(offsets)

    def committed(self, partition):
        """Get the last committed offset for the given partition

        This offset will be used as the position for the consumer
        in the event of a failure.

        This call may block to do a remote call if the partition in question
        isn't assigned to this consumer or if the consumer hasn't yet
        initialized its cache of committed offsets.

        Arguments:
            partition (TopicPartition): the partition to check

        Returns:
            The last committed offset, or None if there was no prior commit.
        """
        assert self.config['api_version'] >= (0, 8, 1), 'Requires >= Kafka 0.8.1'
        assert self.config['group_id'] is not None, 'Requires group_id'
        if not isinstance(partition, TopicPartition):
            raise TypeError('partition must be a TopicPartition namedtuple')
        if self._subscription.is_assigned(partition):
            committed = self._subscription.assignment[partition].committed
            if committed is None:
                self._coordinator.refresh_committed_offsets_if_needed()
                committed = self._subscription.assignment[partition].committed
        else:
            commit_map = self._coordinator.fetch_committed_offsets([partition])
            if partition in commit_map:
                committed = commit_map[partition].offset
            else:
                committed = None
        return committed

    def topics(self):
        """Get all topics the user is authorized to view.

        Returns:
            set: topics
        """
        cluster = self._client.cluster
        if self._client._metadata_refresh_in_progress and self._client._topics:
            future = cluster.request_update()
            self._client.poll(future=future)
        stash = cluster.need_all_topic_metadata
        cluster.need_all_topic_metadata = True
        future = cluster.request_update()
        self._client.poll(future=future)
        cluster.need_all_topic_metadata = stash
        return cluster.topics()

    def partitions_for_topic(self, topic):
        """Get metadata about the partitions for a given topic.

        Arguments:
            topic (str): topic to check

        Returns:
            set: partition ids
        """
        return self._client.cluster.partitions_for_topic(topic)

    def poll(self, timeout_ms=0, max_records=None):
        """Fetch data from assigned topics / partitions.

        Records are fetched and returned in batches by topic-partition.
        On each poll, consumer will try to use the last consumed offset as the
        starting offset and fetch sequentially. The last consumed offset can be
        manually set through seek(partition, offset) or automatically set as
        the last committed offset for the subscribed list of partitions.

        Incompatible with iterator interface -- use one or the other, not both.

        Arguments:
            timeout_ms (int, optional): milliseconds spent waiting in poll if
                data is not available in the buffer. If 0, returns immediately
                with any records that are available currently in the buffer,
                else returns empty. Must not be negative. Default: 0
            max_records (int, optional): The maximum number of records returned
                in a single call to :meth:`poll`. Default: Inherit value from
                max_poll_records.

        Returns:
            dict: topic to list of records since the last fetch for the
                subscribed list of topics and partitions
        """
        assert timeout_ms >= 0, 'Timeout must not be negative'
        if max_records is None:
            max_records = self.config['max_poll_records']

        # poll for new data until the timeout expires
        start = time.time()
        remaining = timeout_ms
        while True:
            records = self._poll_once(remaining, max_records)
            if records:
                return records

            elapsed_ms = (time.time() - start) * 1000
            remaining = timeout_ms - elapsed_ms

            if remaining <= 0:
                return {}

    def _poll_once(self, timeout_ms, max_records):
        """
        Do one round of polling. In addition to checking for new data, this does
        any needed heart-beating, auto-commits, and offset updates.

        Arguments:
            timeout_ms (int): The maximum time in milliseconds to block

        Returns:
            dict: map of topic to list of records (may be empty)
        """
        if self._use_consumer_group():
            self._coordinator.ensure_coordinator_known()
            self._coordinator.ensure_active_group()

        # 0.8.2 brokers support kafka-backed offset storage via group coordinator
        elif self.config['group_id'] is not None and self.config['api_version'] >= (0, 8, 2):
            self._coordinator.ensure_coordinator_known()

        # fetch positions if we have partitions we're subscribed to that we
        # don't know the offset for
        if not self._subscription.has_all_fetch_positions():
            self._update_fetch_positions(self._subscription.missing_fetch_positions())

        # if data is available already, e.g. from a previous network client
        # poll() call to commit, then just return it immediately
        records, partial = self._fetcher.fetched_records(max_records)
        if records:
            # before returning the fetched records, we can send off the
            # next round of fetches and avoid block waiting for their
            # responses to enable pipelining while the user is handling the
            # fetched records.
            if not partial:
                self._fetcher.send_fetches()
            return records

        # send any new fetches (won't resend pending fetches)
        self._fetcher.send_fetches()

        self._client.poll(timeout_ms=timeout_ms, sleep=True)
        records, _ = self._fetcher.fetched_records(max_records)
        return records

    def position(self, partition):
        """Get the offset of the next record that will be fetched

        Arguments:
            partition (TopicPartition): partition to check

        Returns:
            int: offset
        """
        if not isinstance(partition, TopicPartition):
            raise TypeError('partition must be a TopicPartition namedtuple')
        assert self._subscription.is_assigned(partition), 'Partition is not assigned'
        offset = self._subscription.assignment[partition].position
        if offset is None:
            self._update_fetch_positions([partition])
            offset = self._subscription.assignment[partition].position
        return offset

    def highwater(self, partition):
        """Last known highwater offset for a partition

        A highwater offset is the offset that will be assigned to the next
        message that is produced. It may be useful for calculating lag, by
        comparing with the reported position. Note that both position and
        highwater refer to the *next* offset -- i.e., highwater offset is
        one greater than the newest available message.

        Highwater offsets are returned in FetchResponse messages, so will
        not be available if no FetchRequests have been sent for this partition
        yet.

        Arguments:
            partition (TopicPartition): partition to check

        Returns:
            int or None: offset if available
        """
        if not isinstance(partition, TopicPartition):
            raise TypeError('partition must be a TopicPartition namedtuple')
        assert self._subscription.is_assigned(partition), 'Partition is not assigned'
        return self._subscription.assignment[partition].highwater

    def pause(self, *partitions):
        """Suspend fetching from the requested partitions.

        Future calls to poll() will not return any records from these partitions
        until they have been resumed using resume(). Note that this method does
        not affect partition subscription. In particular, it does not cause a
        group rebalance when automatic assignment is used.

        Arguments:
            *partitions (TopicPartition): partitions to pause
        """
        if not all([isinstance(p, TopicPartition) for p in partitions]):
            raise TypeError('partitions must be TopicPartition namedtuples')
        for partition in partitions:
            log.debug("Pausing partition %s", partition)
            self._subscription.pause(partition)

    def paused(self):
        """Get the partitions that were previously paused by a call to pause().

        Returns:
            set: {partition (TopicPartition), ...}
        """
        return self._subscription.paused_partitions()

    def resume(self, *partitions):
        """Resume fetching from the specified (paused) partitions.

        Arguments:
            *partitions (TopicPartition): partitions to resume
        """
        if not all([isinstance(p, TopicPartition) for p in partitions]):
            raise TypeError('partitions must be TopicPartition namedtuples')
        for partition in partitions:
            log.debug("Resuming partition %s", partition)
            self._subscription.resume(partition)

    def seek(self, partition, offset):
        """Manually specify the fetch offset for a TopicPartition.

        Overrides the fetch offsets that the consumer will use on the next
        poll(). If this API is invoked for the same partition more than once,
        the latest offset will be used on the next poll(). Note that you may
        lose data if this API is arbitrarily used in the middle of consumption,
        to reset the fetch offsets.

        Arguments:
            partition (TopicPartition): partition for seek operation
            offset (int): message offset in partition

        Raises:
            AssertionError: if offset is not an int >= 0; or if partition is not
                currently assigned.
        """
        if not isinstance(partition, TopicPartition):
            raise TypeError('partition must be a TopicPartition namedtuple')
        assert isinstance(offset, int) and offset >= 0, 'Offset must be >= 0'
        assert partition in self._subscription.assigned_partitions(), 'Unassigned partition'
        log.debug("Seeking to offset %s for partition %s", offset, partition)
        self._subscription.assignment[partition].seek(offset)

    def seek_to_beginning(self, *partitions):
        """Seek to the oldest available offset for partitions.

        Arguments:
            *partitions: optionally provide specific TopicPartitions, otherwise
                default to all assigned partitions

        Raises:
            AssertionError: if any partition is not currently assigned, or if
                no partitions are assigned
        """
        if not all([isinstance(p, TopicPartition) for p in partitions]):
            raise TypeError('partitions must be TopicPartition namedtuples')
        if not partitions:
            partitions = self._subscription.assigned_partitions()
            assert partitions, 'No partitions are currently assigned'
        else:
            for p in partitions:
                assert p in self._subscription.assigned_partitions(), 'Unassigned partition'

        for tp in partitions:
            log.debug("Seeking to beginning of partition %s", tp)
            self._subscription.need_offset_reset(tp, OffsetResetStrategy.EARLIEST)

    def seek_to_end(self, *partitions):
        """Seek to the most recent available offset for partitions.

        Arguments:
            *partitions: optionally provide specific TopicPartitions, otherwise
                default to all assigned partitions

        Raises:
            AssertionError: if any partition is not currently assigned, or if
                no partitions are assigned
        """
        if not all([isinstance(p, TopicPartition) for p in partitions]):
            raise TypeError('partitions must be TopicPartition namedtuples')
        if not partitions:
            partitions = self._subscription.assigned_partitions()
            assert partitions, 'No partitions are currently assigned'
        else:
            for p in partitions:
                assert p in self._subscription.assigned_partitions(), 'Unassigned partition'

        for tp in partitions:
            log.debug("Seeking to end of partition %s", tp)
            self._subscription.need_offset_reset(tp, OffsetResetStrategy.LATEST)

    def subscribe(self, topics=(), pattern=None, listener=None):
        """Subscribe to a list of topics, or a topic regex pattern

        Partitions will be dynamically assigned via a group coordinator.
        Topic subscriptions are not incremental: this list will replace the
        current assignment (if there is one).

        This method is incompatible with assign()

        Arguments:
            topics (list): List of topics for subscription.
            pattern (str): Pattern to match available topics. You must provide
                either topics or pattern, but not both.
            listener (ConsumerRebalanceListener): Optionally include listener
                callback, which will be called before and after each rebalance
                operation.

                As part of group management, the consumer will keep track of the
                list of consumers that belong to a particular group and will
                trigger a rebalance operation if one of the following events
                trigger:

                * Number of partitions change for any of the subscribed topics
                * Topic is created or deleted
                * An existing member of the consumer group dies
                * A new member is added to the consumer group

                When any of these events are triggered, the provided listener
                will be invoked first to indicate that the consumer's assignment
                has been revoked, and then again when the new assignment has
                been received. Note that this listener will immediately override
                any listener set in a previous call to subscribe. It is
                guaranteed, however, that the partitions revoked/assigned
                through this interface are from topics subscribed in this call.

        Raises:
            IllegalStateError: if called after previously calling assign()
            AssertionError: if neither topics or pattern is provided
            TypeError: if listener is not a ConsumerRebalanceListener
        """
        # SubscriptionState handles error checking
        self._subscription.subscribe(topics=topics,
                                     pattern=pattern,
                                     listener=listener)

        # regex will need all topic metadata
        if pattern is not None:
            self._client.cluster.need_all_topic_metadata = True
            self._client.set_topics([])
            log.debug("Subscribed to topic pattern: %s", pattern)
        else:
            self._client.cluster.need_all_topic_metadata = False
            self._client.set_topics(self._subscription.group_subscription())
            log.debug("Subscribed to topic(s): %s", topics)

    def subscription(self):
        """Get the current topic subscription.

        Returns:
            set: {topic, ...}
        """
        return self._subscription.subscription

    def unsubscribe(self):
        """Unsubscribe from all topics and clear all assigned partitions."""
        self._subscription.unsubscribe()
        self._coordinator.close()
        self._client.cluster.need_all_topic_metadata = False
        self._client.set_topics([])
        log.debug("Unsubscribed all topics or patterns and assigned partitions")

    def metrics(self, raw=False):
        """Warning: this is an unstable interface.
        It may change in future releases without warning"""
        if raw:
            return self._metrics.metrics

        metrics = {}
        for k, v in self._metrics.metrics.items():
            if k.group not in metrics:
                metrics[k.group] = {}
            if k.name not in metrics[k.group]:
                metrics[k.group][k.name] = {}
            metrics[k.group][k.name] = v.value()
        return metrics

    def _use_consumer_group(self):
        """Return True iff this consumer can/should join a broker-coordinated group."""
        if self.config['api_version'] < (0, 9):
            return False
        elif self.config['group_id'] is None:
            return False
        elif not self._subscription.partitions_auto_assigned():
            return False
        return True

    def _update_fetch_positions(self, partitions):
        """
        Set the fetch position to the committed position (if there is one)
        or reset it using the offset reset policy the user has configured.

        Arguments:
            partitions (List[TopicPartition]): The partitions that need
                updating fetch positions

        Raises:
            NoOffsetForPartitionError: If no offset is stored for a given
                partition and no offset reset policy is defined
        """
        if (self.config['api_version'] >= (0, 8, 1)
            and self.config['group_id'] is not None):

            # refresh commits for all assigned partitions
            self._coordinator.refresh_committed_offsets_if_needed()

        # then do any offset lookups in case some positions are not known
        self._fetcher.update_fetch_positions(partitions)

    def _message_generator(self):
        assert self.assignment() or self.subscription() is not None, 'No topic subscription or manual partition assignment'
        while time.time() < self._consumer_timeout:

            if self._use_consumer_group():
                self._coordinator.ensure_coordinator_known()
                self._coordinator.ensure_active_group()

            # 0.8.2 brokers support kafka-backed offset storage via group coordinator
            elif self.config['group_id'] is not None and self.config['api_version'] >= (0, 8, 2):
                self._coordinator.ensure_coordinator_known()

            # fetch offsets for any subscribed partitions that we arent tracking yet
            if not self._subscription.has_all_fetch_positions():
                partitions = self._subscription.missing_fetch_positions()
                self._update_fetch_positions(partitions)

            poll_ms = 1000 * (self._consumer_timeout - time.time())
            if not self._fetcher.in_flight_fetches():
                poll_ms = 0
            self._client.poll(timeout_ms=poll_ms, sleep=True)

            # We need to make sure we at least keep up with scheduled tasks,
            # like heartbeats, auto-commits, and metadata refreshes
            timeout_at = self._next_timeout()

            # Because the consumer client poll does not sleep unless blocking on
            # network IO, we need to explicitly sleep when we know we are idle
            # because we haven't been assigned any partitions to fetch / consume
            if self._use_consumer_group() and not self.assignment():
                sleep_time = max(timeout_at - time.time(), 0)
                if sleep_time > 0 and not self._client.in_flight_request_count():
                    log.debug('No partitions assigned; sleeping for %s', sleep_time)
                    time.sleep(sleep_time)
                    continue

            # Short-circuit the fetch iterator if we are already timed out
            # to avoid any unintentional interaction with fetcher setup
            if time.time() > timeout_at:
                continue

            for msg in self._fetcher:
                yield msg
                if time.time() > timeout_at:
                    log.debug("internal iterator timeout - breaking for poll")
                    break

            # an else block on a for loop only executes if there was no break
            # so this should only be called on a StopIteration from the fetcher
            # and we assume that it is safe to init_fetches when fetcher is done
            # i.e., there are no more records stored internally
            else:
                self._fetcher.send_fetches()

    def _next_timeout(self):
        timeout = min(self._consumer_timeout,
                      self._client._delayed_tasks.next_at() + time.time(),
                      self._client.cluster.ttl() / 1000.0 + time.time())

        # Although the delayed_tasks timeout above should cover processing
        # HeartbeatRequests, it is still possible that HeartbeatResponses
        # are left unprocessed during a long _fetcher iteration without
        # an intermediate poll(). And because tasks are responsible for
        # rescheduling themselves, an unprocessed response will prevent
        # the next heartbeat from being sent. This check should help
        # avoid that.
        if self._use_consumer_group():
            heartbeat = time.time() + self._coordinator.heartbeat.ttl()
            timeout = min(timeout, heartbeat)
        return timeout

    def __iter__(self):  # pylint: disable=non-iterator-returned
        return self

    def __next__(self):
        if not self._iterator:
            self._iterator = self._message_generator()

        self._set_consumer_timeout()
        try:
            return next(self._iterator)
        except StopIteration:
            self._iterator = None
            raise

    def _set_consumer_timeout(self):
        # consumer_timeout_ms can be used to stop iteration early
        if self.config['consumer_timeout_ms'] >= 0:
            self._consumer_timeout = time.time() + (
                self.config['consumer_timeout_ms'] / 1000.0)

    # old KafkaConsumer methods are deprecated
    def configure(self, **configs):
        raise NotImplementedError(
            'deprecated -- initialize a new consumer')

    def set_topic_partitions(self, *topics):
        raise NotImplementedError(
            'deprecated -- use subscribe() or assign()')

    def fetch_messages(self):
        raise NotImplementedError(
            'deprecated -- use poll() or iterator interface')

    def get_partition_offsets(self, topic, partition,
                              request_time_ms, max_num_offsets):
        raise NotImplementedError(
            'deprecated -- send an OffsetRequest with KafkaClient')

    def offsets(self, group=None):
        raise NotImplementedError('deprecated -- use committed(partition)')

    def task_done(self, message):
        raise NotImplementedError(
            'deprecated -- commit offsets manually if needed')
Esempio n. 57
0
    def __init__(self, bootstrap_servers):

        self.client = KafkaClient(bootstrap_servers=bootstrap_servers)
        self.client.check_version()
Esempio n. 58
0
class KafkaConsumerLag:

    def __init__(self, bootstrap_servers):

        self.client = KafkaClient(bootstrap_servers=bootstrap_servers)
        self.client.check_version()

    def _send(self, broker_id, request, response_type=None):

        f = self.client.send(broker_id, request)
        response = self.client.poll(future=f)

        if response_type:
            if response and len(response) > 0:
                for r in response:
                    if isinstance(r, response_type):
                        return r
        else:
            if response and len(response) > 0:
                return response[0]

        return None

    def check(self, group_topics=None, discovery=None):
        """
        {
            "<group>": {
                "state": <str>,
                "topics": {
                    "<topic>": {
                        "consumer_lag": <int>,
                        "partitions": {
                            "<partition>": {
                                "offset_first": <int>,
                                "offset_consumed": <int>,
                                "offset_last": <int>,
                                "lag": <int>
                            }
                        }
                    }
                }
            }
        }
        :param persist_groups:
        :return: consumer statistics
        """
        cluster = self.client.cluster
        brokers = cluster.brokers()

        # Consumer group ID -> list(topics)
        if group_topics is None:
            group_topics = {}

            if discovery is None:
                discovery = True
        else:
            group_topics = copy.deepcopy(group_topics)

        # Set of consumer group IDs
        consumer_groups = set(group_topics.iterkeys())

        # Set of all known topics
        topics = set(itertools.chain(*group_topics.itervalues()))

        # Consumer group ID -> coordinating broker
        consumer_coordinator = {}

        # Coordinating broker - > list(consumer group IDs)
        coordinator_consumers = {}

        results = {}

        for consumer_group in group_topics.iterkeys():
            results[consumer_group] = {'state': None, 'topics': {}}

        # Ensure connections to all brokers
        for broker in brokers:
            while not self.client.is_ready(broker.nodeId):
                self.client.ready(broker.nodeId)

        # Collect all active consumer groups
        if discovery:
            for broker in brokers:
                response = self._send(broker.nodeId, _ListGroupsRequest(), _ListGroupsResponse)

                if response:
                    for group in response.groups:
                        consumer_groups.add(group[0])

        # Identify which broker is coordinating each consumer group
        for group in consumer_groups:

            response = self._send(next(iter(brokers)).nodeId, _GroupCoordinatorRequest(group), _GroupCoordinatorResponse)

            if response:
                consumer_coordinator[group] = response.coordinator_id

                if response.coordinator_id not in coordinator_consumers:
                    coordinator_consumers[response.coordinator_id] = []

                coordinator_consumers[response.coordinator_id].append(group)

        # Populate consumer groups into dict
        for group in consumer_groups:
            if group not in group_topics:
                group_topics[group] = []

        # Add groups to results dict
        for group, topic_list in group_topics.iteritems():
            results[group] = {'state': None, 'topics': {}}

        # Identify group information and topics read by each consumer group
        for coordinator, consumers in coordinator_consumers.iteritems():

            response = self._send(coordinator, _DescribeGroupsRequest(consumers), _DescribeGroupsResponse)

            for group in response.groups:

                if group[1] in results:
                    results[group[1]]['state'] = group[2]
                    # TODO Also include member data?

                if discovery:
                    members = group[5]
                    for member in members:
                        try:
                            assignment = MemberAssignment.decode(member[4])
                            if assignment:
                                for partition in assignment.partition_assignment:
                                    topic = partition[0]

                                    # Add topic to topic set
                                    topics.add(topic)

                                    # Add topic to group
                                    group_topics[group[1]].append(topic)
                        except:
                            pass

        # Add topics to groups in results dict
        for group, topic_list in group_topics.iteritems():
            for topic in topic_list:
                results[group]['topics'][topic] = {'consumer_lag': 0, 'partitions': {}}

        # For storing the latest offset for all partitions of all topics
        # topic -> partition -> offset
        start_offsets = {}
        end_offsets = {}

        # Identify all the topic partitions that each broker is leader for
        # and request next new offset for each partition
        for broker, partitions in cluster._broker_partitions.iteritems():

            # topic -> List(partition, time, max_offsets)
            request_partitions = {}

            for tp in partitions:
                if tp.topic in topics:
                    if tp.topic not in request_partitions:
                        request_partitions[tp.topic] = []

                    # Time value '-2' is to get the offset for first available message
                    request_partitions[tp.topic].append((tp.partition, -2, 1))

            # List(topic, List(partition, time, max_offsets))
            topic_partitions = []

            for tp in request_partitions.iteritems():
                topic_partitions.append(tp)

            # Request partition start offsets
            response = self._send(broker, _OffsetRequest(-1, topic_partitions), _OffsetResponse)

            if response:
                for offset in response.topics:
                    topic = offset[0]
                    if topic not in start_offsets:
                        start_offsets[topic] = {}

                    for p in offset[1]:
                        start_offsets[topic][p[0]] = p[2][0]

            for tp in topic_partitions:
                for i, ptm in enumerate(tp[1]):
                    # Time value '-1' is to get the offset for next new message
                    tp[1][i] = (ptm[0], -1, 1)

            # Request partition end offsets
            response = self._send(broker, _OffsetRequest(-1, topic_partitions), _OffsetResponse)

            if response:
                for offset in response.topics:
                    topic = offset[0]
                    if topic not in end_offsets:
                        end_offsets[topic] = {}

                    for p in offset[1]:
                        end_offsets[topic][p[0]] = p[2][0]

        # Populate with offset values
        for group, topics in group_topics.iteritems():

            coordinator = consumer_coordinator[group]

            # topic -> list(partition)
            request_partitions = {}

            for topic in topics:
                results[group]['topics'][topic]['consumer_lag'] = 0
                results[group]['topics'][topic]['partitions'] = {}

                if topic in start_offsets:
                    for p in start_offsets[topic]:
                        results[group]['topics'][topic]['partitions'][p] = {
                            'offset_first': start_offsets[topic][p],
                            'offset_last': end_offsets[topic][p],
                            'offset_consumed': 0,
                            'lag' : 0}

                        if topic not in request_partitions:
                            request_partitions[topic] = []
                        request_partitions[topic].append(p)

            # List(topic -> list(partition))
            topic_partitions = []

            for tp in request_partitions.iteritems():
                topic_partitions.append(tp)

            response = self._send(coordinator, _OffsetFetchRequest(group, topic_partitions), _OffsetFetchResponse)

            if response:
                for offset in response.topics:
                    topic = offset[0]
                    offsets = offset[1]

                    if topic not in results[group]['topics']:
                        continue

                    for p_offset in offsets:
                        partition = p_offset[0]
                        offset_consumed = p_offset[1]
                        p_results = results[group]['topics'][topic]['partitions'][partition]

                        if offset_consumed != -1:
                            p_results['offset_consumed'] = offset_consumed
                            p_results['lag'] = p_results['offset_last'] - offset_consumed
                        else:
                            p_results['offset_consumed'] = 0
                            p_results['lag'] = p_results['offset_last'] - p_results['offset_first']

                        results[group]['topics'][topic]['consumer_lag'] += p_results['lag']

        return results

    def close(self):
        if self.client:
            self.client.close()
Esempio n. 59
0
class KafkaConsumer(six.Iterator):
    """Consume records from a Kafka cluster.

    The consumer will transparently handle the failure of servers in the Kafka
    cluster, and adapt as topic-partitions are created or migrate between
    brokers. It also interacts with the assigned kafka Group Coordinator node
    to allow multiple consumers to load balance consumption of topics (requires
    kafka >= 0.9.0.0).

    Arguments:
        *topics (str): optional list of topics to subscribe to. If not set,
            call subscribe() or assign() before consuming records.

    Keyword Arguments:
        bootstrap_servers: 'host[:port]' string (or list of 'host[:port]'
            strings) that the consumer should contact to bootstrap initial
            cluster metadata. This does not have to be the full node list.
            It just needs to have at least one broker that will respond to a
            Metadata API Request. Default port is 9092. If no servers are
            specified, will default to localhost:9092.
        client_id (str): a name for this client. This string is passed in
            each request to servers and can be used to identify specific
            server-side log entries that correspond to this client. Also
            submitted to GroupCoordinator for logging with respect to
            consumer group administration. Default: 'kafka-python-{version}'
        group_id (str): name of the consumer group to join for dynamic
            partition assignment (if enabled), and to use for fetching and
            committing offsets. Default: 'kafka-python-default-group'
        key_deserializer (callable): Any callable that takes a
            raw message key and returns a deserialized key.
        value_deserializer (callable, optional): Any callable that takes a
            raw message value and returns a deserialized value.
        fetch_min_bytes (int): Minimum amount of data the server should
            return for a fetch request, otherwise wait up to
            fetch_max_wait_ms for more data to accumulate. Default: 1024.
        fetch_max_wait_ms (int): The maximum amount of time in milliseconds
            the server will block before answering the fetch request if
            there isn't sufficient data to immediately satisfy the
            requirement given by fetch_min_bytes. Default: 500.
        max_partition_fetch_bytes (int): The maximum amount of data
            per-partition the server will return. The maximum total memory
            used for a request = #partitions * max_partition_fetch_bytes.
            This size must be at least as large as the maximum message size
            the server allows or else it is possible for the producer to
            send messages larger than the consumer can fetch. If that
            happens, the consumer can get stuck trying to fetch a large
            message on a certain partition. Default: 1048576.
        request_timeout_ms (int): Client request timeout in milliseconds.
            Default: 40000.
        retry_backoff_ms (int): Milliseconds to backoff when retrying on
            errors. Default: 100.
        reconnect_backoff_ms (int): The amount of time in milliseconds to
            wait before attempting to reconnect to a given host.
            Default: 50.
        max_in_flight_requests_per_connection (int): Requests are pipelined
            to kafka brokers up to this number of maximum requests per
            broker connection. Default: 5.
        auto_offset_reset (str): A policy for resetting offsets on
            OffsetOutOfRange errors: 'earliest' will move to the oldest
            available message, 'latest' will move to the most recent. Any
            ofther value will raise the exception. Default: 'latest'.
        enable_auto_commit (bool): If true the consumer's offset will be
            periodically committed in the background. Default: True.
        auto_commit_interval_ms (int): milliseconds between automatic
            offset commits, if enable_auto_commit is True. Default: 5000.
        default_offset_commit_callback (callable): called as
            callback(offsets, response) response will be either an Exception
            or a OffsetCommitResponse struct. This callback can be used to
            trigger custom actions when a commit request completes.
        check_crcs (bool): Automatically check the CRC32 of the records
            consumed. This ensures no on-the-wire or on-disk corruption to
            the messages occurred. This check adds some overhead, so it may
            be disabled in cases seeking extreme performance. Default: True
        metadata_max_age_ms (int): The period of time in milliseconds after
            which we force a refresh of metadata even if we haven't seen any
            partition leadership changes to proactively discover any new
            brokers or partitions. Default: 300000
        partition_assignment_strategy (list): List of objects to use to
            distribute partition ownership amongst consumer instances when
            group management is used. Default: [RoundRobinPartitionAssignor]
        heartbeat_interval_ms (int): The expected time in milliseconds
            between heartbeats to the consumer coordinator when using
            Kafka's group management feature. Heartbeats are used to ensure
            that the consumer's session stays active and to facilitate
            rebalancing when new consumers join or leave the group. The
            value must be set lower than session_timeout_ms, but typically
            should be set no higher than 1/3 of that value. It can be
            adjusted even lower to control the expected time for normal
            rebalances. Default: 3000
        session_timeout_ms (int): The timeout used to detect failures when
            using Kafka's group managementment facilities. Default: 30000
        send_buffer_bytes (int): The size of the TCP send buffer
            (SO_SNDBUF) to use when sending data. Default: 131072
        receive_buffer_bytes (int): The size of the TCP receive buffer
            (SO_RCVBUF) to use when reading data. Default: 32768
        consumer_timeout_ms (int): number of millisecond to throw a timeout
            exception to the consumer if no message is available for
            consumption. Default: -1 (dont throw exception)
        api_version (str): specify which kafka API version to use.
            0.9 enables full group coordination features; 0.8.2 enables
            kafka-storage offset commits; 0.8.1 enables zookeeper-storage
            offset commits; 0.8.0 is what is left. If set to 'auto', will
            attempt to infer the broker version by probing various APIs.
            Default: auto

    Note:
        Configuration parameters are described in more detail at
        https://kafka.apache.org/090/configuration.html#newconsumerconfigs
    """

    DEFAULT_CONFIG = {
        "bootstrap_servers": "localhost",
        "client_id": "kafka-python-" + __version__,
        "group_id": "kafka-python-default-group",
        "key_deserializer": None,
        "value_deserializer": None,
        "fetch_max_wait_ms": 500,
        "fetch_min_bytes": 1024,
        "max_partition_fetch_bytes": 1 * 1024 * 1024,
        "request_timeout_ms": 40 * 1000,
        "retry_backoff_ms": 100,
        "reconnect_backoff_ms": 50,
        "max_in_flight_requests_per_connection": 5,
        "auto_offset_reset": "latest",
        "enable_auto_commit": True,
        "auto_commit_interval_ms": 5000,
        "check_crcs": True,
        "metadata_max_age_ms": 5 * 60 * 1000,
        "partition_assignment_strategy": (RoundRobinPartitionAssignor,),
        "heartbeat_interval_ms": 3000,
        "session_timeout_ms": 30000,
        "send_buffer_bytes": 128 * 1024,
        "receive_buffer_bytes": 32 * 1024,
        "consumer_timeout_ms": -1,
        "api_version": "auto",
        "connections_max_idle_ms": 9 * 60 * 1000,  # not implemented yet
        #'metric_reporters': None,
        #'metrics_num_samples': 2,
        #'metrics_sample_window_ms': 30000,
    }

    def __init__(self, *topics, **configs):
        self.config = copy.copy(self.DEFAULT_CONFIG)
        for key in self.config:
            if key in configs:
                self.config[key] = configs.pop(key)

        # Only check for extra config keys in top-level class
        assert not configs, "Unrecognized configs: %s" % configs

        deprecated = {"smallest": "earliest", "largest": "latest"}
        if self.config["auto_offset_reset"] in deprecated:
            new_config = deprecated[self.config["auto_offset_reset"]]
            log.warning("use auto_offset_reset=%s (%s is deprecated)", new_config, self.config["auto_offset_reset"])
            self.config["auto_offset_reset"] = new_config

        self._client = KafkaClient(**self.config)

        # Check Broker Version if not set explicitly
        if self.config["api_version"] == "auto":
            self.config["api_version"] = self._client.check_version()
        assert self.config["api_version"] in ("0.9", "0.8.2", "0.8.1", "0.8.0")

        # Convert api_version config to tuple for easy comparisons
        self.config["api_version"] = tuple(map(int, self.config["api_version"].split(".")))

        self._subscription = SubscriptionState(self.config["auto_offset_reset"])
        self._fetcher = Fetcher(self._client, self._subscription, **self.config)
        self._coordinator = ConsumerCoordinator(
            self._client, self._subscription, assignors=self.config["partition_assignment_strategy"], **self.config
        )
        self._closed = False
        self._iterator = None
        self._consumer_timeout = float("inf")

        # self.metrics = None
        if topics:
            self._subscription.subscribe(topics=topics)
            self._client.set_topics(topics)

    def assign(self, partitions):
        """Manually assign a list of TopicPartitions to this consumer.

        Arguments:
            partitions (list of TopicPartition): assignment for this instance.

        Raises:
            IllegalStateError: if consumer has already called subscribe()

        Warning:
            It is not possible to use both manual partition assignment with
            assign() and group assignment with subscribe().

        Note:
            This interface does not support incremental assignment and will
            replace the previous assignment (if there was one).

        Note:
            Manual topic assignment through this method does not use the
            consumer's group management functionality. As such, there will be
            no rebalance operation triggered when group membership or cluster
            and topic metadata change.
        """
        self._subscription.assign_from_user(partitions)
        self._client.set_topics([tp.topic for tp in partitions])

    def assignment(self):
        """Get the TopicPartitions currently assigned to this consumer.

        If partitions were directly assigned using assign(), then this will
        simply return the same partitions that were previously assigned.
        If topics were subscribed using subscribe(), then this will give the
        set of topic partitions currently assigned to the consumer (which may
        be none if the assignment hasn't happened yet, or if the partitions are
        in the process of being reassigned).

        Returns:
            set: {TopicPartition, ...}
        """
        return self._subscription.assigned_partitions()

    def close(self):
        """Close the consumer, waiting indefinitely for any needed cleanup."""
        if self._closed:
            return
        log.debug("Closing the KafkaConsumer.")
        self._closed = True
        self._coordinator.close()
        # self.metrics.close()
        self._client.close()
        try:
            self.config["key_deserializer"].close()
        except AttributeError:
            pass
        try:
            self.config["value_deserializer"].close()
        except AttributeError:
            pass
        log.debug("The KafkaConsumer has closed.")

    def commit_async(self, offsets=None, callback=None):
        """Commit offsets to kafka asynchronously, optionally firing callback

        This commits offsets only to Kafka. The offsets committed using this API
        will be used on the first fetch after every rebalance and also on
        startup. As such, if you need to store offsets in anything other than
        Kafka, this API should not be used.

        This is an asynchronous call and will not block. Any errors encountered
        are either passed to the callback (if provided) or discarded.

        Arguments:
            offsets (dict, optional): {TopicPartition: OffsetAndMetadata} dict
                to commit with the configured group_id. Defaults to current
                consumed offsets for all subscribed partitions.
            callback (callable, optional): called as callback(offsets, response)
                with response as either an Exception or a OffsetCommitResponse
                struct. This callback can be used to trigger custom actions when
                a commit request completes.

        Returns:
            kafka.future.Future
        """
        assert self.config["api_version"] >= (0, 8, 1)
        if offsets is None:
            offsets = self._subscription.all_consumed_offsets()
        log.debug("Committing offsets: %s", offsets)
        future = self._coordinator.commit_offsets_async(offsets, callback=callback)
        return future

    def commit(self, offsets=None):
        """Commit offsets to kafka, blocking until success or error

        This commits offsets only to Kafka. The offsets committed using this API
        will be used on the first fetch after every rebalance and also on
        startup. As such, if you need to store offsets in anything other than
        Kafka, this API should not be used.

        Blocks until either the commit succeeds or an unrecoverable error is
        encountered (in which case it is thrown to the caller).

        Currently only supports kafka-topic offset storage (not zookeeper)

        Arguments:
            offsets (dict, optional): {TopicPartition: OffsetAndMetadata} dict
                to commit with the configured group_id. Defaults to current
                consumed offsets for all subscribed partitions.
        """
        assert self.config["api_version"] >= (0, 8, 1)
        if offsets is None:
            offsets = self._subscription.all_consumed_offsets()
        self._coordinator.commit_offsets_sync(offsets)

    def committed(self, partition):
        """Get the last committed offset for the given partition

        This offset will be used as the position for the consumer
        in the event of a failure.

        This call may block to do a remote call if the partition in question
        isn't assigned to this consumer or if the consumer hasn't yet
        initialized its cache of committed offsets.

        Arguments:
            partition (TopicPartition): the partition to check

        Returns:
            The last committed offset, or None if there was no prior commit.
        """
        assert self.config["api_version"] >= (0, 8, 1)
        if self._subscription.is_assigned(partition):
            committed = self._subscription.assignment[partition].committed
            if committed is None:
                self._coordinator.refresh_committed_offsets_if_needed()
                committed = self._subscription.assignment[partition].committed
        else:
            commit_map = self._coordinator.fetch_committed_offsets([partition])
            if partition in commit_map:
                committed = commit_map[partition].offset
            else:
                committed = None
        return committed

    def topics(self):
        """Get all topic metadata topics the user is authorized to view.

        [Not Implemented Yet]

        Returns:
            {topic: [partition_info]}
        """
        raise NotImplementedError("TODO")

    def partitions_for_topic(self, topic):
        """Get metadata about the partitions for a given topic.

        Arguments:
            topic (str): topic to check

        Returns:
            set: partition ids
        """
        return self._client.cluster.partitions_for_topic(topic)

    def poll(self, timeout_ms=0):
        """Fetch data from assigned topics / partitions.

        Records are fetched and returned in batches by topic-partition.
        On each poll, consumer will try to use the last consumed offset as the
        starting offset and fetch sequentially. The last consumed offset can be
        manually set through seek(partition, offset) or automatically set as
        the last committed offset for the subscribed list of partitions.

        Incompatible with iterator interface -- use one or the other, not both.

        Arguments:
            timeout_ms (int, optional): milliseconds to spend waiting in poll if
                data is not available. If 0, returns immediately with any
                records that are available now. Must not be negative. Default: 0

        Returns:
            dict: topic to list of records since the last fetch for the
                subscribed list of topics and partitions
        """
        assert timeout_ms >= 0, "Timeout must not be negative"
        assert self._iterator is None, "Incompatible with iterator interface"

        # poll for new data until the timeout expires
        start = time.time()
        remaining = timeout_ms
        while True:
            records = self._poll_once(remaining)
            if records:
                # before returning the fetched records, we can send off the
                # next round of fetches and avoid block waiting for their
                # responses to enable pipelining while the user is handling the
                # fetched records.
                self._fetcher.init_fetches()
                return records

            elapsed_ms = (time.time() - start) * 1000
            remaining = timeout_ms - elapsed_ms

            if remaining <= 0:
                return {}

    def _poll_once(self, timeout_ms):
        """
        Do one round of polling. In addition to checking for new data, this does
        any needed heart-beating, auto-commits, and offset updates.

        Arguments:
            timeout_ms (int): The maximum time in milliseconds to block

        Returns:
            dict: map of topic to list of records (may be empty)
        """
        if self.config["api_version"] >= (0, 8, 2):
            # TODO: Sub-requests should take into account the poll timeout (KAFKA-1894)
            self._coordinator.ensure_coordinator_known()

        if self.config["api_version"] >= (0, 9):
            # ensure we have partitions assigned if we expect to
            if self._subscription.partitions_auto_assigned():
                self._coordinator.ensure_active_group()

        # fetch positions if we have partitions we're subscribed to that we
        # don't know the offset for
        if not self._subscription.has_all_fetch_positions():
            self._update_fetch_positions(self._subscription.missing_fetch_positions())

        # init any new fetches (won't resend pending fetches)
        records = self._fetcher.fetched_records()

        # if data is available already, e.g. from a previous network client
        # poll() call to commit, then just return it immediately
        if records:
            return records

        self._fetcher.init_fetches()
        self._client.poll(timeout_ms)
        return self._fetcher.fetched_records()

    def position(self, partition):
        """Get the offset of the next record that will be fetched

        Arguments:
            partition (TopicPartition): partition to check
        """
        assert self._subscription.is_assigned(partition)

        offset = self._subscription.assignment[partition].position
        if offset is None:
            self._update_fetch_positions(partition)
            offset = self._subscription.assignment[partition].position
        return offset

    def pause(self, *partitions):
        """Suspend fetching from the requested partitions.

        Future calls to poll() will not return any records from these partitions
        until they have been resumed using resume(). Note that this method does
        not affect partition subscription. In particular, it does not cause a
        group rebalance when automatic assignment is used.

        Arguments:
            *partitions (TopicPartition): partitions to pause
        """
        for partition in partitions:
            log.debug("Pausing partition %s", partition)
            self._subscription.pause(partition)

    def resume(self, *partitions):
        """Resume fetching from the specified (paused) partitions.

        Arguments:
            *partitions (TopicPartition): partitions to resume
        """
        for partition in partitions:
            log.debug("Resuming partition %s", partition)
            self._subscription.resume(partition)

    def seek(self, partition, offset):
        """Manually specify the fetch offset for a TopicPartition.

        Overrides the fetch offsets that the consumer will use on the next
        poll(). If this API is invoked for the same partition more than once,
        the latest offset will be used on the next poll(). Note that you may
        lose data if this API is arbitrarily used in the middle of consumption,
        to reset the fetch offsets.

        Arguments:
            partition (TopicPartition): partition for seek operation
            offset (int): message offset in partition
        """
        assert offset >= 0
        log.debug("Seeking to offset %s for partition %s", offset, partition)
        self._subscription.assignment[partition].seek(offset)

    def seek_to_beginning(self, *partitions):
        """Seek to the oldest available offset for partitions.

        Arguments:
            *partitions: optionally provide specific TopicPartitions, otherwise
                default to all assigned partitions
        """
        if not partitions:
            partitions = self._subscription.assigned_partitions()
        for tp in partitions:
            log.debug("Seeking to beginning of partition %s", tp)
            self._subscription.need_offset_reset(tp, OffsetResetStrategy.EARLIEST)

    def seek_to_end(self, *partitions):
        """Seek to the most recent available offset for partitions.

        Arguments:
            *partitions: optionally provide specific TopicPartitions, otherwise
                default to all assigned partitions
        """
        if not partitions:
            partitions = self._subscription.assigned_partitions()
        for tp in partitions:
            log.debug("Seeking to end of partition %s", tp)
            self._subscription.need_offset_reset(tp, OffsetResetStrategy.LATEST)

    def subscribe(self, topics=(), pattern=None, listener=None):
        """Subscribe to a list of topics, or a topic regex pattern

        Partitions will be dynamically assigned via a group coordinator.
        Topic subscriptions are not incremental: this list will replace the
        current assignment (if there is one).

        This method is incompatible with assign()

        Arguments:
            topics (list): List of topics for subscription.
            pattern (str): Pattern to match available topics. You must provide
                either topics or pattern, but not both.
            listener (ConsumerRebalanceListener): Optionally include listener
                callback, which will be called before and after each rebalance
                operation.

                As part of group management, the consumer will keep track of the
                list of consumers that belong to a particular group and will
                trigger a rebalance operation if one of the following events
                trigger:

                * Number of partitions change for any of the subscribed topics
                * Topic is created or deleted
                * An existing member of the consumer group dies
                * A new member is added to the consumer group

                When any of these events are triggered, the provided listener
                will be invoked first to indicate that the consumer's assignment
                has been revoked, and then again when the new assignment has
                been received. Note that this listener will immediately override
                any listener set in a previous call to subscribe. It is
                guaranteed, however, that the partitions revoked/assigned
                through this interface are from topics subscribed in this call.
        """
        if not topics:
            self.unsubscribe()
        else:
            self._subscription.subscribe(topics=topics, pattern=pattern, listener=listener)
            # regex will need all topic metadata
            if pattern is not None:
                self._client.cluster.need_metadata_for_all = True
                log.debug("Subscribed to topic pattern: %s", topics)
            else:
                self._client.set_topics(self._subscription.group_subscription())
                log.debug("Subscribed to topic(s): %s", topics)

    def subscription(self):
        """Get the current topic subscription.

        Returns:
            set: {topic, ...}
        """
        return self._subscription.subscription

    def unsubscribe(self):
        """Unsubscribe from all topics and clear all assigned partitions."""
        self._subscription.unsubscribe()
        self._coordinator.close()
        self._client.cluster.need_metadata_for_all_topics = False
        log.debug("Unsubscribed all topics or patterns and assigned partitions")

    def _update_fetch_positions(self, partitions):
        """
        Set the fetch position to the committed position (if there is one)
        or reset it using the offset reset policy the user has configured.

        Arguments:
            partitions (List[TopicPartition]): The partitions that need
                updating fetch positions

        Raises:
            NoOffsetForPartitionError: If no offset is stored for a given
                partition and no offset reset policy is defined
        """
        if self.config["api_version"] >= (0, 8, 1):
            # refresh commits for all assigned partitions
            self._coordinator.refresh_committed_offsets_if_needed()

        # then do any offset lookups in case some positions are not known
        self._fetcher.update_fetch_positions(partitions)

    def _message_generator(self):
        assert self.assignment() or self.subscription() is not None
        while time.time() < self._consumer_timeout:
            if self.config["api_version"] >= (0, 8, 2):
                self._coordinator.ensure_coordinator_known()

            if self.config["api_version"] >= (0, 9):
                # ensure we have partitions assigned if we expect to
                if self._subscription.partitions_auto_assigned():
                    self._coordinator.ensure_active_group()

            # fetch positions if we have partitions we're subscribed to that we
            # don't know the offset for
            if not self._subscription.has_all_fetch_positions():
                partitions = self._subscription.missing_fetch_positions()
                self._update_fetch_positions(partitions)

            # We need to make sure we at least keep up with scheduled tasks,
            # like heartbeats, auto-commits, and metadata refreshes
            timeout_at = min(
                self._consumer_timeout,
                self._client._delayed_tasks.next_at() + time.time(),
                self._client.cluster.ttl() / 1000.0 + time.time(),
            )

            if self.config["api_version"] >= (0, 9):
                if not self.assignment():
                    sleep_time = time.time() - timeout_at
                    log.debug("No partitions assigned; sleeping for %s", sleep_time)
                    time.sleep(sleep_time)
                    continue

            poll_ms = 1000 * (time.time() - self._consumer_timeout)

            # Dont bother blocking if there are no fetches
            if not self._fetcher.in_flight_fetches():
                poll_ms = 0

            self._client.poll(poll_ms)

            if time.time() > timeout_at:
                continue

            for msg in self._fetcher:
                yield msg
                if time.time() > timeout_at:
                    log.debug("internal iterator timeout - breaking for poll")
                    break

            # an else block on a for loop only executes if there was no break
            # so this should only be called on a StopIteration from the fetcher
            # and we assume that it is safe to init_fetches when fetcher is done
            # i.e., there are no more records stored internally
            else:
                self._fetcher.init_fetches()

    def __iter__(self):  # pylint: disable=non-iterator-returned
        return self

    def __next__(self):
        if not self._iterator:
            self._iterator = self._message_generator()

        self._set_consumer_timeout()
        try:
            return next(self._iterator)
        except StopIteration:
            self._iterator = None
            raise

    def _set_consumer_timeout(self):
        # consumer_timeout_ms can be used to stop iteration early
        if self.config["consumer_timeout_ms"] >= 0:
            self._consumer_timeout = time.time() + (self.config["consumer_timeout_ms"] / 1000.0)

    # old KafkaConsumer methods are deprecated
    def configure(self, **configs):
        raise NotImplementedError("deprecated -- initialize a new consumer")

    def set_topic_partitions(self, *topics):
        raise NotImplementedError("deprecated -- use subscribe() or assign()")

    def fetch_messages(self):
        raise NotImplementedError("deprecated -- use poll() or iterator interface")

    def get_partition_offsets(self, topic, partition, request_time_ms, max_num_offsets):
        raise NotImplementedError("deprecated -- send an OffsetRequest with KafkaClient")

    def offsets(self, group=None):
        raise NotImplementedError("deprecated -- use committed(partition)")

    def task_done(self, message):
        raise NotImplementedError("deprecated -- commit offsets manually if needed")
Esempio n. 60
0
def test_close(mocker, conn):
    cli = KafkaClient()
    mocker.patch.object(cli, '_selector')

    # Unknown node - silent
    cli.close(2)

    # Single node close
    cli._maybe_connect(0)
    assert not conn.close.call_count
    cli.close(0)
    assert conn.close.call_count == 1

    # All node close
    cli._maybe_connect(1)
    cli.close()
    assert conn.close.call_count == 3