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()
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, 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: 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 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': 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, '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'), '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) 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), '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. 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 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): """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['group_id'] is not None: 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(): 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 Returns: int: offset """ 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 availabel message. Highwater offsets are returned in FetchResponse messages, so will not be available if not FetchRequests have been sent for this partition yet. Arguments: partition (TopicPartition): partition to check Returns: int or None: offset if available """ 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 """ 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 Raises: AssertionError: if offset is not an int >= 0; or if partition is not currently assigned. """ 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 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 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 _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.config['group_id'] is not None: 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) poll_ms = 1000 * (self._consumer_timeout - time.time()) if not self._fetcher.in_flight_fetches(): poll_ms = 0 self._client.poll(poll_ms) # 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() if self.config['api_version'] >= (0, 9): if self.config['group_id'] is not None 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 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 _next_timeout(self): return min(self._consumer_timeout, self._client._delayed_tasks.next_at() + time.time(), self._client.cluster.ttl() / 1000.0 + time.time()) 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')
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 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: [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 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. 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. 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) 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 veriication. 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. 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': 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, 'send_buffer_bytes': None, 'receive_buffer_bytes': None, 'consumer_timeout_ms': -1, 'security_protocol': 'PLAINTEXT', 'ssl_context': None, 'ssl_check_hostname': True, 'ssl_cafile': None, 'ssl_certfile': None, 'ssl_keyfile': None, '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'), '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) 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): """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 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._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()) # 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=timeout_ms, sleep=True) 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 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 not 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 _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.init_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')
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")
class KafkaAdminClient(object): """A class for administering the Kafka cluster. Warning: This is an unstable interface that was recently added and is subject to change without warning. In particular, many methods currently return raw protocol tuples. In future releases, we plan to make these into nicer, more pythonic objects. Unfortunately, this will likely break those interfaces. The KafkaAdminClient class will negotiate for the latest version of each message protocol format supported by both the kafka-python client library and the Kafka broker. Usage of optional fields from protocol versions that are not supported by the broker will result in IncompatibleBrokerVersion exceptions. Use of this class requires a minimum broker version >= 0.10.0.0. 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}' reconnect_backoff_ms (int): The amount of time in milliseconds to wait before attempting to reconnect to a given host. Default: 50. reconnect_backoff_max_ms (int): The maximum amount of time in milliseconds to wait when reconnecting to a broker that has repeatedly failed to connect. If provided, the backoff per host will increase exponentially for each consecutive connection failure, up to this maximum. To avoid connection storms, a randomization factor of 0.2 will be applied to the backoff resulting in a random range between 20% below and 20% above the computed value. Default: 1000. request_timeout_ms (int): Client request timeout in milliseconds. Default: 30000. connections_max_idle_ms: Close idle connections after the number of milliseconds specified by this config. The broker closes idle connections after connections.max.idle.ms, so this avoids hitting unexpected socket disconnected errors on the client. Default: 540000 retry_backoff_ms (int): Milliseconds to backoff when retrying on errors. Default: 100. 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. receive_buffer_bytes (int): The size of the TCP receive buffer (SO_RCVBUF) to use when reading data. Default: None (relies on system defaults). 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). 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)] 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 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 broker's hostname. Default: True. ssl_cafile (str): Optional filename of CA file to use in certificate veriication. 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, KafkaClient will attempt to infer the broker version by probing various APIs. Example: (0, 10, 2). Default: None 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 is None selector (selectors.BaseSelector): Provide a specific selector implementation to use for I/O multiplexing. Default: selectors.DefaultSelector metrics (kafka.metrics.Metrics): Optionally provide a metrics instance for capturing network IO stats. Default: None. metric_group_prefix (str): Prefix for metric names. Default: '' sasl_mechanism (str): Authentication mechanism when security_protocol is configured for SASL_PLAINTEXT or SASL_SSL. Valid values are: PLAIN, GSSAPI, OAUTHBEARER. sasl_plain_username (str): username for sasl PLAIN authentication. Required if sasl_mechanism is PLAIN. sasl_plain_password (str): password for sasl PLAIN authentication. Required if sasl_mechanism is PLAIN. sasl_kerberos_service_name (str): Service name to include in GSSAPI sasl mechanism handshake. Default: 'kafka' sasl_oauth_token_provider (AbstractTokenProvider): OAuthBearer token provider instance. (See kafka.oauth.abstract). Default: None """ DEFAULT_CONFIG = { # client configs 'bootstrap_servers': 'localhost', 'client_id': 'kafka-python-' + __version__, 'request_timeout_ms': 30000, 'connections_max_idle_ms': 9 * 60 * 1000, 'reconnect_backoff_ms': 50, 'reconnect_backoff_max_ms': 1000, 'max_in_flight_requests_per_connection': 5, 'receive_buffer_bytes': None, 'send_buffer_bytes': None, 'socket_options': [(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)], 'sock_chunk_bytes': 4096, # undocumented experimental option 'sock_chunk_buffer_count': 1000, # undocumented experimental option 'retry_backoff_ms': 100, 'metadata_max_age_ms': 300000, 'security_protocol': 'PLAINTEXT', 'ssl_context': None, 'ssl_check_hostname': True, 'ssl_cafile': None, 'ssl_certfile': None, 'ssl_keyfile': None, 'ssl_password': None, 'ssl_crlfile': None, 'api_version': None, 'api_version_auto_timeout_ms': 2000, 'selector': selectors.DefaultSelector, 'sasl_mechanism': None, 'sasl_plain_username': None, 'sasl_plain_password': None, 'sasl_kerberos_service_name': 'kafka', 'sasl_oauth_token_provider': None, # metrics configs 'metric_reporters': [], 'metrics_num_samples': 2, 'metrics_sample_window_ms': 30000, } 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.") def close(self): """Close the KafkaAdminClient connection to the Kafka broker.""" if not hasattr(self, '_closed') or self._closed: log.info("KafkaAdminClient already closed.") return self._metrics.close() self._client.close() self._closed = True log.debug("KafkaAdminClient is now closed.") def _matching_api_version(self, operation): """Find the latest version of the protocol operation supported by both this library and the broker. This resolves to the lesser of either the latest api version this library supports, or the max version supported by the broker. :param operation: A list of protocol operation versions from kafka.protocol. :return: The max matching version number between client and broker. """ version = min( len(operation) - 1, self._client.get_api_versions()[operation[0].API_KEY][1]) if version < self._client.get_api_versions()[operation[0].API_KEY][0]: # max library version is less than min broker version. Currently, # no Kafka versions specify a min msg version. Maybe in the future? raise IncompatibleBrokerVersion( "No version of the '{}' Kafka protocol is supported by both the client and broker." .format(operation.__name__)) return version def _validate_timeout(self, timeout_ms): """Validate the timeout is set or use the configuration default. :param timeout_ms: The timeout provided by api call, in milliseconds. :return: The timeout to use for the operation. """ return timeout_ms or self.config['request_timeout_ms'] def _refresh_controller_id(self): """Determine the Kafka cluster controller.""" version = self._matching_api_version(MetadataRequest) if 1 <= version <= 6: request = MetadataRequest[version]() future = self._send_request_to_node( self._client.least_loaded_node(), request) self._wait_for_futures([future]) response = future.value controller_id = response.controller_id # verify the controller is new enough to support our requests controller_version = self._client.check_version(controller_id) if controller_version < (0, 10, 0): raise IncompatibleBrokerVersion( "The controller appears to be running Kafka {}. KafkaAdminClient requires brokers >= 0.10.0.0." .format(controller_version)) self._controller_id = controller_id else: raise UnrecognizedBrokerVersion( "Kafka Admin interface cannot determine the controller using MetadataRequest_v{}." .format(version)) def _find_group_coordinator_id(self, group_id): """Find the broker node_id of the coordinator of the given group. Sends a FindCoordinatorRequest message to the cluster. Will block until the FindCoordinatorResponse is received. Any errors are immediately raised. :param group_id: The consumer group ID. This is typically the group name as a string. :return: The node_id of the broker that is the coordinator. """ # Note: Java may change how this is implemented in KAFKA-6791. # # TODO add support for dynamically picking version of # GroupCoordinatorRequest which was renamed to FindCoordinatorRequest. # When I experimented with this, GroupCoordinatorResponse_v1 didn't # match GroupCoordinatorResponse_v0 and I couldn't figure out why. gc_request = GroupCoordinatorRequest[0](group_id) future = self._send_request_to_node(self._client.least_loaded_node(), gc_request) self._wait_for_futures([future]) gc_response = future.value # use the extra error checking in add_group_coordinator() rather than # immediately returning the group coordinator. success = self._client.cluster.add_group_coordinator( group_id, gc_response) if not success: error_type = Errors.for_code(gc_response.error_code) assert error_type is not Errors.NoError # Note: When error_type.retriable, Java will retry... see # KafkaAdminClient's handleFindCoordinatorError method raise error_type( "Could not identify group coordinator for group_id '{}' from response '{}'." .format(group_id, gc_response)) group_coordinator = self._client.cluster.coordinator_for_group( group_id) # will be None if the coordinator was never populated, which should never happen here assert group_coordinator is not None # will be -1 if add_group_coordinator() failed... but by this point the # error should have been raised. assert group_coordinator != -1 return group_coordinator def _send_request_to_node(self, node_id, request): """Send a Kafka protocol message to a specific broker. Returns a future that may be polled for status and results. :param node_id: The broker id to which to send the message. :param request: The message to send. :return: A future object that may be polled for status and results. :exception: The exception if the message could not be sent. """ while not self._client.ready(node_id): # poll until the connection to broker is ready, otherwise send() # will fail with NodeNotReadyError self._client.poll() return self._client.send(node_id, request) def _send_request_to_controller(self, request): """Send a Kafka protocol message to the cluster controller. Will block until the message result is received. :param request: The message to send. :return: The Kafka protocol response for the message. """ tries = 2 # in case our cached self._controller_id is outdated while tries: tries -= 1 future = self._send_request_to_node(self._controller_id, request) self._wait_for_futures([future]) response = future.value # In Java, the error fieldname is inconsistent: # - CreateTopicsResponse / CreatePartitionsResponse uses topic_errors # - DeleteTopicsResponse uses topic_error_codes # So this is a little brittle in that it assumes all responses have # one of these attributes and that they always unpack into # (topic, error_code) tuples. topic_error_tuples = (response.topic_errors if hasattr( response, 'topic_errors') else response.topic_error_codes) # Also small py2/py3 compatibility -- py3 can ignore extra values # during unpack via: for x, y, *rest in list_of_values. py2 cannot. # So for now we have to map across the list and explicitly drop any # extra values (usually the error_message) for topic, error_code in map(lambda e: e[:2], topic_error_tuples): error_type = Errors.for_code(error_code) if tries and error_type is NotControllerError: # No need to inspect the rest of the errors for # non-retriable errors because NotControllerError should # either be thrown for all errors or no errors. self._refresh_controller_id() break elif error_type is not Errors.NoError: raise error_type( "Request '{}' failed with response '{}'.".format( request, response)) else: return response raise RuntimeError( "This should never happen, please file a bug with full stacktrace if encountered" ) @staticmethod def _convert_new_topic_request(new_topic): return ( new_topic.name, new_topic.num_partitions, new_topic.replication_factor, [ (partition_id, replicas) for partition_id, replicas in new_topic.replica_assignments.items() ], [(config_key, config_value) for config_key, config_value in new_topic.topic_configs.items()]) def create_topics(self, new_topics, timeout_ms=None, validate_only=False): """Create new topics in the cluster. :param new_topics: A list of NewTopic objects. :param timeout_ms: Milliseconds to wait for new topics to be created before the broker returns. :param validate_only: If True, don't actually create new topics. Not supported by all versions. Default: False :return: Appropriate version of CreateTopicResponse class. """ version = self._matching_api_version(CreateTopicsRequest) timeout_ms = self._validate_timeout(timeout_ms) if version == 0: if validate_only: raise IncompatibleBrokerVersion( "validate_only requires CreateTopicsRequest >= v1, which is not supported by Kafka {}." .format(self.config['api_version'])) request = CreateTopicsRequest[version](create_topic_requests=[ self._convert_new_topic_request(new_topic) for new_topic in new_topics ], timeout=timeout_ms) elif version <= 2: request = CreateTopicsRequest[version](create_topic_requests=[ self._convert_new_topic_request(new_topic) for new_topic in new_topics ], timeout=timeout_ms, validate_only=validate_only) else: raise NotImplementedError( "Support for CreateTopics v{} has not yet been added to KafkaAdminClient." .format(version)) # TODO convert structs to a more pythonic interface # TODO raise exceptions if errors return self._send_request_to_controller(request) def delete_topics(self, topics, timeout_ms=None): """Delete topics from the cluster. :param topics: A list of topic name strings. :param timeout_ms: Milliseconds to wait for topics to be deleted before the broker returns. :return: Appropriate version of DeleteTopicsResponse class. """ version = self._matching_api_version(DeleteTopicsRequest) timeout_ms = self._validate_timeout(timeout_ms) if version <= 1: request = DeleteTopicsRequest[version](topics=topics, timeout=timeout_ms) response = self._send_request_to_controller(request) else: raise NotImplementedError( "Support for DeleteTopics v{} has not yet been added to KafkaAdminClient." .format(version)) return response # list topics functionality is in ClusterMetadata # Note: if implemented here, send the request to the least_loaded_node() # describe topics functionality is in ClusterMetadata # Note: if implemented here, send the request to the controller # describe cluster functionality is in ClusterMetadata # Note: if implemented here, send the request to the least_loaded_node() # describe_acls protocol not yet implemented # Note: send the request to the least_loaded_node() # create_acls protocol not yet implemented # Note: send the request to the least_loaded_node() # delete_acls protocol not yet implemented # Note: send the request to the least_loaded_node() @staticmethod def _convert_describe_config_resource_request(config_resource): return (config_resource.resource_type, config_resource.name, [ config_key for config_key, config_value in config_resource.configs.items() ] if config_resource.configs else None) def describe_configs(self, config_resources, include_synonyms=False): """Fetch configuration parameters for one or more Kafka resources. :param config_resources: An list of ConfigResource objects. Any keys in ConfigResource.configs dict will be used to filter the result. Setting the configs dict to None will get all values. An empty dict will get zero values (as per Kafka protocol). :param include_synonyms: If True, return synonyms in response. Not supported by all versions. Default: False. :return: Appropriate version of DescribeConfigsResponse class. """ version = self._matching_api_version(DescribeConfigsRequest) if version == 0: if include_synonyms: raise IncompatibleBrokerVersion( "include_synonyms requires DescribeConfigsRequest >= v1, which is not supported by Kafka {}." .format(self.config['api_version'])) request = DescribeConfigsRequest[version](resources=[ self._convert_describe_config_resource_request(config_resource) for config_resource in config_resources ]) elif version == 1: request = DescribeConfigsRequest[version]( resources=[ self._convert_describe_config_resource_request( config_resource) for config_resource in config_resources ], include_synonyms=include_synonyms) else: raise NotImplementedError( "Support for DescribeConfigs v{} has not yet been added to KafkaAdminClient." .format(version)) future = self._send_request_to_node(self._client.least_loaded_node(), request) self._wait_for_futures([future]) return future.value @staticmethod def _convert_alter_config_resource_request(config_resource): return (config_resource.resource_type, config_resource.name, [ (config_key, config_value) for config_key, config_value in config_resource.configs.items() ]) def alter_configs(self, config_resources): """Alter configuration parameters of one or more Kafka resources. Warning: This is currently broken for BROKER resources because those must be sent to that specific broker, versus this always picks the least-loaded node. See the comment in the source code for details. We would happily accept a PR fixing this. :param config_resources: A list of ConfigResource objects. :return: Appropriate version of AlterConfigsResponse class. """ version = self._matching_api_version(AlterConfigsRequest) if version == 0: request = AlterConfigsRequest[version](resources=[ self._convert_alter_config_resource_request(config_resource) for config_resource in config_resources ]) else: raise NotImplementedError( "Support for AlterConfigs v{} has not yet been added to KafkaAdminClient." .format(version)) # TODO the Java client has the note: # // We must make a separate AlterConfigs request for every BROKER resource we want to alter # // and send the request to that specific broker. Other resources are grouped together into # // a single request that may be sent to any broker. # # So this is currently broken as it always sends to the least_loaded_node() future = self._send_request_to_node(self._client.least_loaded_node(), request) self._wait_for_futures([future]) return future.value # alter replica logs dir protocol not yet implemented # Note: have to lookup the broker with the replica assignment and send the request to that broker # describe log dirs protocol not yet implemented # Note: have to lookup the broker with the replica assignment and send the request to that broker @staticmethod def _convert_create_partitions_request(topic_name, new_partitions): return (topic_name, (new_partitions.total_count, new_partitions.new_assignments)) def create_partitions(self, topic_partitions, timeout_ms=None, validate_only=False): """Create additional partitions for an existing topic. :param topic_partitions: A map of topic name strings to NewPartition objects. :param timeout_ms: Milliseconds to wait for new partitions to be created before the broker returns. :param validate_only: If True, don't actually create new partitions. Default: False :return: Appropriate version of CreatePartitionsResponse class. """ version = self._matching_api_version(CreatePartitionsRequest) timeout_ms = self._validate_timeout(timeout_ms) if version == 0: request = CreatePartitionsRequest[version]( topic_partitions=[ self._convert_create_partitions_request( topic_name, new_partitions) for topic_name, new_partitions in topic_partitions.items() ], timeout=timeout_ms, validate_only=validate_only) else: raise NotImplementedError( "Support for CreatePartitions v{} has not yet been added to KafkaAdminClient." .format(version)) return self._send_request_to_controller(request) # delete records protocol not yet implemented # Note: send the request to the partition leaders # create delegation token protocol not yet implemented # Note: send the request to the least_loaded_node() # renew delegation token protocol not yet implemented # Note: send the request to the least_loaded_node() # expire delegation_token protocol not yet implemented # Note: send the request to the least_loaded_node() # describe delegation_token protocol not yet implemented # Note: send the request to the least_loaded_node() def describe_consumer_groups(self, group_ids, group_coordinator_id=None): """Describe a set of consumer groups. Any errors are immediately raised. :param group_ids: A list of consumer group IDs. These are typically the group names as strings. :param group_coordinator_id: The node_id of the groups' coordinator broker. If set to None, it will query the cluster for each group to find that group's coordinator. Explicitly specifying this can be useful for avoiding extra network round trips if you already know the group coordinator. This is only useful when all the group_ids have the same coordinator, otherwise it will error. Default: None. :return: A list of group descriptions. For now the group descriptions are the raw results from the DescribeGroupsResponse. Long-term, we plan to change this to return namedtuples as well as decoding the partition assignments. """ group_descriptions = [] futures = [] version = self._matching_api_version(DescribeGroupsRequest) for group_id in group_ids: if group_coordinator_id is not None: this_groups_coordinator_id = group_coordinator_id else: this_groups_coordinator_id = self._find_group_coordinator_id( group_id) if version <= 1: # Note: KAFKA-6788 A potential optimization is to group the # request per coordinator and send one request with a list of # all consumer groups. Java still hasn't implemented this # because the error checking is hard to get right when some # groups error and others don't. request = DescribeGroupsRequest[version](groups=(group_id, )) futures.append( self._send_request_to_node(this_groups_coordinator_id, request)) else: raise NotImplementedError( "Support for DescribeGroups v{} has not yet been added to KafkaAdminClient." .format(version)) self._wait_for_futures(futures) for future in futures: response = future.value assert len(response.groups) == 1 # TODO need to implement converting the response tuple into # a more accessible interface like a namedtuple and then stop # hardcoding tuple indices here. Several Java examples, # including KafkaAdminClient.java group_description = response.groups[0] error_code = group_description[0] error_type = Errors.for_code(error_code) # Java has the note: KAFKA-6789, we can retry based on the error code if error_type is not Errors.NoError: raise error_type( "Request '{}' failed with response '{}'.".format( request, response)) # TODO Java checks the group protocol type, and if consumer # (ConsumerProtocol.PROTOCOL_TYPE) or empty string, it decodes # the members' partition assignments... that hasn't yet been # implemented here so just return the raw struct results group_descriptions.append(group_description) return group_descriptions def list_consumer_groups(self, broker_ids=None): """List all consumer groups known to the cluster. This returns a list of Consumer Group tuples. The tuples are composed of the consumer group name and the consumer group protocol type. Only consumer groups that store their offsets in Kafka are returned. The protocol type will be an empty string for groups created using Kafka < 0.9 APIs because, although they store their offsets in Kafka, they don't use Kafka for group coordination. For groups created using Kafka >= 0.9, the protocol type will typically be "consumer". As soon as any error is encountered, it is immediately raised. :param broker_ids: A list of broker node_ids to query for consumer groups. If set to None, will query all brokers in the cluster. Explicitly specifying broker(s) can be useful for determining which consumer groups are coordinated by those broker(s). Default: None :return list: List of tuples of Consumer Groups. :exception GroupCoordinatorNotAvailableError: The coordinator is not available, so cannot process requests. :exception GroupLoadInProgressError: The coordinator is loading and hence can't process requests. """ # While we return a list, internally use a set to prevent duplicates # because if a group coordinator fails after being queried, and its # consumer groups move to new brokers that haven't yet been queried, # then the same group could be returned by multiple brokers. consumer_groups = set() futures = [] if broker_ids is None: broker_ids = [ broker.nodeId for broker in self._client.cluster.brokers() ] version = self._matching_api_version(ListGroupsRequest) if version <= 2: request = ListGroupsRequest[version]() for broker_id in broker_ids: futures.append(self._send_request_to_node(broker_id, request)) self._wait_for_futures(futures) for future in futures: response = future.value error_type = Errors.for_code(response.error_code) if error_type is not Errors.NoError: raise error_type( "Request '{}' failed with response '{}'.".format( request, response)) consumer_groups.update(response.groups) else: raise NotImplementedError( "Support for ListGroups v{} has not yet been added to KafkaAdminClient." .format(version)) return list(consumer_groups) def list_consumer_group_offsets(self, group_id, group_coordinator_id=None, partitions=None): """Fetch Consumer Group Offsets. Note: This does not verify that the group_id or partitions actually exist in the cluster. As soon as any error is encountered, it is immediately raised. :param group_id: The consumer group id name for which to fetch offsets. :param group_coordinator_id: The node_id of the group's coordinator broker. If set to None, will query the cluster to find the group coordinator. Explicitly specifying this can be useful to prevent that extra network round trip if you already know the group coordinator. Default: None. :param partitions: A list of TopicPartitions for which to fetch offsets. On brokers >= 0.10.2, this can be set to None to fetch all known offsets for the consumer group. Default: None. :return dictionary: A dictionary with TopicPartition keys and OffsetAndMetada values. Partitions that are not specified and for which the group_id does not have a recorded offset are omitted. An offset value of `-1` indicates the group_id has no offset for that TopicPartition. A `-1` can only happen for partitions that are explicitly specified. """ group_offsets_listing = {} if group_coordinator_id is None: group_coordinator_id = self._find_group_coordinator_id(group_id) version = self._matching_api_version(OffsetFetchRequest) if version <= 3: if partitions is None: if version <= 1: raise ValueError( """OffsetFetchRequest_v{} requires specifying the partitions for which to fetch offsets. Omitting the partitions is only supported on brokers >= 0.10.2. For details, see KIP-88.""".format(version)) topics_partitions = None else: # transform from [TopicPartition("t1", 1), TopicPartition("t1", 2)] to [("t1", [1, 2])] topics_partitions_dict = defaultdict(set) for topic, partition in partitions: topics_partitions_dict[topic].add(partition) topics_partitions = list(six.iteritems(topics_partitions_dict)) request = OffsetFetchRequest[version](group_id, topics_partitions) future = self._send_request_to_node(group_coordinator_id, request) self._wait_for_futures([future]) response = future.value if version > 1: # OffsetFetchResponse_v1 lacks a top-level error_code error_type = Errors.for_code(response.error_code) if error_type is not Errors.NoError: # optionally we could retry if error_type.retriable raise error_type( "Request '{}' failed with response '{}'.".format( request, response)) # transform response into a dictionary with TopicPartition keys and # OffsetAndMetada values--this is what the Java AdminClient returns for topic, partitions in response.topics: for partition, offset, metadata, error_code in partitions: error_type = Errors.for_code(error_code) if error_type is not Errors.NoError: raise error_type( "Unable to fetch offsets for group_id {}, topic {}, partition {}" .format(group_id, topic, partition)) group_offsets_listing[TopicPartition( topic, partition)] = OffsetAndMetadata(offset, metadata) else: raise NotImplementedError( "Support for OffsetFetch v{} has not yet been added to KafkaAdminClient." .format(version)) return group_offsets_listing # delete groups protocol not yet implemented # Note: send the request to the group's coordinator. def _wait_for_futures(self, futures): while not all(future.succeeded() for future in futures): for future in futures: self._client.poll(future=future) if future.failed(): raise future.exception # pylint: disable-msg=raising-bad-type
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()
class KafkaAdmin(object): """An class for administering the kafka cluster. The KafkaAdmin class will negotiate for the latest version of each message protocol format supported by both the kafka-python client library and the kafka broker. Usage of optional fields from protocol versions that are not supported by the broker will result in UnsupportedVersionError exceptions. Use of this class requires a minimum broker version >= 0.10.0.0. 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}' reconnect_backoff_ms (int): The amount of time in milliseconds to wait before attempting to reconnect to a given host. Default: 50. reconnect_backoff_max_ms (int): The maximum amount of time in milliseconds to wait when reconnecting to a broker that has repeatedly failed to connect. If provided, the backoff per host will increase exponentially for each consecutive connection failure, up to this maximum. To avoid connection storms, a randomization factor of 0.2 will be applied to the backoff resulting in a random range between 20% below and 20% above the computed value. Default: 1000. request_timeout_ms (int): Client request timeout in milliseconds. Default: 30000. connections_max_idle_ms: Close idle connections after the number of milliseconds specified by this config. The broker closes idle connections after connections.max.idle.ms, so this avoids hitting unexpected socket disconnected errors on the client. Default: 540000 retry_backoff_ms (int): Milliseconds to backoff when retrying on errors. Default: 100. 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. receive_buffer_bytes (int): The size of the TCP receive buffer (SO_RCVBUF) to use when reading data. Default: None (relies on system defaults). 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). 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)] 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 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 broker's hostname. Default: True. ssl_cafile (str): Optional filename of CA file to use in certificate veriication. 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, KafkaClient will attempt to infer the broker version by probing various APIs. Example: (0, 10, 2). Default: None 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 is None selector (selectors.BaseSelector): Provide a specific selector implementation to use for I/O multiplexing. Default: selectors.DefaultSelector metrics (kafka.metrics.Metrics): Optionally provide a metrics instance for capturing network IO stats. Default: None. metric_group_prefix (str): Prefix for metric names. Default: '' 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 sasl_kerberos_service_name (str): Service name to include in GSSAPI sasl mechanism handshake. Default: 'kafka' """ DEFAULT_CONFIG = { # client configs 'bootstrap_servers': 'localhost', 'client_id': 'kafka-python-' + __version__, 'request_timeout_ms': 30000, 'connections_max_idle_ms': 9 * 60 * 1000, 'reconnect_backoff_ms': 50, 'reconnect_backoff_max_ms': 1000, 'max_in_flight_requests_per_connection': 5, 'receive_buffer_bytes': None, 'send_buffer_bytes': None, 'socket_options': [(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)], 'sock_chunk_bytes': 4096, # undocumented experimental option 'sock_chunk_buffer_count': 1000, # undocumented experimental option 'retry_backoff_ms': 100, 'metadata_max_age_ms': 300000, 'security_protocol': 'PLAINTEXT', 'ssl_context': None, 'ssl_check_hostname': True, 'ssl_cafile': None, 'ssl_certfile': None, 'ssl_keyfile': None, 'ssl_password': None, 'ssl_crlfile': None, 'api_version': None, 'api_version_auto_timeout_ms': 2000, 'selector': selectors.DefaultSelector, 'sasl_mechanism': None, 'sasl_plain_username': None, 'sasl_plain_password': None, 'sasl_kerberos_service_name': 'kafka', # metrics configs 'metric_reporters': [], 'metrics_num_samples': 2, 'metrics_sample_window_ms': 30000, } 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') def close(self): """Close the administration connection to the kafka broker""" if not hasattr(self, '_closed') or self._closed: log.info('Kafka administration interface already closed') return self._metrics.close() self._client.close() self._closed = True log.debug('Kafka administartion interface has closed') def _matching_api_version(self, operation): """Find matching api version, the lesser of either the latest api version the library supports, or the max version supported by the broker :param operation: An operation array from kafka.protocol :return: The max matching version number between client and broker """ version = min( len(operation) - 1, self._client.get_api_versions()[operation[0].API_KEY][1]) if version < self._client.get_api_versions()[operation[0].API_KEY][0]: # max library version is less than min broker version. Not sure any brokers # actually set a min version greater than 0 right now, tho. But maybe in the future? raise UnsupportedVersionError( "Could not find matching protocol version for {}".format( operation.__name__)) return version def _validate_timeout(self, timeout_ms): """Validate the timeout is set or use the configuration default :param timeout_ms: The timeout provided by api call, in milliseconds :return: The timeout to use for the operation """ return timeout_ms or self.config['request_timeout_ms'] def _refresh_controller_id(self): """Determine the kafka cluster controller """ response = self._send_request_to_node(self._client.least_loaded_node(), MetadataRequest[1]([])) self._controller_id = response.controller_id version = self._client.check_version(self._controller_id) if version < (0, 10, 0): raise UnsupportedVersionError( "Kafka Admin interface not supported for cluster controller version {} < 0.10.0.0" .format(version)) def _send_request_to_node(self, node, request): """Send a kafka protocol message to a specific broker. Will block until the message result is received. :param node: The broker id to which to send the message :param request: The message to send :return: The kafka protocol response for the message :exception: The exception if the message could not be sent """ while not self._client.ready(node): # connection to broker not ready, poll until it is or send will fail with NodeNotReadyError self._client.poll() future = self._client.send(node, request) self._client.poll(future=future) if future.succeeded(): return future.value else: raise future.exception # pylint: disable-msg=raising-bad-type def _send(self, request): """Send a kafka protocol message to the cluster controller. Will block until the message result is received. :param request: The message to send :return The kafka protocol response for the message :exception NodeNotReadyError: If the controller connection can't be established """ remaining_tries = 2 while remaining_tries > 0: remaining_tries = remaining_tries - 1 try: return self._send_request_to_node(self._controller_id, request) except (NotControllerError, KafkaConnectionError) as e: # controller changed? refresh it self._refresh_controller_id() raise NodeNotReadyError(self._controller_id) @staticmethod def _convert_new_topic_request(new_topic): return ( new_topic.name, new_topic.num_partitions, new_topic.replication_factor, [ (partition_id, replicas) for partition_id, replicas in new_topic.replica_assignments.items() ], [(config_key, config_value) for config_key, config_value in new_topic.topic_configs.items()]) def create_topics(self, new_topics, timeout_ms=None, validate_only=None): """Create new topics in the cluster. :param new_topics: Array of NewTopic objects :param timeout_ms: Milliseconds to wait for new topics to be created before broker returns :param validate_only: If True, don't actually create new topics. Not supported by all versions. :return: Appropriate version of CreateTopicResponse class """ version = self._matching_api_version(CreateTopicsRequest) timeout_ms = self._validate_timeout(timeout_ms) if version == 0: if validate_only: raise UnsupportedVersionError( "validate_only not supported on cluster version {}".format( self.config['api_version'])) request = CreateTopicsRequest[version](create_topic_requests=[ self._convert_new_topic_request(new_topic) for new_topic in new_topics ], timeout=timeout_ms) elif version <= 2: validate_only = validate_only or False request = CreateTopicsRequest[version](create_topic_requests=[ self._convert_new_topic_request(new_topic) for new_topic in new_topics ], timeout=timeout_ms, validate_only=validate_only) else: raise UnsupportedVersionError( "missing implementation of CreateTopics for library supported version {}" .format(version)) return self._send(request) def delete_topics(self, topics, timeout_ms=None): """Delete topics from the cluster :param topics: Array of topic name strings :param timeout_ms: Milliseconds to wait for topics to be deleted before broker returns :return: Appropriate version of DeleteTopicsResponse class """ version = self._matching_api_version(DeleteTopicsRequest) timeout_ms = self._validate_timeout(timeout_ms) if version <= 1: request = DeleteTopicsRequest[version](topics=topics, timeout=timeout_ms) else: raise UnsupportedVersionError( "missing implementation of DeleteTopics for library supported version {}" .format(version)) return self._send(request) # list topics functionality is in ClusterMetadata # describe topics functionality is in ClusterMetadata # describe cluster functionality is in ClusterMetadata # describe_acls protocol not implemented # create_acls protocol not implemented # delete_acls protocol not implemented @staticmethod def _convert_describe_config_resource_request(config_resource): return (config_resource.resource_type, config_resource.name, [ config_key for config_key, config_value in config_resource.configs.items() ] if config_resource.configs else None) def describe_configs(self, config_resources, include_synonyms=None): """Fetch configuration parameters for one or more kafka resources. :param config_resources: An array of ConfigResource objects. Any keys in ConfigResource.configs dict will be used to filter the result. The configs dict should be None to get all values. An empty dict will get zero values (as per kafka protocol). :param include_synonyms: If True, return synonyms in response. Not supported by all versions. :return: Appropriate version of DescribeConfigsResponse class """ version = self._matching_api_version(DescribeConfigsRequest) if version == 0: if include_synonyms: raise UnsupportedVersionError( "include_synonyms not supported on cluster version {}". format(self.config['api_version'])) request = DescribeConfigsRequest[version](resources=[ self._convert_describe_config_resource_request(config_resource) for config_resource in config_resources ]) elif version <= 1: include_synonyms = include_synonyms or False request = DescribeConfigsRequest[version]( resources=[ self._convert_describe_config_resource_request( config_resource) for config_resource in config_resources ], include_synonyms=include_synonyms) else: raise UnsupportedVersionError( "missing implementation of DescribeConfigs for library supported version {}" .format(version)) return self._send(request) @staticmethod def _convert_alter_config_resource_request(config_resource): return (config_resource.resource_type, config_resource.name, [ (config_key, config_value) for config_key, config_value in config_resource.configs.items() ]) def alter_configs(self, config_resources): """Alter configuration parameters of one or more kafka resources. :param config_resources: An array of ConfigResource objects. :return: Appropriate version of AlterConfigsResponse class """ version = self._matching_api_version(AlterConfigsRequest) if version == 0: request = AlterConfigsRequest[version](resources=[ self._convert_alter_config_resource_request(config_resource) for config_resource in config_resources ]) else: raise UnsupportedVersionError( "missing implementation of AlterConfigs for library supported version {}" .format(version)) return self._send(request) # alter replica logs dir protocol not implemented # describe log dirs protocol not implemented @staticmethod def _convert_create_partitions_request(topic_name, new_partitions): return (topic_name, (new_partitions.total_count, new_partitions.new_assignments)) def create_partitions(self, topic_partitions, timeout_ms=None, validate_only=None): """Create additional partitions for an existing topic. :param topic_partitions: A map of topic name strings to NewPartition objects :param timeout_ms: Milliseconds to wait for new partitions to be created before broker returns :param validate_only: If True, don't actually create new partitions. :return: Appropriate version of CreatePartitionsResponse class """ version = self._matching_api_version(CreatePartitionsRequest) timeout_ms = self._validate_timeout(timeout_ms) validate_only = validate_only or False if version == 0: request = CreatePartitionsRequest[version]( topic_partitions=[ self._convert_create_partitions_request( topic_name, new_partitions) for topic_name, new_partitions in topic_partitions.items() ], timeout=timeout_ms, validate_only=validate_only) else: raise UnsupportedVersionError( "missing implementation of CreatePartitions for library supported version {}" .format(version)) return self._send(request) # delete records protocol not implemented # create delegation token protocol not implemented # renew delegation token protocol not implemented # expire delegation_token protocol not implemented # describe delegation_token protocol not implemented def describe_consumer_groups(self, group_ids): """Describe a set of consumer groups. :param group_ids: A list of consumer group id names :return: Appropriate version of DescribeGroupsResponse class """ version = self._matching_api_version(DescribeGroupsRequest) if version <= 1: request = DescribeGroupsRequest[version](groups=group_ids) else: raise UnsupportedVersionError( "missing implementation of DescribeGroups for library supported version {}" .format(version)) return self._send(request) def list_consumer_groups(self): """List all consumer groups known to the cluster. :return: Appropriate version of ListGroupsResponse class """ version = self._matching_api_version(ListGroupsRequest) if version <= 1: request = ListGroupsRequest[version]() else: raise UnsupportedVersionError( "missing implementation of ListGroups for library supported version {}" .format(version)) return self._send(request)
if sequence_number is None or processed_stream_data is None: # Handle bad data return None payload['sequence'] = sequence_number payload['data'] = processed_stream_data # print(processed_stream_data) return payload if __name__ == '__main__': # First, check if the topic already exists in kafka kafka_client = KafkaClient(bootstrap_servers=KAFKA_SERVER, api_version=(2, 5, 0)) version = kafka_client.check_version() future = kafka_client.cluster.request_update() kafka_client.poll(future=future) metadata = kafka_client.cluster current_topics = metadata.topics() kafka_client.close() print('Active topics:', current_topics) if KAFKA_TOPIC not in current_topics: print(f'Creating topic {KAFKA_TOPIC}...') kafka_admin_client = KafkaAdminClient(bootstrap_servers=KAFKA_SERVER, api_version=(2, 5, 0))
class KafkaAdminClient(object): """A class for administering the Kafka cluster. Warning: This is an unstable interface that was recently added and is subject to change without warning. In particular, many methods currently return raw protocol tuples. In future releases, we plan to make these into nicer, more pythonic objects. Unfortunately, this will likely break those interfaces. The KafkaAdminClient class will negotiate for the latest version of each message protocol format supported by both the kafka-python client library and the Kafka broker. Usage of optional fields from protocol versions that are not supported by the broker will result in IncompatibleBrokerVersion exceptions. Use of this class requires a minimum broker version >= 0.10.0.0. 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}' reconnect_backoff_ms (int): The amount of time in milliseconds to wait before attempting to reconnect to a given host. Default: 50. reconnect_backoff_max_ms (int): The maximum amount of time in milliseconds to wait when reconnecting to a broker that has repeatedly failed to connect. If provided, the backoff per host will increase exponentially for each consecutive connection failure, up to this maximum. To avoid connection storms, a randomization factor of 0.2 will be applied to the backoff resulting in a random range between 20% below and 20% above the computed value. Default: 1000. request_timeout_ms (int): Client request timeout in milliseconds. Default: 30000. connections_max_idle_ms: Close idle connections after the number of milliseconds specified by this config. The broker closes idle connections after connections.max.idle.ms, so this avoids hitting unexpected socket disconnected errors on the client. Default: 540000 retry_backoff_ms (int): Milliseconds to backoff when retrying on errors. Default: 100. 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. receive_buffer_bytes (int): The size of the TCP receive buffer (SO_RCVBUF) to use when reading data. Default: None (relies on system defaults). 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). 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)] 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 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 broker's hostname. Default: True. ssl_cafile (str): Optional filename of CA file to use in certificate veriication. 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, KafkaClient will attempt to infer the broker version by probing various APIs. Example: (0, 10, 2). Default: None 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 is None selector (selectors.BaseSelector): Provide a specific selector implementation to use for I/O multiplexing. Default: selectors.DefaultSelector metrics (kafka.metrics.Metrics): Optionally provide a metrics instance for capturing network IO stats. Default: None. metric_group_prefix (str): Prefix for metric names. Default: '' sasl_mechanism (str): Authentication mechanism when security_protocol is configured for SASL_PLAINTEXT or SASL_SSL. Valid values are: PLAIN, GSSAPI, OAUTHBEARER. sasl_plain_username (str): username for sasl PLAIN authentication. Required if sasl_mechanism is PLAIN. sasl_plain_password (str): password for sasl PLAIN authentication. Required if sasl_mechanism is PLAIN. sasl_kerberos_service_name (str): Service name to include in GSSAPI sasl mechanism handshake. Default: 'kafka' sasl_oauth_token_provider (AbstractTokenProvider): OAuthBearer token provider instance. (See kafka.oauth.abstract). Default: None """ DEFAULT_CONFIG = { # client configs 'bootstrap_servers': 'localhost', 'client_id': 'kafka-python-' + __version__, 'request_timeout_ms': 30000, 'connections_max_idle_ms': 9 * 60 * 1000, 'reconnect_backoff_ms': 50, 'reconnect_backoff_max_ms': 1000, 'max_in_flight_requests_per_connection': 5, 'receive_buffer_bytes': None, 'send_buffer_bytes': None, 'socket_options': [(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)], 'sock_chunk_bytes': 4096, # undocumented experimental option 'sock_chunk_buffer_count': 1000, # undocumented experimental option 'retry_backoff_ms': 100, 'metadata_max_age_ms': 300000, 'security_protocol': 'PLAINTEXT', 'ssl_context': None, 'ssl_check_hostname': True, 'ssl_cafile': None, 'ssl_certfile': None, 'ssl_keyfile': None, 'ssl_password': None, 'ssl_crlfile': None, 'api_version': None, 'api_version_auto_timeout_ms': 2000, 'selector': selectors.DefaultSelector, 'sasl_mechanism': None, 'sasl_plain_username': None, 'sasl_plain_password': None, 'sasl_kerberos_service_name': 'kafka', 'sasl_oauth_token_provider': None, # metrics configs 'metric_reporters': [], 'metrics_num_samples': 2, 'metrics_sample_window_ms': 30000, } 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.") def close(self): """Close the KafkaAdminClient connection to the Kafka broker.""" if not hasattr(self, '_closed') or self._closed: log.info("KafkaAdminClient already closed.") return self._metrics.close() self._client.close() self._closed = True log.debug("KafkaAdminClient is now closed.") def _matching_api_version(self, operation): """Find the latest version of the protocol operation supported by both this library and the broker. This resolves to the lesser of either the latest api version this library supports, or the max version supported by the broker. :param operation: A list of protocol operation versions from kafka.protocol. :return: The max matching version number between client and broker. """ version = min(len(operation) - 1, self._client.get_api_versions()[operation[0].API_KEY][1]) if version < self._client.get_api_versions()[operation[0].API_KEY][0]: # max library version is less than min broker version. Currently, # no Kafka versions specify a min msg version. Maybe in the future? raise IncompatibleBrokerVersion( "No version of the '{}' Kafka protocol is supported by both the client and broker." .format(operation.__name__)) return version def _validate_timeout(self, timeout_ms): """Validate the timeout is set or use the configuration default. :param timeout_ms: The timeout provided by api call, in milliseconds. :return: The timeout to use for the operation. """ return timeout_ms or self.config['request_timeout_ms'] def _refresh_controller_id(self): """Determine the Kafka cluster controller.""" version = self._matching_api_version(MetadataRequest) if 1 <= version <= 6: request = MetadataRequest[version]() response = self._send_request_to_node(self._client.least_loaded_node(), request) controller_id = response.controller_id # verify the controller is new enough to support our requests controller_version = self._client.check_version(controller_id) if controller_version < (0, 10, 0): raise IncompatibleBrokerVersion( "The controller appears to be running Kafka {}. KafkaAdminClient requires brokers >= 0.10.0.0." .format(controller_version)) self._controller_id = controller_id else: raise UnrecognizedBrokerVersion( "Kafka Admin interface cannot determine the controller using MetadataRequest_v{}." .format(version)) def _find_group_coordinator_id(self, group_id): """Find the broker node_id of the coordinator of the given group. Sends a FindCoordinatorRequest message to the cluster. Will block until the FindCoordinatorResponse is received. Any errors are immediately raised. :param group_id: The consumer group ID. This is typically the group name as a string. :return: The node_id of the broker that is the coordinator. """ # Note: Java may change how this is implemented in KAFKA-6791. # # TODO add support for dynamically picking version of # GroupCoordinatorRequest which was renamed to FindCoordinatorRequest. # When I experimented with this, GroupCoordinatorResponse_v1 didn't # match GroupCoordinatorResponse_v0 and I couldn't figure out why. gc_request = GroupCoordinatorRequest[0](group_id) gc_response = self._send_request_to_node(self._client.least_loaded_node(), gc_request) # use the extra error checking in add_group_coordinator() rather than # immediately returning the group coordinator. success = self._client.cluster.add_group_coordinator(group_id, gc_response) if not success: error_type = Errors.for_code(gc_response.error_code) assert error_type is not Errors.NoError # Note: When error_type.retriable, Java will retry... see # KafkaAdminClient's handleFindCoordinatorError method raise error_type( "Could not identify group coordinator for group_id '{}' from response '{}'." .format(group_id, gc_response)) group_coordinator = self._client.cluster.coordinator_for_group(group_id) # will be None if the coordinator was never populated, which should never happen here assert group_coordinator is not None # will be -1 if add_group_coordinator() failed... but by this point the # error should have been raised. assert group_coordinator != -1 return group_coordinator def _send_request_to_node(self, node_id, request): """Send a Kafka protocol message to a specific broker. Will block until the message result is received. :param node_id: The broker id to which to send the message. :param request: The message to send. :return: The Kafka protocol response for the message. :exception: The exception if the message could not be sent. """ while not self._client.ready(node_id): # poll until the connection to broker is ready, otherwise send() # will fail with NodeNotReadyError self._client.poll() future = self._client.send(node_id, request) self._client.poll(future=future) if future.succeeded(): return future.value else: raise future.exception # pylint: disable-msg=raising-bad-type def _send_request_to_controller(self, request): """Send a Kafka protocol message to the cluster controller. Will block until the message result is received. :param request: The message to send. :return: The Kafka protocol response for the message. """ tries = 2 # in case our cached self._controller_id is outdated while tries: tries -= 1 response = self._send_request_to_node(self._controller_id, request) # In Java, the error fieldname is inconsistent: # - CreateTopicsResponse / CreatePartitionsResponse uses topic_errors # - DeleteTopicsResponse uses topic_error_codes # So this is a little brittle in that it assumes all responses have # one of these attributes and that they always unpack into # (topic, error_code) tuples. topic_error_tuples = (response.topic_errors if hasattr(response, 'topic_errors') else response.topic_error_codes) # Also small py2/py3 compatibility -- py3 can ignore extra values # during unpack via: for x, y, *rest in list_of_values. py2 cannot. # So for now we have to map across the list and explicitly drop any # extra values (usually the error_message) for topic, error_code in map(lambda e: e[:2], topic_error_tuples): error_type = Errors.for_code(error_code) if tries and error_type is NotControllerError: # No need to inspect the rest of the errors for # non-retriable errors because NotControllerError should # either be thrown for all errors or no errors. self._refresh_controller_id() break elif error_type is not Errors.NoError: raise error_type( "Request '{}' failed with response '{}'." .format(request, response)) else: return response raise RuntimeError("This should never happen, please file a bug with full stacktrace if encountered") @staticmethod def _convert_new_topic_request(new_topic): return ( new_topic.name, new_topic.num_partitions, new_topic.replication_factor, [ (partition_id, replicas) for partition_id, replicas in new_topic.replica_assignments.items() ], [ (config_key, config_value) for config_key, config_value in new_topic.topic_configs.items() ] ) def create_topics(self, new_topics, timeout_ms=None, validate_only=False): """Create new topics in the cluster. :param new_topics: A list of NewTopic objects. :param timeout_ms: Milliseconds to wait for new topics to be created before the broker returns. :param validate_only: If True, don't actually create new topics. Not supported by all versions. Default: False :return: Appropriate version of CreateTopicResponse class. """ version = self._matching_api_version(CreateTopicsRequest) timeout_ms = self._validate_timeout(timeout_ms) if version == 0: if validate_only: raise IncompatibleBrokerVersion( "validate_only requires CreateTopicsRequest >= v1, which is not supported by Kafka {}." .format(self.config['api_version'])) request = CreateTopicsRequest[version]( create_topic_requests=[self._convert_new_topic_request(new_topic) for new_topic in new_topics], timeout=timeout_ms ) elif version <= 2: request = CreateTopicsRequest[version]( create_topic_requests=[self._convert_new_topic_request(new_topic) for new_topic in new_topics], timeout=timeout_ms, validate_only=validate_only ) else: raise NotImplementedError( "Support for CreateTopics v{} has not yet been added to KafkaAdminClient." .format(version)) # TODO convert structs to a more pythonic interface # TODO raise exceptions if errors return self._send_request_to_controller(request) def delete_topics(self, topics, timeout_ms=None): """Delete topics from the cluster. :param topics: A list of topic name strings. :param timeout_ms: Milliseconds to wait for topics to be deleted before the broker returns. :return: Appropriate version of DeleteTopicsResponse class. """ version = self._matching_api_version(DeleteTopicsRequest) timeout_ms = self._validate_timeout(timeout_ms) if version <= 1: request = DeleteTopicsRequest[version]( topics=topics, timeout=timeout_ms ) response = self._send_request_to_controller(request) else: raise NotImplementedError( "Support for DeleteTopics v{} has not yet been added to KafkaAdminClient." .format(version)) return response # list topics functionality is in ClusterMetadata # Note: if implemented here, send the request to the least_loaded_node() # describe topics functionality is in ClusterMetadata # Note: if implemented here, send the request to the controller # describe cluster functionality is in ClusterMetadata # Note: if implemented here, send the request to the least_loaded_node() # describe_acls protocol not yet implemented # Note: send the request to the least_loaded_node() # create_acls protocol not yet implemented # Note: send the request to the least_loaded_node() # delete_acls protocol not yet implemented # Note: send the request to the least_loaded_node() @staticmethod def _convert_describe_config_resource_request(config_resource): return ( config_resource.resource_type, config_resource.name, [ config_key for config_key, config_value in config_resource.configs.items() ] if config_resource.configs else None ) def describe_configs(self, config_resources, include_synonyms=False): """Fetch configuration parameters for one or more Kafka resources. :param config_resources: An list of ConfigResource objects. Any keys in ConfigResource.configs dict will be used to filter the result. Setting the configs dict to None will get all values. An empty dict will get zero values (as per Kafka protocol). :param include_synonyms: If True, return synonyms in response. Not supported by all versions. Default: False. :return: Appropriate version of DescribeConfigsResponse class. """ version = self._matching_api_version(DescribeConfigsRequest) if version == 0: if include_synonyms: raise IncompatibleBrokerVersion( "include_synonyms requires DescribeConfigsRequest >= v1, which is not supported by Kafka {}." .format(self.config['api_version'])) request = DescribeConfigsRequest[version]( resources=[self._convert_describe_config_resource_request(config_resource) for config_resource in config_resources] ) elif version == 1: request = DescribeConfigsRequest[version]( resources=[self._convert_describe_config_resource_request(config_resource) for config_resource in config_resources], include_synonyms=include_synonyms ) else: raise NotImplementedError( "Support for DescribeConfigs v{} has not yet been added to KafkaAdminClient." .format(version)) return self._send_request_to_node(self._client.least_loaded_node(), request) @staticmethod def _convert_alter_config_resource_request(config_resource): return ( config_resource.resource_type, config_resource.name, [ (config_key, config_value) for config_key, config_value in config_resource.configs.items() ] ) def alter_configs(self, config_resources): """Alter configuration parameters of one or more Kafka resources. Warning: This is currently broken for BROKER resources because those must be sent to that specific broker, versus this always picks the least-loaded node. See the comment in the source code for details. We would happily accept a PR fixing this. :param config_resources: A list of ConfigResource objects. :return: Appropriate version of AlterConfigsResponse class. """ version = self._matching_api_version(AlterConfigsRequest) if version == 0: request = AlterConfigsRequest[version]( resources=[self._convert_alter_config_resource_request(config_resource) for config_resource in config_resources] ) else: raise NotImplementedError( "Support for AlterConfigs v{} has not yet been added to KafkaAdminClient." .format(version)) # TODO the Java client has the note: # // We must make a separate AlterConfigs request for every BROKER resource we want to alter # // and send the request to that specific broker. Other resources are grouped together into # // a single request that may be sent to any broker. # # So this is currently broken as it always sends to the least_loaded_node() return self._send_request_to_node(self._client.least_loaded_node(), request) # alter replica logs dir protocol not yet implemented # Note: have to lookup the broker with the replica assignment and send the request to that broker # describe log dirs protocol not yet implemented # Note: have to lookup the broker with the replica assignment and send the request to that broker @staticmethod def _convert_create_partitions_request(topic_name, new_partitions): return ( topic_name, ( new_partitions.total_count, new_partitions.new_assignments ) ) def create_partitions(self, topic_partitions, timeout_ms=None, validate_only=False): """Create additional partitions for an existing topic. :param topic_partitions: A map of topic name strings to NewPartition objects. :param timeout_ms: Milliseconds to wait for new partitions to be created before the broker returns. :param validate_only: If True, don't actually create new partitions. Default: False :return: Appropriate version of CreatePartitionsResponse class. """ version = self._matching_api_version(CreatePartitionsRequest) timeout_ms = self._validate_timeout(timeout_ms) if version == 0: request = CreatePartitionsRequest[version]( topic_partitions=[self._convert_create_partitions_request(topic_name, new_partitions) for topic_name, new_partitions in topic_partitions.items()], timeout=timeout_ms, validate_only=validate_only ) else: raise NotImplementedError( "Support for CreatePartitions v{} has not yet been added to KafkaAdminClient." .format(version)) return self._send_request_to_controller(request) # delete records protocol not yet implemented # Note: send the request to the partition leaders # create delegation token protocol not yet implemented # Note: send the request to the least_loaded_node() # renew delegation token protocol not yet implemented # Note: send the request to the least_loaded_node() # expire delegation_token protocol not yet implemented # Note: send the request to the least_loaded_node() # describe delegation_token protocol not yet implemented # Note: send the request to the least_loaded_node() def describe_consumer_groups(self, group_ids, group_coordinator_id=None): """Describe a set of consumer groups. Any errors are immediately raised. :param group_ids: A list of consumer group IDs. These are typically the group names as strings. :param group_coordinator_id: The node_id of the groups' coordinator broker. If set to None, it will query the cluster for each group to find that group's coordinator. Explicitly specifying this can be useful for avoiding extra network round trips if you already know the group coordinator. This is only useful when all the group_ids have the same coordinator, otherwise it will error. Default: None. :return: A list of group descriptions. For now the group descriptions are the raw results from the DescribeGroupsResponse. Long-term, we plan to change this to return namedtuples as well as decoding the partition assignments. """ group_descriptions = [] version = self._matching_api_version(DescribeGroupsRequest) for group_id in group_ids: if group_coordinator_id is not None: this_groups_coordinator_id = group_coordinator_id else: this_groups_coordinator_id = self._find_group_coordinator_id(group_id) if version <= 1: # Note: KAFKA-6788 A potential optimization is to group the # request per coordinator and send one request with a list of # all consumer groups. Java still hasn't implemented this # because the error checking is hard to get right when some # groups error and others don't. request = DescribeGroupsRequest[version](groups=(group_id,)) response = self._send_request_to_node(this_groups_coordinator_id, request) assert len(response.groups) == 1 # TODO need to implement converting the response tuple into # a more accessible interface like a namedtuple and then stop # hardcoding tuple indices here. Several Java examples, # including KafkaAdminClient.java group_description = response.groups[0] error_code = group_description[0] error_type = Errors.for_code(error_code) # Java has the note: KAFKA-6789, we can retry based on the error code if error_type is not Errors.NoError: raise error_type( "Request '{}' failed with response '{}'." .format(request, response)) # TODO Java checks the group protocol type, and if consumer # (ConsumerProtocol.PROTOCOL_TYPE) or empty string, it decodes # the members' partition assignments... that hasn't yet been # implemented here so just return the raw struct results group_descriptions.append(group_description) else: raise NotImplementedError( "Support for DescribeGroups v{} has not yet been added to KafkaAdminClient." .format(version)) return group_descriptions def list_consumer_groups(self, broker_ids=None): """List all consumer groups known to the cluster. This returns a list of Consumer Group tuples. The tuples are composed of the consumer group name and the consumer group protocol type. Only consumer groups that store their offsets in Kafka are returned. The protocol type will be an empty string for groups created using Kafka < 0.9 APIs because, although they store their offsets in Kafka, they don't use Kafka for group coordination. For groups created using Kafka >= 0.9, the protocol type will typically be "consumer". As soon as any error is encountered, it is immediately raised. :param broker_ids: A list of broker node_ids to query for consumer groups. If set to None, will query all brokers in the cluster. Explicitly specifying broker(s) can be useful for determining which consumer groups are coordinated by those broker(s). Default: None :return list: List of tuples of Consumer Groups. :exception GroupCoordinatorNotAvailableError: The coordinator is not available, so cannot process requests. :exception GroupLoadInProgressError: The coordinator is loading and hence can't process requests. """ # While we return a list, internally use a set to prevent duplicates # because if a group coordinator fails after being queried, and its # consumer groups move to new brokers that haven't yet been queried, # then the same group could be returned by multiple brokers. consumer_groups = set() if broker_ids is None: broker_ids = [broker.nodeId for broker in self._client.cluster.brokers()] version = self._matching_api_version(ListGroupsRequest) if version <= 2: request = ListGroupsRequest[version]() for broker_id in broker_ids: response = self._send_request_to_node(broker_id, request) error_type = Errors.for_code(response.error_code) if error_type is not Errors.NoError: raise error_type( "Request '{}' failed with response '{}'." .format(request, response)) consumer_groups.update(response.groups) else: raise NotImplementedError( "Support for ListGroups v{} has not yet been added to KafkaAdminClient." .format(version)) return list(consumer_groups) def list_consumer_group_offsets(self, group_id, group_coordinator_id=None, partitions=None): """Fetch Consumer Group Offsets. Note: This does not verify that the group_id or partitions actually exist in the cluster. As soon as any error is encountered, it is immediately raised. :param group_id: The consumer group id name for which to fetch offsets. :param group_coordinator_id: The node_id of the group's coordinator broker. If set to None, will query the cluster to find the group coordinator. Explicitly specifying this can be useful to prevent that extra network round trip if you already know the group coordinator. Default: None. :param partitions: A list of TopicPartitions for which to fetch offsets. On brokers >= 0.10.2, this can be set to None to fetch all known offsets for the consumer group. Default: None. :return dictionary: A dictionary with TopicPartition keys and OffsetAndMetada values. Partitions that are not specified and for which the group_id does not have a recorded offset are omitted. An offset value of `-1` indicates the group_id has no offset for that TopicPartition. A `-1` can only happen for partitions that are explicitly specified. """ group_offsets_listing = {} if group_coordinator_id is None: group_coordinator_id = self._find_group_coordinator_id(group_id) version = self._matching_api_version(OffsetFetchRequest) if version <= 3: if partitions is None: if version <= 1: raise ValueError( """OffsetFetchRequest_v{} requires specifying the partitions for which to fetch offsets. Omitting the partitions is only supported on brokers >= 0.10.2. For details, see KIP-88.""".format(version)) topics_partitions = None else: # transform from [TopicPartition("t1", 1), TopicPartition("t1", 2)] to [("t1", [1, 2])] topics_partitions_dict = defaultdict(set) for topic, partition in partitions: topics_partitions_dict[topic].add(partition) topics_partitions = list(six.iteritems(topics_partitions_dict)) request = OffsetFetchRequest[version](group_id, topics_partitions) response = self._send_request_to_node(group_coordinator_id, request) if version > 1: # OffsetFetchResponse_v1 lacks a top-level error_code error_type = Errors.for_code(response.error_code) if error_type is not Errors.NoError: # optionally we could retry if error_type.retriable raise error_type( "Request '{}' failed with response '{}'." .format(request, response)) # transform response into a dictionary with TopicPartition keys and # OffsetAndMetada values--this is what the Java AdminClient returns for topic, partitions in response.topics: for partition, offset, metadata, error_code in partitions: error_type = Errors.for_code(error_code) if error_type is not Errors.NoError: raise error_type( "Unable to fetch offsets for group_id {}, topic {}, partition {}" .format(group_id, topic, partition)) group_offsets_listing[TopicPartition(topic, partition)] = OffsetAndMetadata(offset, metadata) else: raise NotImplementedError( "Support for OffsetFetch v{} has not yet been added to KafkaAdminClient." .format(version)) return group_offsets_listing