class kafka.KafkaProducer(**configs)[source]

A Kafka client that publishes records to the Kafka cluster.

The producer is thread safe and sharing a single producer instance across threads will generally be faster than having multiple instances.

The producer consists of a pool of buffer space that holds records that haven’t yet been transmitted to the server as well as a background I/O thread that is responsible for turning these records into requests and transmitting them to the cluster.

send() is asynchronous. When called it adds the record to a buffer of pending record sends and immediately returns. This allows the producer to batch together individual records for efficiency.

The ‘acks’ config controls the criteria under which requests are considered complete. The “all” setting will result in blocking on the full commit of the record, the slowest but most durable setting.

If the request fails, the producer can automatically retry, unless ‘retries’ is configured to 0. Enabling retries also opens up the possibility of duplicates (see the documentation on message delivery semantics for details: http://kafka.apache.org/documentation.html#semantics ).

The producer maintains buffers of unsent records for each partition. These buffers are of a size specified by the ‘batch_size’ config. Making this larger can result in more batching, but requires more memory (since we will generally have one of these buffers for each active partition).

By default a buffer is available to send immediately even if there is additional unused space in the buffer. However if you want to reduce the number of requests you can set ‘linger_ms’ to something greater than 0. This will instruct the producer to wait up to that number of milliseconds before sending a request in hope that more records will arrive to fill up the same batch. This is analogous to Nagle’s algorithm in TCP. Note that records that arrive close together in time will generally batch together even with linger_ms=0 so under heavy load batching will occur regardless of the linger configuration; however setting this to something larger than 0 can lead to fewer, more efficient requests when not under maximal load at the cost of a small amount of latency.

The buffer_memory controls the total amount of memory available to the producer for buffering. If records are sent faster than they can be transmitted to the server then this buffer space will be exhausted. When the buffer space is exhausted additional send calls will block.

The key_serializer and value_serializer instruct how to turn the key and value objects the user provides into bytes.

Keyword Arguments:
  • bootstrap_servers – ‘host[:port]’ string (or list of ‘host[:port]’ strings) that the producer 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. Default: ‘kafka-python-producer-#’ (appended with a unique number per instance)
  • key_serializer (callable) – used to convert user-supplied keys to bytes If not None, called as f(key), should return bytes. Default: None.
  • value_serializer (callable) – used to convert user-supplied message values to bytes. If not None, called as f(value), should return bytes. Default: None.
  • acks (0, 1, 'all') –

    The number of acknowledgments the producer requires the leader to have received before considering a request complete. This controls the durability of records that are sent. The following settings are common:

    0: Producer will not wait for any acknowledgment from the server.
    The message will immediately be added to the socket buffer and considered sent. No guarantee can be made that the server has received the record in this case, and the retries configuration will not take effect (as the client won’t generally know of any failures). The offset given back for each record will always be set to -1.
    1: Wait for leader to write the record to its local log only.
    Broker will respond without awaiting full acknowledgement from all followers. In this case should the leader fail immediately after acknowledging the record but before the followers have replicated it then the record will be lost.
    all: Wait for the full set of in-sync replicas to write the record.
    This guarantees that the record will not be lost as long as at least one in-sync replica remains alive. This is the strongest available guarantee.

    If unset, defaults to acks=1.

  • compression_type (str) – The compression type for all data generated by the producer. Valid values are ‘gzip’, ‘snappy’, ‘lz4’, or None. Compression is of full batches of data, so the efficacy of batching will also impact the compression ratio (more batching means better compression). Default: None.
  • retries (int) – Setting a value greater than zero will cause the client to resend any record whose send fails with a potentially transient error. Note that this retry is no different than if the client resent the record upon receiving the error. Allowing retries without setting max_in_flight_requests_per_connection to 1 will potentially change the ordering of records because if two batches are sent to a single partition, and the first fails and is retried but the second succeeds, then the records in the second batch may appear first. Default: 0.
  • batch_size (int) – Requests sent to brokers will contain multiple batches, one for each partition with data available to be sent. A small batch size will make batching less common and may reduce throughput (a batch size of zero will disable batching entirely). Default: 16384
  • linger_ms (int) – The producer groups together any records that arrive in between request transmissions into a single batched request. Normally this occurs only under load when records arrive faster than they can be sent out. However in some circumstances the client may want to reduce the number of requests even under moderate load. This setting accomplishes this by adding a small amount of artificial delay; that is, rather than immediately sending out a record the producer will wait for up to the given delay to allow other records to be sent so that the sends can be batched together. This can be thought of as analogous to Nagle’s algorithm in TCP. This setting gives the upper bound on the delay for batching: once we get batch_size worth of records for a partition it will be sent immediately regardless of this setting, however if we have fewer than this many bytes accumulated for this partition we will ‘linger’ for the specified time waiting for more records to show up. This setting defaults to 0 (i.e. no delay). Setting linger_ms=5 would have the effect of reducing the number of requests sent but would add up to 5ms of latency to records sent in the absense of load. Default: 0.
  • partitioner (callable) – Callable used to determine which partition each message is assigned to. Called (after key serialization): partitioner(key_bytes, all_partitions, available_partitions). The default partitioner implementation hashes each non-None key using the same murmur2 algorithm as the java client so that messages with the same key are assigned to the same partition. When a key is None, the message is delivered to a random partition (filtered to partitions with available leaders only, if possible).
  • buffer_memory (int) – The total bytes of memory the producer should use to buffer records waiting to be sent to the server. If records are sent faster than they can be delivered to the server the producer will block up to max_block_ms, raising an exception on timeout. In the current implementation, this setting is an approximation. Default: 33554432 (32MB)
  • max_block_ms (int) – Number of milliseconds to block during send() and partitions_for(). These methods can be blocked either because the buffer is full or metadata unavailable. Blocking in the user-supplied serializers or partitioner will not be counted against this timeout. Default: 60000.
  • max_request_size (int) – The maximum size of a request. This is also effectively a cap on the maximum record size. Note that the server has its own cap on record size which may be different from this. This setting will limit the number of record batches the producer will send in a single request to avoid sending huge requests. Default: 1048576.
  • 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
  • retry_backoff_ms (int) – Milliseconds to backoff when retrying on errors. Default: 100.
  • request_timeout_ms (int) – Client request timeout in milliseconds. Default: 30000.
  • 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)]
  • 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.
  • 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.
  • security_protocol (str) – Protocol used to communicate with brokers. Valid values are: PLAINTEXT, SSL, SASL_PLAINTEXT, SASL_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.
  • ssl_password (str) – optional password to be used when loading the certificate chain. default: none.
  • ssl_crlfile (str) – optional filename containing the CRL to check for certificate expiration. By default, no CRL check is done. When providing a file, only the leaf certificate will be checked against this CRL. The CRL can only be checked with Python 3.4+ or 2.7.9+. default: none.
  • api_version (tuple) – Specify which Kafka API version to use. If set to None, the client will attempt to infer the broker version by probing various APIs. For a full list of supported versions, see KafkaClient.API_VERSIONS. 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 set to ‘auto’
  • metric_reporters (list) – A list of classes to use as metrics reporters. Implementing the AbstractMetricsReporter interface allows plugging in classes that will be notified of new metric creation. Default: []
  • metrics_num_samples (int) – The number of samples maintained to compute metrics. Default: 2
  • metrics_sample_window_ms (int) – The maximum age in milliseconds of samples used to compute metrics. Default: 30000
  • selector (selectors.BaseSelector) – Provide a specific selector implementation to use for I/O multiplexing. Default: selectors.DefaultSelector
  • 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


Configuration parameters are described in more detail at https://kafka.apache.org/0100/configuration.html#producerconfigs


Close this producer.

Parameters:timeout (float, optional) – timeout in seconds to wait for completion.

Invoking this method makes all buffered records immediately available to send (even if linger_ms is greater than 0) and blocks on the completion of the requests associated with these records. The post-condition of flush() is that any previously sent record will have completed (e.g. Future.is_done() == True). A request is considered completed when either it is successfully acknowledged according to the ‘acks’ configuration for the producer, or it results in an error.

Other threads can continue sending messages while one thread is blocked waiting for a flush call to complete; however, no guarantee is made about the completion of messages sent after the flush call begins.

Parameters:timeout (float, optional) – timeout in seconds to wait for completion.
Raises:KafkaTimeoutError – failure to flush buffered records within the provided timeout

Warning: this is an unstable interface. It may change in future releases without warning


Returns set of all known partitions for the topic.

send(topic, value=None, key=None, partition=None, timestamp_ms=None)[source]

Publish a message to a topic.

  • topic (str) – topic where the message will be published
  • value (optional) – message value. Must be type bytes, or be serializable to bytes via configured value_serializer. If value is None, key is required and message acts as a ‘delete’. See kafka compaction documentation for more details: http://kafka.apache.org/documentation.html#compaction (compaction requires kafka >= 0.8.1)
  • partition (int, optional) – optionally specify a partition. If not set, the partition will be selected using the configured ‘partitioner’.
  • key (optional) – a key to associate with the message. Can be used to determine which partition to send the message to. If partition is None (and producer’s partitioner config is left as default), then messages with the same key will be delivered to the same partition (but if key is None, partition is chosen randomly). Must be type bytes, or be serializable to bytes via configured key_serializer.
  • timestamp_ms (int, optional) – epoch milliseconds (from Jan 1 1970 UTC) to use as the message timestamp. Defaults to current time.

resolves to RecordMetadata

Return type:



KafkaTimeoutError – if unable to fetch topic metadata, or unable to obtain memory buffer prior to configured max_block_ms