Skip to content

Latest commit

 

History

History
2018 lines (1534 loc) · 103 KB

INTRODUCTION.md

File metadata and controls

2018 lines (1534 loc) · 103 KB
Raw

Introduction to librdkafka - the Apache Kafka C/C++ client library

librdkafka is a high performance C implementation of the Apache Kafka client, providing a reliable and performant client for production use. librdkafka also provides a native C++ interface.

Table of Contents

Performance

librdkafka is a multi-threaded library designed for use on modern hardware and it attempts to keep memory copying to a minimum. The payload of produced or consumed messages may pass through without any copying (if so desired by the application) putting no limit on message sizes.

librdkafka allows you to decide if high throughput is the name of the game, or if a low latency service is required, or a balance between the two, all through the configuration property interface.

The single most important configuration properties for performance tuning is linger.ms - how long to wait for batch.num.messages or batch.size to fill up in the local per-partition queue before sending the batch of messages to the broker.

In low throughput scenarios, a lower value improves latency. As throughput increases, the cost of each broker request becomes significant impacting both maximum throughput and latency. For higher throughput applications, latency will typically be lower using a higher linger.ms due to larger batches resulting in a lesser number of requests, yielding decreased per-message load on the broker. A good general purpose setting is 5ms. For applications seeking maximum throughput, the recommended value is >= 50ms.

High throughput

The key to high throughput is message batching - waiting for a certain amount of messages to accumulate in the local queue before sending them off in one large message set or batch to the peer. This amortizes the messaging overhead and eliminates the adverse effect of the round trip time (rtt).

linger.ms (also called queue.buffering.max.ms) allows librdkafka to wait up to the specified amount of time to accumulate up to batch.num.messages or batch.size in a single batch (MessageSet) before sending to the broker. The larger the batch the higher the throughput. Enabling msg debugging (set debug property to msg) will emit log messages for the accumulation process which lets you see what batch sizes are being produced.

Example using linger.ms=1:

... test [0]: MessageSet with 1514 message(s) delivered
... test [3]: MessageSet with 1690 message(s) delivered
... test [0]: MessageSet with 1720 message(s) delivered
... test [3]: MessageSet with 2 message(s) delivered
... test [3]: MessageSet with 4 message(s) delivered
... test [0]: MessageSet with 4 message(s) delivered
... test [3]: MessageSet with 11 message(s) delivered

Example using linger.ms=1000:

... test [0]: MessageSet with 10000 message(s) delivered
... test [0]: MessageSet with 10000 message(s) delivered
... test [0]: MessageSet with 4667 message(s) delivered
... test [3]: MessageSet with 10000 message(s) delivered
... test [3]: MessageSet with 10000 message(s) delivered
... test [3]: MessageSet with 4476 message(s) delivered

The default setting of linger.ms=0.1 is not suitable for high throughput, it is recommended to set this value to >50ms, with throughput leveling out somewhere around 100-1000ms depending on message produce pattern and sizes.

These setting are set globally (rd_kafka_conf_t) but applies on a per topic+partition basis.

Low latency

When low latency messaging is required the linger.ms should be tuned to the maximum permitted producer-side latency. Setting linger.ms to 0 or 0.1 will make sure messages are sent as soon as possible. Lower buffering time leads to smaller batches and larger per-message overheads, increasing network, memory and CPU usage for producers, brokers and consumers.

See How to decrease message latency for more info.

Latency measurement

End-to-end latency is preferably measured by synchronizing clocks on producers and consumers and using the message timestamp on the consumer to calculate the full latency. Make sure the topic's log.message.timestamp.type is set to the default CreateTime (Kafka topic configuration, not librdkafka topic).

Latencies are typically incurred by the producer, network and broker, the consumer effect on end-to-end latency is minimal.

To break down the end-to-end latencies and find where latencies are adding up there are a number of metrics available through librdkafka statistics on the producer:

  • brokers[].int_latency is the time, per message, between produce() and the message being written to a MessageSet and ProduceRequest. High int_latency indicates CPU core contention: check CPU load and, involuntary context switches (/proc/<..>/status). Consider using a machine/instance with more CPU cores. This metric is only relevant on the producer.

  • brokers[].outbuf_latency is the time, per protocol request (such as ProduceRequest), between the request being enqueued (which happens right after int_latency) and the time the request is written to the TCP socket connected to the broker. High outbuf_latency indicates CPU core contention or network congestion: check CPU load and socket SendQ (netstat -anp | grep :9092).

  • brokers[].rtt is the time, per protocol request, between the request being written to the TCP socket and the time the response is received from the broker. High rtt indicates broker load or network congestion: check broker metrics, local socket SendQ, network performance, etc.

  • brokers[].throttle is the time, per throttled protocol request, the broker throttled/delayed handling of a request due to usage quotas. The throttle time will also be reflected in rtt.

  • topics[].batchsize is the size of individual Producer MessageSet batches. See below.

  • topics[].batchcnt is the number of messages in individual Producer MessageSet batches. Due to Kafka protocol overhead a batch with few messages will have a higher relative processing and size overhead than a batch with many messages. Use the linger.ms client configuration property to set the maximum amount of time allowed for accumulating a single batch, the larger the value the larger the batches will grow, thus increasing efficiency. When producing messages at a high rate it is recommended to increase linger.ms, which will improve throughput and in some cases also latency.

See STATISTICS.md for the full definition of metrics. A JSON schema for the statistics is available in statistics-schema.json.

Compression

Producer message compression is enabled through the compression.codec configuration property.

Compression is performed on the batch of messages in the local queue, the larger the batch the higher likelyhood of a higher compression ratio. The local batch queue size is controlled through the batch.num.messages, batch.size, and linger.ms configuration properties as described in the High throughput chapter above.

Message reliability

Message reliability is an important factor of librdkafka - an application can rely fully on librdkafka to deliver a message according to the specified configuration (request.required.acks and message.send.max.retries, etc).

If the topic configuration property request.required.acks is set to wait for message commit acknowledgements from brokers (any value but 0, see CONFIGURATION.md for specifics) then librdkafka will hold on to the message until all expected acks have been received, gracefully handling the following events:

  • Broker connection failure
  • Topic leader change
  • Produce errors signaled by the broker
  • Network problems

We recommend request.required.acks to be set to all to make sure produced messages are acknowledged by all in-sync replica brokers.

This is handled automatically by librdkafka and the application does not need to take any action at any of the above events. The message will be resent up to message.send.max.retries times before reporting a failure back to the application.

The delivery report callback is used by librdkafka to signal the status of a message back to the application, it will be called once for each message to report the status of message delivery:

  • If error_code is non-zero the message delivery failed and the error_code indicates the nature of the failure (rd_kafka_resp_err_t enum).
  • If error_code is zero the message has been successfully delivered.

See Producer API chapter for more details on delivery report callback usage.

The delivery report callback is optional but highly recommended.

Producer message delivery success

When a ProduceRequest is successfully handled by the broker and a ProduceResponse is received (also called the ack) without an error code the messages from the ProduceRequest are enqueued on the delivery report queue (if a delivery report callback has been set) and will be passed to the application on the next invocation rd_kafka_poll().

Producer message delivery failure

The following sub-chapters explains how different produce errors are handled.

If the error is retryable and there are remaining retry attempts for the given message(s), an automatic retry will be scheduled by librdkafka, these retries are not visible to the application.

Only permanent errors and temporary errors that have reached their maximum retry count will generate a delivery report event to the application with an error code set.

The application should typically not attempt to retry producing the message on failure, but instead configure librdkafka to perform these retries using the retries and retry.backoff.ms configuration properties.

Error: Timed out in transmission queue

Internal error ERR__TIMED_OUT_QUEUE.

The connectivity to the broker may be stalled due to networking contention, local or remote system issues, etc, and the request has not yet been sent.

The producer can be certain that the message has not been sent to the broker.

This is a retryable error, but is not counted as a retry attempt since the message was never actually transmitted.

A retry by librdkafka at this point will not cause duplicate messages.

Error: Timed out in flight to/from broker

Internal error ERR__TIMED_OUT, ERR__TRANSPORT.

Same reasons as for Timed out in transmission queue above, with the difference that the message may have been sent to the broker and might be stalling waiting for broker replicas to ack the message, or the response could be stalled due to networking issues. At this point the producer can't know if the message reached the broker, nor if the broker wrote the message to disk and replicas.

This is a retryable error.

A retry by librdkafka at this point may cause duplicate messages.

Error: Temporary broker-side error

Broker errors ERR_REQUEST_TIMED_OUT, ERR_NOT_ENOUGH_REPLICAS, ERR_NOT_ENOUGH_REPLICAS_AFTER_APPEND.

These errors are considered temporary and librdkafka is will retry them if permitted by configuration.

Error: Temporary errors due to stale metadata

Broker errors ERR_LEADER_NOT_AVAILABLE, ERR_NOT_LEADER_FOR_PARTITION.

These errors are considered temporary and a retry is warranted, a metadata request is automatically sent to find a new leader for the partition.

A retry by librdkafka at this point will not cause duplicate messages.

Error: Local time out

Internal error ERR__MSG_TIMED_OUT.

The message could not be successfully transmitted before message.timeout.ms expired, typically due to no leader being available or no broker connection. The message may have been retried due to other errors but those error messages are abstracted by the ERR__MSG_TIMED_OUT error code.

Since the message.timeout.ms has passed there will be no more retries by librdkafka.

Error: Permanent errors

Any other error is considered a permanent error and the message will fail immediately, generating a delivery report event with the distinctive error code.

The full list of permanent errors depend on the broker version and will likely grow in the future.

Typical permanent broker errors are:

  • ERR_CORRUPT_MESSAGE
  • ERR_MSG_SIZE_TOO_LARGE - adjust client's or broker's message.max.bytes.
  • ERR_UNKNOWN_TOPIC_OR_PART - topic or partition does not exist, automatic topic creation is disabled on the broker or the application is specifying a partition that does not exist.
  • ERR_RECORD_LIST_TOO_LARGE
  • ERR_INVALID_REQUIRED_ACKS
  • ERR_TOPIC_AUTHORIZATION_FAILED
  • ERR_UNSUPPORTED_FOR_MESSAGE_FORMAT
  • ERR_CLUSTER_AUTHORIZATION_FAILED

Producer retries

The ProduceRequest itself is not retried, instead the messages are put back on the internal partition queue by an insert sort that maintains their original position (the message order is defined at the time a message is initially appended to a partition queue, i.e., after partitioning). A backoff time (retry.backoff.ms) is set on the retried messages which effectively blocks retry attempts until the backoff time has expired.

Reordering

As for all retries, if max.in.flight > 1 and retries > 0, retried messages may be produced out of order, since a sub-sequent message in a sub-sequent ProduceRequest may already be in-flight (and accepted by the broker) by the time the retry for the failing message is sent.

Using the Idempotent Producer prevents reordering even with max.in.flight > 1, see Idempotent Producer below for more information.

Idempotent Producer

librdkafka supports the idempotent producer which provides strict ordering and and exactly-once producer guarantees. The idempotent producer is enabled by setting the enable.idempotence configuration property to true, this will automatically adjust a number of other configuration properties to adhere to the idempotency requirements, see the documentation of enable.idempotence in CONFIGURATION.md for more information. Producer instantiation will fail if the user supplied an incompatible value for any of the automatically adjusted properties, e.g., it is an error to explicitly set acks=1 when enable.idempotence=true is set.

Guarantees

There are three types of guarantees that the idempotent producer can satisfy:

  • Exactly-once - a message is only written to the log once. Does NOT cover the exactly-once consumer case.
  • Ordering - a series of messages are written to the log in the order they were produced.
  • Gap-less - EXPERIMENTAL a series of messages are written once and in order without risk of skipping messages. The sequence of messages may be cut short and fail before all messages are written, but may not fail individual messages in the series. This guarantee is disabled by default, but may be enabled by setting enable.gapless.guarantee if individual message failure is a concern. Messages that fail due to exceeded timeout (message.timeout.ms), are permitted by the gap-less guarantee and may cause gaps in the message series without raising a fatal error. See Message timeout considerations below for more info. WARNING: This is an experimental property subject to change or removal.

All three guarantees are in effect when idempotence is enabled, only gap-less may be disabled individually.

Ordering and message sequence numbers

librdkafka maintains the original produce() ordering per-partition for all messages produced, using an internal per-partition 64-bit counter called the msgid which starts at 1. This msgid allows messages to be re-inserted in the partition message queue in the original order in the case of retries.

The Idempotent Producer functionality in the Kafka protocol also has a per-message sequence number, which is a signed 32-bit wrapping counter that is reset each time the Producer's ID (PID) or Epoch changes.

The librdkafka msgid is used, along with a base msgid value stored at the time the PID/Epoch was bumped, to calculate the Kafka protocol's message sequence number.

With Idempotent Producer enabled there is no risk of reordering despite max.in.flight > 1 (capped at 5).

Note: "MsgId" in log messages refer to the librdkafka msgid, while "seq" refers to the protocol message sequence, "baseseq" is the seq of the first message in a batch. MsgId starts at 1, while message seqs start at 0.

The producer statistics also maintain two metrics for tracking the next expected response sequence:

  • next_ack_seq - the next sequence to expect an acknowledgement for, which is the last successfully produced MessageSet's last sequence + 1.
  • next_err_seq - the next sequence to expect an error for, which is typically the same as next_ack_seq until an error occurs, in which case the next_ack_seq can't be incremented (since no messages were acked on error). next_err_seq is used to properly handle sub-sequent errors due to a failing first request.

Note: Both are exposed in partition statistics.

Partitioner considerations

Strict ordering is guaranteed on a per partition basis.

An application utilizing the idempotent producer should not mix producing to explicit partitions with partitioner-based partitions since messages produced for the latter are queued separately until a topic's partition count is known, which would insert these messages after the partition-explicit messages regardless of produce order.

Message timeout considerations

If messages time out (due to message.timeout.ms) while in the producer queue there will be gaps in the series of produced messages.

E.g., Messages 1,2,3,4,5 are produced by the application. While messages 2,3,4 are transmitted to the broker the connection to the broker goes down. While the broker is down the message timeout expires for message 2 and 3. As the connection comes back up messages 4, 5 are transmitted to the broker, resulting in a final written message sequence of 1, 4, 5.

The producer gracefully handles this case by draining the in-flight requests for a given partition when one or more of its queued (not transmitted) messages are timed out. When all requests are drained the Epoch is bumped and the base sequence number is reset to the first message in the queue, effectively skipping the timed out messages as if they had never existed from the broker's point of view. The message status for timed out queued messages will be RD_KAFKA_MSG_STATUS_NOT_PERSISTED.

If messages time out while in-flight to the broker (also due to message.timeout.ms), the protocol request will fail, the broker connection will be closed by the client, and the timed out messages will be removed from the producer queue. In this case the in-flight messages may be written to the topic log by the broker, even though a delivery report with error ERR__MSG_TIMED_OUT will be raised, since the producer timed out the request before getting an acknowledgement back from the broker. The message status for timed out in-flight messages will be RD_KAFKA_MSG_STATUS_POSSIBLY_PERSISTED, indicating that the producer does not know if the messages were written and acked by the broker, or dropped in-flight.

An application may inspect the message status by calling rd_kafka_message_status() on the message in the delivery report callback, to see if the message was (possibly) persisted (written to the topic log) by the broker or not.

Despite the graceful handling of timeouts, we recommend to use a large message.timeout.ms to minimize the risk of timeouts.

Warning: enable.gapless.guarantee does not apply to timed-out messages.

Note: delivery.timeout.ms is an alias for message.timeout.ms.

Leader change

There are corner cases where an Idempotent Producer has outstanding ProduceRequests in-flight to the previous leader while a new leader is elected.

A leader change is typically triggered by the original leader failing or terminating, which has the risk of also failing (some of) the in-flight ProduceRequests to that broker. To recover the producer to a consistent state it will not send any ProduceRequests for these partitions to the new leader broker until all responses for any outstanding ProduceRequests to the previous partition leader has been received, or these requests have timed out. This drain may take up to min(socket.timeout.ms, message.timeout.ms). If the connection to the previous broker goes down the outstanding requests are failed immediately.

Error handling

Background: The error handling for the Idempotent Producer, as initially proposed in the EOS design document, missed some corner cases which are now being addressed in KIP-360. There were some intermediate fixes and workarounds prior to KIP-360 that proved to be incomplete and made the error handling in the client overly complex. With the benefit of hindsight the librdkafka implementation will attempt to provide correctness from the lessons learned in the Java client and provide stricter and less complex error handling.

The follow sections describe librdkafka's handling of the Idempotent Producer specific errors that may be returned by the broker.

RD_KAFKA_RESP_ERR_OUT_OF_ORDER_SEQUENCE_NUMBER

This error is returned by the broker when the sequence number in the ProduceRequest is larger than the expected next sequence for the given PID+Epoch+Partition (last BaseSeq + msgcount + 1). Note: sequence 0 is always accepted.

If the failed request is the head-of-line (next expected sequence to be acked) it indicates desynchronization between the client and broker: the client thinks the sequence number is correct but the broker disagrees. There is no way for the client to recover from this scenario without risking message loss or duplication, and it is not safe for the application to manually retry messages. A fatal error (RD_KAFKA_RESP_ERR_OUT_OF_ORDER_SEQUENCE_NUMBER) is raised.

When the request is not head-of-line the previous request failed (for any reason), which means the messages in the current request can be retried after waiting for all outstanding requests for this partition to drain and then reset the Producer ID and start over.

Java Producer behaviour: Fail the batch, reset the pid, and then continue producing (and retrying sub-sequent) messages. This will lead to gaps in the message series.

RD_KAFKA_RESP_ERR_DUPLICATE_SEQUENCE_NUMBER

Returned by broker when the request's base sequence number is less than the expected sequence number (which is the last written sequence + msgcount). Note: sequence 0 is always accepted.

This error is typically benign and occurs upon retrying a previously successful send that was not acknowledged.

The messages will be considered successfully produced but will have neither timestamp or offset set.

Java Producer behaviour: Treats the message as successfully delivered.

RD_KAFKA_RESP_ERR_UNKNOWN_PRODUCER_ID

Returned by broker when the PID+Epoch is unknown, which may occur when the PID's state has expired (due to topic retention, DeleteRercords, or compaction).

The Java producer added quite a bit of error handling for this case, extending the ProduceRequest protocol to return the logStartOffset to give the producer a chance to differentiate between an actual UNKNOWN_PRODUCER_ID or topic retention having deleted the last message for this producer (effectively voiding the Producer ID cache). This workaround proved to be error prone (see explanation in KIP-360) when the partition leader changed.

KIP-360 suggests removing this error checking in favour of failing fast, librdkafka follows suite.

If the response is for the first ProduceRequest in-flight and there are no messages waiting to be retried nor any ProduceRequests unaccounted for, then the error is ignored and the epoch is incremented, this is likely to happen for an idle producer who's last written message has been deleted from the log, and thus its PID state. Otherwise the producer raises a fatal error (RD_KAFKA_RESP_ERR_UNKNOWN_PRODUCER_ID) since the delivery guarantees can't be satisfied.

Java Producer behaviour: Retries the send in some cases (but KIP-360 will change this). Not a fatal error in any case.

Standard errors

All the standard Produce errors are handled in the usual way, permanent errors will fail the messages in the batch, while temporary errors will be retried (if retry count permits).

If a permanent error is returned for a batch in a series of in-flight batches, the sub-sequent batches will fail with RD_KAFKA_RESP_ERR_OUT_OF_ORDER_SEQUENCE_NUMBER since the sequence number of the failed batched was never written to the topic log and next expected sequence thus not incremented on the broker.

A fatal error (RD_KAFKA_RESP_ERR__GAPLESS_GUARANTEE) is raised to satisfy the gap-less guarantee (if enable.gapless.guarantee is set) by failing all queued messages.

Message persistence status

To help the application decide what to do in these error cases, a new per-message API is introduced, rd_kafka_message_status(), which returns one of the following values:

  • RD_KAFKA_MSG_STATUS_NOT_PERSISTED - the message has never been transmitted to the broker, or failed with an error indicating it was not written to the log. Application retry will risk ordering, but not duplication.
  • RD_KAFKA_MSG_STATUS_POSSIBLY_PERSISTED - the message was transmitted to the broker, but no acknowledgement was received. Application retry will risk ordering and duplication.
  • RD_KAFKA_MSG_STATUS_PERSISTED - the message was written to the log by the broker and fully acknowledged. No reason for application to retry.

This method should be called by the application on delivery report error.

Transactional Producer

Error handling

Using the transactional producer simplifies error handling compared to the standard or idempotent producer, a transactional application will only need to care about these different types of errors:

  • Retriable errors - the operation failed due to temporary problems, such as network timeouts, the operation may be safely retried. Use rd_kafka_error_is_retriable() to distinguish this case.
  • Abortable errors - if any of the transactional APIs return a non-fatal error code the current transaction has failed and the application must call rd_kafka_abort_transaction(), rewind its input to the point before the current transaction started, and attempt a new transaction by calling rd_kafka_begin_transaction(), etc. Use rd_kafka_error_txn_requires_abort() to distinguish this case.
  • Fatal errors - the application must cease operations and destroy the producer instance. Use rd_kafka_error_is_fatal() to distinguish this case.
  • For all other errors returned from the transactional API: the current recommendation is to treat any error that has neither retriable, abortable, or fatal set, as a fatal error.

While the application should log the actual fatal or abortable errors, there is no need for the application to handle the underlying errors specifically.

Old producer fencing

If a new transactional producer instance is started with the same transactional.id, any previous still running producer instance will be fenced off at the next produce, commit or abort attempt, by raising a fatal error with the error code set to RD_KAFKA_RESP_ERR__FENCED.

Configuration considerations

To make sure messages time out (in case of connectivity problems, etc) within the transaction, the message.timeout.ms configuration property must be set lower than the transaction.timeout.ms, this is enforced when creating the producer instance. If message.timeout.ms is not explicitly configured it will be adjusted automatically.

Exactly Once Semantics (EOS) and transactions

librdkafka supports Exactly One Semantics (EOS) as defined in KIP-98. For more on the use of transactions, see Transactions in Apache Kafka.

See examples/transactions.c for an example transactional EOS application.

Warning If the broker version is older than Apache Kafka 2.5.0 then one transactional producer instance per consumed input partition is required. For 2.5.0 and later a single producer instance may be used regardless of the number of input partitions. See KIP-447 for more information.

Usage

Documentation

The librdkafka API is documented in the rdkafka.h header file, the configuration properties are documented in CONFIGURATION.md

Initialization

The application needs to instantiate a top-level object rd_kafka_t which is the base container, providing global configuration and shared state. It is created by calling rd_kafka_new().

It also needs to instantiate one or more topics (rd_kafka_topic_t) to be used for producing to or consuming from. The topic object holds topic-specific configuration and will be internally populated with a mapping of all available partitions and their leader brokers. It is created by calling rd_kafka_topic_new().

Both rd_kafka_t and rd_kafka_topic_t comes with a configuration API which is optional. Not using the API will cause librdkafka to use its default values which are documented in CONFIGURATION.md.

Note: An application may create multiple rd_kafka_t objects and they share no state.

Note: An rd_kafka_topic_t object may only be used with the rd_kafka_t object it was created from.

Configuration

To ease integration with the official Apache Kafka software and lower the learning curve, librdkafka implements identical configuration properties as found in the official clients of Apache Kafka.

Configuration is applied prior to object creation using the rd_kafka_conf_set() and rd_kafka_topic_conf_set() APIs.

Note: The rd_kafka.._conf_t objects are not reusable after they have been passed to rd_kafka.._new(). The application does not need to free any config resources after a rd_kafka.._new() call.

Example

    rd_kafka_conf_t *conf;
    rd_kafka_conf_res_t res;
    rd_kafka_t *rk;
    char errstr[512];

    conf = rd_kafka_conf_new();

    res = rd_kafka_conf_set(conf, "compression.codec", "snappy",
                            errstr, sizeof(errstr));
    if (res != RD_KAFKA_CONF_OK)
        fail("%s\n", errstr);

    res = rd_kafka_conf_set(conf, "batch.num.messages", "100",
                            errstr, sizeof(errstr));
    if (res != RD_KAFKA_CONF_OK)
        fail("%s\n", errstr);

    rk = rd_kafka_new(RD_KAFKA_PRODUCER, conf, errstr, sizeof(errstr));
    if (!rk) {
        rd_kafka_conf_destroy(rk);
        fail("Failed to create producer: %s\n", errstr);
    }

    /* Note: librdkafka takes ownership of the conf object on success */

Configuration properties may be set in any order (except for interceptors) and may be overwritten before being passed to rd_kafka_new(). rd_kafka_new() will verify that the passed configuration is consistent and will fail and return an error if incompatible configuration properties are detected. It will also emit log warnings for deprecated and problematic configuration properties.

Termination

librdkafka is asynchronous in its nature and performs most operation in its background threads.

Calling the librdkafka handle destructor tells the librdkafka background threads to finalize their work, close network connections, clean up, etc, and may thus take some time. The destructor (rd_kafka_destroy()) will block until all background threads have terminated.

If the destructor blocks indefinitely it typically means there is an outstanding object reference, such as a message or topic object, that was not destroyed prior to destroying the client handle.

All objects except for the handle (C: rd_kafka_t, C++: Consumer,KafkaConsumer,Producer), such as topic objects, messages, topic_partition_t, TopicPartition, events, etc, MUST be destroyed/deleted prior to destroying or closing the handle.

For C, make sure the following objects are destroyed prior to calling rd_kafka_consumer_close() and rd_kafka_destroy():

  • rd_kafka_message_t
  • rd_kafka_topic_t
  • rd_kafka_topic_partition_t
  • rd_kafka_topic_partition_list_t
  • rd_kafka_event_t
  • rd_kafka_queue_t

For C++ make sure the following objects are deleted prior to calling KafkaConsumer::close() and delete on the Consumer, KafkaConsumer or Producer handle:

  • Message
  • Topic
  • TopicPartition
  • Event
  • Queue

High-level KafkaConsumer

Proper termination sequence for the high-level KafkaConsumer is:

     /* 1) Leave the consumer group, commit final offsets, etc. */
     rd_kafka_consumer_close(rk);

     /* 2) Destroy handle object */
     rd_kafka_destroy(rk);

NOTE: There is no need to unsubscribe prior to calling rd_kafka_consumer_close().

NOTE: Any topic objects created must be destroyed prior to rd_kafka_destroy()

Effects of not doing the above, for:

  1. Final offsets are not committed and the consumer will not actively leave the group, it will be kicked out of the group after the session.timeout.ms expires. It is okay to omit the rd_kafka_consumer_close() call in case the application does not want to wait for the blocking close call.
  2. librdkafka will continue to operate on the handle. Actual memory leaks.

Producer

The proper termination sequence for Producers is:

     /* 1) Make sure all outstanding requests are transmitted and handled. */
     rd_kafka_flush(rk, 60*1000); /* One minute timeout */

     /* 2) Destroy the topic and handle objects */
     rd_kafka_topic_destroy(rkt);  /* Repeat for all topic objects held */
     rd_kafka_destroy(rk);

Effects of not doing the above, for:

  1. Messages in-queue or in-flight will be dropped.
  2. librdkafka will continue to operate on the handle. Actual memory leaks.

Admin API client

Unlike the Java Admin client, the Admin APIs in librdkafka are available on any type of client instance and can be used in combination with the client type's main functionality, e.g., it is perfectly fine to call CreateTopics() in your running producer, or DeleteRecords() in your consumer.

If you need a client instance to only perform Admin API operations the recommendation is to create a producer instance since it requires less configuration (no group.id) than the consumer and is generally more cost efficient. We do recommend that you set allow.auto.create.topics=false to avoid topic metadata lookups to unexpectedly have the broker create topics.

Speeding up termination

To speed up the termination of librdkafka an application can set a termination signal that will be used internally by librdkafka to quickly cancel any outstanding I/O waits. Make sure you block this signal in your application.

   char tmp[16];
   snprintf(tmp, sizeof(tmp), "%i", SIGIO);  /* Or whatever signal you decide */
   rd_kafka_conf_set(rk_conf, "internal.termination.signal", tmp, errstr, sizeof(errstr));

Threads and callbacks

librdkafka uses multiple threads internally to fully utilize modern hardware. The API is completely thread-safe and the calling application may call any of the API functions from any of its own threads at any time.

A poll-based API is used to provide signaling back to the application, the application should call rd_kafka_poll() at regular intervals. The poll API will call the following configured callbacks (optional):

  • dr_msg_cb - Message delivery report callback - signals that a message has been delivered or failed delivery, allowing the application to take action and to release any application resources used in the message.
  • error_cb - Error callback - signals an error. These errors are usually of an informational nature, i.e., failure to connect to a broker, and the application usually does not need to take any action. The type of error is passed as a rd_kafka_resp_err_t enum value, including both remote broker errors as well as local failures. An application typically does not have to perform any action when an error is raised through the error callback, the client will automatically try to recover from all errors, given that the client and cluster is correctly configured. In some specific cases a fatal error may occur which will render the client more or less inoperable for further use: if the error code in the error callback is set to RD_KAFKA_RESP_ERR__FATAL the application should retrieve the underlying fatal error and reason using the rd_kafka_fatal_error() call, and then begin terminating the instance. The Event API's EVENT_ERROR has a rd_kafka_event_error_is_fatal() function, and the C++ EventCb has a fatal() method, to help the application determine if an error is fatal or not.
  • stats_cb - Statistics callback - triggered if statistics.interval.ms is configured to a non-zero value, emitting metrics and internal state in JSON format, see [STATISTICS.md].
  • throttle_cb - Throttle callback - triggered whenever a broker has throttled (delayed) a request.

These callbacks will also be triggered by rd_kafka_flush(), rd_kafka_consumer_poll(), and any other functions that serve queues.

Optional callbacks not triggered by poll, these may be called spontaneously from any thread at any time:

  • log_cb - Logging callback - allows the application to output log messages generated by librdkafka.
  • partitioner - Partitioner callback - application provided message partitioner. The partitioner may be called in any thread at any time, it may be called multiple times for the same key. Partitioner function contraints:
    • MUST NOT call any rd_kafka_*() functions
    • MUST NOT block or execute for prolonged periods of time.
    • MUST return a value between 0 and partition_cnt-1, or the special RD_KAFKA_PARTITION_UA value if partitioning could not be performed.

Brokers

On initialization, librdkafka only needs a partial list of brokers (at least one), called the bootstrap brokers. The client will connect to the bootstrap brokers specified by the bootstrap.servers configuration property and query cluster Metadata information which contains the full list of brokers, topic, partitions and their leaders in the Kafka cluster.

Broker names are specified as host[:port] where the port is optional (default 9092) and the host is either a resolvable hostname or an IPv4 or IPv6 address. If host resolves to multiple addresses librdkafka will round-robin the addresses for each connection attempt. A DNS record containing all broker address can thus be used to provide a reliable bootstrap broker.

SSL

If the client is to connect to a broker's SSL endpoints/listeners the client needs to be configured with security.protocol=SSL for just SSL transport or security.protocol=SASL_SSL for SASL authentication and SSL transport. The client will try to verify the broker's certificate by checking the CA root certificates, if the broker's certificate can't be verified the connection is closed (and retried). This is to protect the client from connecting to rogue brokers.

The CA root certificate defaults are system specific:

  • On Linux, Mac OSX, and other Unix-like system the OpenSSL default CA path will be used, also called the OPENSSLDIR, which is typically /etc/ssl/certs (on Linux, typcially in the ca-certificates package) and /usr/local/etc/openssl on Mac OSX (Homebrew).
  • On Windows the Root certificate store is used, unless ssl.ca.certificate.stores is configured in which case certificates are read from the specified stores.
  • If OpenSSL is linked statically, librdkafka will set the default CA location to the first of a series of probed paths (see below).

If the system-provided default CA root certificates are not sufficient to verify the broker's certificate, such as when a self-signed certificate or a local CA authority is used, the CA certificate must be specified explicitly so that the client can find it. This can be done either by providing a PEM file (e.g., cacert.pem) as the ssl.ca.location configuration property, or by passing an in-memory PEM, X.509/DER or PKCS#12 certificate to rd_kafka_conf_set_ssl_cert().

It is also possible to disable broker certificate verification completely by setting enable.ssl.certificate.verification=false, but this is not recommended since it allows for rogue brokers and man-in-the-middle attacks, and should only be used for testing and troubleshooting purposes.

CA location probe paths (see rdkafka_ssl.c for full list) used when OpenSSL is statically linked:

"/etc/pki/tls/certs/ca-bundle.crt",
"/etc/ssl/certs/ca-bundle.crt",
"/etc/pki/tls/certs/ca-bundle.trust.crt",
"/etc/pki/ca-trust/extracted/pem/tls-ca-bundle.pem",
"/etc/ssl/ca-bundle.pem",
"/etc/pki/tls/cacert.pem",
"/etc/ssl/cert.pem",
"/etc/ssl/cacert.pem",
"/etc/certs/ca-certificates.crt",
"/etc/ssl/certs/ca-certificates.crt",
"/etc/ssl/certs",
"/usr/local/etc/ssl/cert.pem",
"/usr/local/etc/ssl/cacert.pem",
"/usr/local/etc/ssl/certs/cert.pem",
"/usr/local/etc/ssl/certs/cacert.pem",
etc..

On Windows the Root certificate store is read by default, but any number of certificate stores can be read by setting the ssl.ca.certificate.stores configuration property to a comma-separated list of certificate store names. The predefined system store names are:

  • MY - User certificates
  • Root - System CA certificates (default)
  • CA - Intermediate CA certificates
  • Trust - Trusted publishers

For example, to read both intermediate and root CAs, set ssl.ca.certificate.stores=CA,Root.

Sparse connections

The client will only connect to brokers it needs to communicate with, and only when necessary.

Examples of needed broker connections are:

  • leaders for partitions being consumed from
  • leaders for partitions being produced to
  • consumer group coordinator broker
  • cluster controller for Admin API operations
Random broker selection

When there is no broker connection and a connection to any broker is needed, such as on startup to retrieve metadata, the client randomly selects a broker from its list of brokers, which includes both the configured bootstrap brokers (including brokers manually added with rd_kafka_brokers_add()), as well as the brokers discovered from cluster metadata. Brokers with no prior connection attempt are tried first.

If there is already an available broker connection to any broker it is used, rather than connecting to a new one.

The random broker selection and connection scheduling is triggered when:

  • bootstrap servers are configured (rd_kafka_new())
  • brokers are manually added (rd_kafka_brokers_add()).
  • a consumer group coordinator needs to be found.
  • acquiring a ProducerID for the Idempotent Producer.
  • cluster or topic metadata is being refreshed.

A single connection attempt will be performed, and the broker will return to an idle INIT state on failure to connect.

The random broker selection is rate-limited to: 10 < reconnect.backoff.ms/2 < 1000 milliseconds.

Note: The broker connection will be maintained until it is closed by the broker (idle connection reaper).

Persistent broker connections

While the random broker selection is useful for one-off queries, there is need for the client to maintain persistent connections to certain brokers:

  • Consumer: the group coordinator.
  • Consumer: partition leader for topics being fetched from.
  • Producer: partition leader for topics being produced to.

These dependencies are discovered and maintained automatically, marking matching brokers as persistent, which will make the client maintain connections to these brokers at all times, reconnecting as necessary.

Connection close

A broker connection may be closed by the broker, intermediary network gear, due to network errors, timeouts, etc. When a broker connection is closed, librdkafka will back off the next reconnect attempt (to the given broker) for reconnect.backoff.ms -25% to +50% jitter, this value is increased exponentially for each connect attempt until reconnect.backoff.max.ms is reached, at which time the value is reset to reconnect.backoff.ms.

The broker will disconnect clients that have not sent any protocol requests within connections.max.idle.ms (broker configuration propertion, defaults to 10 minutes), but there is no fool proof way for the client to know that it was a deliberate close by the broker and not an error. To avoid logging these deliberate idle disconnects as errors the client employs some logic to try to classify a disconnect as an idle disconnect if no requests have been sent in the last socket.timeout.ms or there are no outstanding, or queued, requests waiting to be sent. In this case the standard "Disconnect" error log is silenced (will only be seen with debug enabled).

Otherwise, if a connection is closed while there are requests in-flight the logging level will be LOG_WARNING (4), else LOG_INFO (6).

log.connection.close=false may be used to silence all disconnect logs, but it is recommended to instead rely on the above heuristics.

Fetch From Follower

librdkafka supports consuming messages from follower replicas (KIP-392). This is enabled by setting the client.rack configuration property which corresponds to broker.rack on the broker. The actual assignment of consumers to replicas is determined by the configured replica.selector.class on the broker.

Logging

Debug contexts

Extensive debugging of librdkafka can be enabled by setting the debug configuration property to a CSV string of debug contexts:

Debug context Type Description
generic * General client instance level debugging. Includes initialization and termination debugging.
broker * Broker and connection state debugging.
topic * Topic and partition state debugging. Includes leader changes.
metadata * Cluster and topic metadata retrieval debugging.
feature * Kafka protocol feature support as negotiated with the broker.
queue producer Message queue debugging.
msg * Message debugging. Includes information about batching, compression, sizes, etc.
protocol * Kafka protocol request/response debugging. Includes latency (rtt) printouts.
cgrp consumer Low-level consumer group state debugging.
security * Security and authentication debugging.
fetch consumer Consumer message fetch debugging. Includes decision when and why messages are fetched.
interceptor * Interceptor interface debugging.
plugin * Plugin loading debugging.
consumer consumer High-level consumer debugging.
admin admin Admin API debugging.
eos producer Idempotent Producer debugging.
mock * Mock cluster functionality debugging.
assignor consumer Detailed consumer group partition assignor debugging.
conf * Display set configuration properties on startup.
all * All of the above.

Suggested debugging settings for troubleshooting:

Problem space Type Debug setting
Producer not delivering messages to broker producer broker,topic,msg
Consumer not fetching messages consumer Start with consumer, or use cgrp,fetch for detailed information.
Consumer starts reading at unexpected offset consumer consumer or cgrp,fetch
Authentication or connectivity issues * broker,auth
Protocol handling or latency * broker,protocol
Topic leader and state * topic,metadata

Feature discovery

Apache Kafka broker version 0.10.0 added support for the ApiVersionRequest API which allows a client to query a broker for its range of supported API versions.

librdkafka supports this functionality and will query each broker on connect for this information (if api.version.request=true) and use it to enable or disable various protocol features, such as MessageVersion 1 (timestamps), KafkaConsumer, etc.

If the broker fails to respond to the ApiVersionRequest librdkafka will assume the broker is too old to support the API and fall back to an older broker version's API. These fallback versions are hardcoded in librdkafka and is controlled by the broker.version.fallback configuration property.

Producer API

After setting up the rd_kafka_t object with type RD_KAFKA_PRODUCER and one or more rd_kafka_topic_t objects librdkafka is ready for accepting messages to be produced and sent to brokers.

The rd_kafka_produce() function takes the following arguments:

  • rkt - the topic to produce to, previously created with rd_kafka_topic_new()

  • partition - partition to produce to. If this is set to RD_KAFKA_PARTITION_UA (UnAssigned) then the configured partitioner function will be used to select a target partition.

  • msgflags - 0, or one of:

    • RD_KAFKA_MSG_F_COPY - librdkafka will immediately make a copy of the payload. Use this when the payload is in non-persistent memory, such as the stack.
    • RD_KAFKA_MSG_F_FREE - let librdkafka free the payload using free(3) when it is done with it.

    These two flags are mutually exclusive and neither need to be set in which case the payload is neither copied nor freed by librdkafka.

    If RD_KAFKA_MSG_F_COPY flag is not set no data copying will be performed and librdkafka will hold on the payload pointer until the message has been delivered or fails. The delivery report callback will be called when librdkafka is done with the message to let the application regain ownership of the payload memory. The application must not free the payload in the delivery report callback if RD_KAFKA_MSG_F_FREE is set.

  • payload,len - the message payload

  • key,keylen - an optional message key which can be used for partitioning. It will be passed to the topic partitioner callback, if any, and will be attached to the message when sending to the broker.

  • msg_opaque - an optional application-provided per-message opaque pointer that will be provided in the message delivery callback to let the application reference a specific message.

rd_kafka_produce() is a non-blocking API, it will enqueue the message on an internal queue and return immediately. If the number of queued messages would exceed the queue.buffering.max.messages configuration property then rd_kafka_produce() returns -1 and sets errno to ENOBUFS and last_error to RD_KAFKA_RESP_ERR__QUEUE_FULL, thus providing a backpressure mechanism.

rd_kafka_producev() provides an alternative produce API that does not require a topic rkt object and also provides support for extended message fields, such as timestamp and headers.

Note: See examples/rdkafka_performance.c for a producer implementation.

Simple Consumer API (legacy)

NOTE: For the high-level KafkaConsumer interface see rd_kafka_subscribe (rdkafka.h) or KafkaConsumer (rdkafkacpp.h)

The consumer API is a bit more stateful than the producer API. After creating rd_kafka_t with type RD_KAFKA_CONSUMER and rd_kafka_topic_t instances the application must also start the consumer for a given partition by calling rd_kafka_consume_start().

rd_kafka_consume_start() arguments:

  • rkt - the topic to start consuming from, previously created with rd_kafka_topic_new().
  • partition - partition to consume from.
  • offset - message offset to start consuming from. This may either be an absolute message offset or one of the three special offsets: RD_KAFKA_OFFSET_BEGINNING to start consuming from the beginning of the partition's queue (oldest message), or RD_KAFKA_OFFSET_END to start consuming at the next message to be produced to the partition, or RD_KAFKA_OFFSET_STORED to use the offset store.

After a topic+partition consumer has been started librdkafka will attempt to keep queued.min.messages messages in the local queue by repeatedly fetching batches of messages from the broker. librdkafka will fetch all consumed partitions for which that broker is a leader, through a single request.

This local message queue is then served to the application through three different consume APIs:

  • rd_kafka_consume() - consumes a single message
  • rd_kafka_consume_batch() - consumes one or more messages
  • rd_kafka_consume_callback() - consumes all messages in the local queue and calls a callback function for each one.

These three APIs are listed above the ascending order of performance, rd_kafka_consume() being the slowest and rd_kafka_consume_callback() being the fastest. The different consume variants are provided to cater for different application needs.

A consumed message, as provided or returned by each of the consume functions, is represented by the rd_kafka_message_t type.

rd_kafka_message_t members:

  • err - Error signaling back to the application. If this field is non-zero the payload field should be considered an error message and err is an error code (rd_kafka_resp_err_t). If err is zero then the message is a proper fetched message and payload et.al contains message payload data.
  • rkt,partition - Topic and partition for this message or error.
  • payload,len - Message payload data or error message (err!=0).
  • key,key_len - Optional message key as specified by the producer
  • offset - Message offset

Both the payload and key memory, as well as the message as a whole, is owned by librdkafka and must not be used after an rd_kafka_message_destroy() call. librdkafka will share the same messageset receive buffer memory for all message payloads of that messageset to avoid excessive copying which means that if the application decides to hang on to a single rd_kafka_message_t it will hinder the backing memory to be released for all other messages from the same messageset.

When the application is done consuming messages from a topic+partition it should call rd_kafka_consume_stop() to stop the consumer. This will also purge any messages currently in the local queue.

Note: See examples/rdkafka_performance.c for a consumer implementation.

Offset management

Broker based offset management is available for broker version >= 0.9.0 in conjunction with using the high-level KafkaConsumer interface (see rdkafka.h or rdkafkacpp.h)

Offset management is also available through a deprecated local offset file, where the offset is periodically written to a local file for each topic+partition according to the following topic configuration properties:

  • enable.auto.commit
  • auto.commit.interval.ms
  • offset.store.path
  • offset.store.sync.interval.ms

The legacy auto.commit.enable topic configuration property is only to be used with the legacy low-level consumer. Use enable.auto.commit with the modern KafkaConsumer.

Auto offset commit

The consumer will automatically commit offsets every auto.commit.interval.ms when enable.auto.commit is enabled (default).

Offsets to be committed are kept in a local in-memory offset store, this offset store is updated by consumer_poll() (et.al) to store the offset of the last message passed to the application (per topic+partition).

At-least-once processing

Since auto commits are performed in a background thread this may result in the offset for the latest message being committed before the application has finished processing the message. If the application was to crash or exit prior to finishing processing, and the offset had been auto committed, the next incarnation of the consumer application would start at the next message, effectively missing the message that was processed when the application crashed. To avoid this scenario the application can disable the automatic offset store by setting enable.auto.offset.store to false and manually storing offsets after processing by calling rd_kafka_offsets_store(). This gives an application fine-grained control on when a message is eligible for committing without having to perform the commit itself. enable.auto.commit should be set to true when using manual offset storing. The latest stored offset will be automatically committed every auto.commit.interval.ms.

Note: Only greater offsets are committed, e.g., if the latest committed offset was 10 and the application performs an offsets_store() with offset 9, that offset will not be committed.

Auto offset reset

The consumer will by default try to acquire the last committed offsets for each topic+partition it is assigned using its configured group.id. If there is no committed offset available, or the consumer is unable to fetch the committed offsets, the policy of auto.offset.reset will kick in. This configuration property may be set to one the following values:

  • earliest - start consuming the earliest message of the partition.
  • latest - start consuming the next message to be produced to the partition.
  • error - don't start consuming but isntead raise a consumer error with error-code RD_KAFKA_RESP_ERR__AUTO_OFFSET_RESET for the topic+partition. This allows the application to decide what to do in case there is no committed start offset.

Consumer groups

Broker based consumer groups (requires Apache Kafka broker >=0.9) are supported, see KafkaConsumer in rdkafka.h or rdkafkacpp.h

The following diagram visualizes the high-level balanced consumer group state flow and synchronization between the application, librdkafka consumer, group coordinator, and partition leader(s).

Consumer group state diagram

Static consumer groups

By default Kafka consumers are rebalanced each time a new consumer joins the group or an existing member leaves. This is what is known as a dynamic membership. Apache Kafka >= 2.3.0 introduces static membership. Unlike dynamic membership, static members can leave and rejoin a group within the session.timeout.ms without triggering a rebalance, retaining their existing partitions assignment.

To enable static group membership configure each consumer instance in the group with a unique group.instance.id.

Consumers with group.instance.id set will not send a leave group request on close - session timeout, change of subscription, or a new group member joining the group, are the only mechanisms that will trigger a group rebalance for static consumer groups.

If a new consumer joins the group with same group.instance.id as an existing consumer, the existing consumer will be fenced and raise a fatal error. The fatal error is propagated as a consumer error with error code RD_KAFKA_RESP_ERR__FATAL, use rd_kafka_fatal_error() to retrieve the original fatal error code and reason.

To read more about static group membership, see KIP-345.

Topics

Unknown or unauthorized topics

If a consumer application subscribes to non-existent or unauthorized topics a consumer error will be propagated for each unavailable topic with the error code set to either RD_KAFKA_RESP_ERR_UNKNOWN_TOPIC_OR_PART or a broker-specific error code, such as RD_KAFKA_RESP_ERR_TOPIC_AUTHORIZATION_FAILED.

As the topic metadata is refreshed every topic.metadata.refresh.interval.ms the unavailable topics are re-checked for availability, but the same error will not be raised again for the same topic.

If a consumer has Describe (ACL) permissions for a topic but not Read it will be able to join a consumer group and start consuming the topic, but the Fetch requests to retrieve messages from the broker will fail with RD_KAFKA_RESP_ERR_TOPIC_AUTHORIZATION_FAILED. This error will be raised to the application once per partition and assign()/seek() and the fetcher will back off the next fetch 10 times longer than the fetch.error.backoff.ms (but at least 1 second). It is recommended that the application takes appropriate action when this occurs, for instance adjusting its subscription or assignment to exclude the unauthorized topic.

Topic metadata propagation for newly created topics

Due to the asynchronous nature of topic creation in Apache Kafka it may take some time for a newly created topic to be known by all brokers in the cluster. If a client tries to use a topic after topic creation but before the topic has been fully propagated in the cluster it will seem as if the topic does not exist which would raise RD_KAFKA_RESP_ERR__UNKNOWN_TOPIC (et.al) errors to the application. To avoid these temporary errors being raised, the client will not flag a topic as non-existent until a propagation time has elapsed, this propagation defaults to 30 seconds and can be configured with topic.metadata.propagation.max.ms. The per-topic max propagation time starts ticking as soon as the topic is referenced (e.g., by produce()).

If messages are produced to unknown topics during the propagation time, the messages will be queued for later delivery to the broker when the topic metadata has propagated. Should the topic propagation time expire without the topic being seen the produced messages will fail with RD_KAFKA_RESP_ERR__UNKNOWN_TOPIC.

Note: The propagation time will not take affect if a topic is known to the client and then deleted, in this case the topic will immediately be marked as non-existent and remain non-existent until a topic metadata refresh sees the topic again (after the topic has been re-created).

Topic auto creation

Topic auto creation is supported by librdkafka, if a non-existent topic is referenced by the client (by produce to, or consuming from, the topic, etc) the broker will automatically create the topic (with default partition counts and replication factor) if the broker configuration property auto.create.topics.enable=true is set.

Note: A topic that is undergoing automatic creation may be reported as unavailable, with e.g., RD_KAFKA_RESP_ERR_UNKNOWN_TOPIC_OR_PART, during the time the topic is being created and partition leaders are elected.

While topic auto creation may be useful for producer applications, it is not particularily valuable for consumer applications since even if the topic to consume is auto created there is nothing writing messages to the topic. To avoid consumers automatically creating topics the allow.auto.create.topics consumer configuration property is set to false by default, preventing the consumer to trigger automatic topic creation on the broker. This requires broker version v0.11.0.0 or later. The allow.auto.create.topics property may be set to true to allow auto topic creation, which also requires auto.create.topics.enable=true to be configured on the broker.

Metadata

< 0.9.3

Previous to the 0.9.3 release librdkafka's metadata handling was chatty and excessive, which usually isn't a problem in small to medium-sized clusters, but in large clusters with a large amount of librdkafka clients the metadata requests could hog broker CPU and bandwidth.

> 0.9.3

The remaining Metadata sections describe the current behaviour.

Note: "Known topics" in the following section means topics for locally created rd_kafka_topic_t objects.

Query reasons

There are four reasons to query metadata:

  • brokers - update/populate cluster broker list, so the client can find and connect to any new brokers added.

  • specific topic - find leader or partition count for specific topic

  • known topics - same, but for all locally known topics.

  • all topics - get topic names for consumer group wildcard subscription matching

The above list is sorted so that the sub-sequent entries contain the information above, e.g., 'known topics' contains enough information to also satisfy 'specific topic' and 'brokers'.

Caching strategy

The prevalent cache timeout is metadata.max.age.ms, any cached entry will remain authoritative for this long or until a relevant broker error is returned.

  • brokers - eternally cached, the broker list is additative.

  • topics - cached for metadata.max.age.ms

Fatal errors

If an unrecoverable error occurs, a fatal error is triggered in one or more of the follow ways depending on what APIs the application is utilizing:

  • C: the error_cb is triggered with error code RD_KAFKA_RESP_ERR__FATAL, the application should call rd_kafka_fatal_error() to retrieve the underlying fatal error code and error string.
  • C: an RD_KAFKA_EVENT_ERROR event is triggered and rd_kafka_event_error_is_fatal() returns true: the fatal error code and string are available through rd_kafka_event_error(), and .._string().
  • C and C++: any API call may return RD_KAFKA_RESP_ERR__FATAL, use rd_kafka_fatal_error() to retrieve the underlying fatal error code and error string.
  • C++: an EVENT_ERROR event is triggered and event.fatal() returns true: the fatal error code and string are available through event.err() and event.str().

An application may call rd_kafka_fatal_error() at any time to check if a fatal error has been raised.

Fatal producer errors

The idempotent producer guarantees of ordering and no duplicates also requires a way for the client to fail gracefully when these guarantees can't be satisfied.

If a fatal error has been raised, sub-sequent use of the following API calls will fail:

  • rd_kafka_produce()
  • rd_kafka_producev()
  • rd_kafka_produce_batch()

The underlying fatal error code will be returned, depending on the error reporting scheme for each of those APIs.

When a fatal error has occurred the application should call rd_kafka_flush() to wait for all outstanding and queued messages to drain before terminating the application. rd_kafka_purge(RD_KAFKA_PURGE_F_QUEUE) is automatically called by the client when a producer fatal error has occurred, messages in-flight are not purged automatically to allow waiting for the proper acknowledgement from the broker. The purged messages in queue will fail with error code set to RD_KAFKA_RESP_ERR__PURGE_QUEUE.

Fatal consumer errors

A consumer configured for static group membership (group.instance.id) may raise a fatal error if a new consumer instance is started with the same instance id, causing the existing consumer to be fenced by the new consumer.

This fatal error is propagated on the fenced existing consumer in multiple ways:

  • error_cb (if configured) is triggered.
  • rd_kafka_consumer_poll() (et.al) will return a message object with the err field set to RD_KAFKA_ERR__FATAL.
  • any sub-sequent calls to state-changing consumer calls will return RD_KAFKA_ERR___FATAL. This includes rd_kafka_subscribe(), rd_kafka_assign(), rd_kafka_consumer_close(), rd_kafka_commit*(), etc.

The consumer will automatically stop consuming when a fatal error has occurred and no further subscription, assignment, consumption or offset committing will be possible. At this point the application should simply destroy the consumer instance and terminate the application since it has been replaced by a newer instance.

Compatibility

Broker version compatibility

librdkafka supports all released Apache Kafka broker versions since 0.8.0.0.0, but not all features may be available on all broker versions since some features rely on newer broker functionality.

Current defaults:

  • api.version.request=true
  • broker.version.fallback=0.10.0
  • api.version.fallback.ms=0 (never revert to broker.version.fallback)

Depending on what broker version you are using, please configure your librdkafka based client as follows:

Broker version >= 0.10.0.0 (or trunk)

For librdkafka >= v1.0.0 there is no need to set any api.version-related configuration parameters, the defaults are tailored for broker version 0.10.0.0 or later.

For librdkafka < v1.0.0, please specify:

api.version.request=true
api.version.fallback.ms=0

Broker versions 0.9.0.x

api.version.request=false
broker.version.fallback=0.9.0.x  (the exact 0.9.0.. version you are using)

Broker versions 0.8.x.y

api.version.request=false
broker.version.fallback=0.8.x.y  (your exact 0.8... broker version)

Detailed description

Apache Kafka version 0.10.0.0 added support for KIP-35 - querying the broker for supported API request types and versions - allowing the client to figure out what features it can use. But for older broker versions there is no way for the client to reliably know what protocol features the broker supports.

To alleviate this situation librdkafka has three configuration properties:

  • api.version.request=true|false - enables the API version request, this requires a >= 0.10.0.0 broker and will cause a disconnect on brokers 0.8.x - this disconnect is recognized by librdkafka and on the next connection attempt (which is immediate) it will disable the API version request and use broker.version.fallback as a basis of available features. NOTE: Due to a bug in broker version 0.9.0.0 & 0.9.0.1 the broker will not close the connection when receiving the API version request, instead the request will time out in librdkafka after 10 seconds and it will fall back to broker.version.fallback on the next immediate connection attempt.
  • broker.version.fallback=X.Y.Z.N - if the API version request fails (if api.version.request=true) or API version requests are disabled (api.version.request=false) then this tells librdkafka what version the broker is running and adapts its feature set accordingly.
  • api.version.fallback.ms=MS - In the case where api.version.request=true and the API version request fails, this property dictates for how long librdkafka will use broker.version.fallback instead of api.version.request=true. After MS has passed the API version request will be sent on any new connections made for the broker in question. This allows upgrading the Kafka broker to a new version with extended feature set without needing to restart or reconfigure the client (given that api.version.request=true).

Note: These properties applies per broker.

The API version query was disabled by default (api.version.request=false) in librdkafka up to and including v0.9.5 due to the afforementioned bug in broker version 0.9.0.0 & 0.9.0.1, but was changed to true in librdkafka v0.11.0.

Supported KIPs

The Apache Kafka Implementation Proposals (KIPs) supported by librdkafka.

KIP Kafka release Status
KIP-1 - Stop accepting request.required.acks > 1 0.9.0.0 Not enforced on client (due to backwards compat with brokers <0.8.3)
KIP-4 - Metadata protocol changes 0.9.0.0, 0.10.0.0, 0.10.1.0 Supported
KIP-8 - Producer flush() 0.9.0.0 Supported
KIP-12 - SASL Kerberos 0.9.0.0 Supported (uses SSPI/logged-on-user on Windows, full KRB5 keytabs on Unix)
KIP-13 - Protocol request throttling (enforced on broker) 0.9.0.0 Supported
KIP-15 - Producer close with timeout 0.9.0.0 Supported (through flush() + destroy())
KIP-19 - Request timeouts 0.9.0.0 Supported
KIP-22 - Producer pluggable partitioner 0.9.0.0 Supported (not supported by Go, .NET and Python)
KIP-31 - Relative offsets in messagesets 0.10.0.0 Supported
KIP-35 - ApiVersionRequest 0.10.0.0 Supported
KIP-40 - ListGroups and DescribeGroups 0.9.0.0 Supported
KIP-41 - max.poll.records 0.10.0.0 Supported through batch consumption interface (not supported by .NET and Go)
KIP-42 - Producer and Consumer interceptors 0.10.0.0 Supported (not supported by Go, .NET and Python)
KIP-43 - SASL PLAIN and handshake 0.10.0.0 Supported
KIP-48 - Delegation tokens 1.1.0 Not supported
KIP-54 - Sticky partition assignment strategy 0.11.0.0 Supported but not available, use KIP-429 instead.
KIP-57 - Interoperable LZ4 framing 0.10.0.0 Supported
KIP-62 - max.poll.interval and background heartbeats 0.10.1.0 Supported
KIP-70 - Proper client rebalance event on unsubscribe/subscribe 0.10.1.0 Supported
KIP-74 - max.partition.fetch.bytes 0.10.1.0 Supported
KIP-78 - Retrieve Cluster Id 0.10.1.0 Supported (not supported by .NET)
KIP-79 - OffsetsForTimes 0.10.1.0 Supported
KIP-81 - Consumer pre-fetch buffer size 2.4.0 (WIP) Supported
KIP-82 - Record Headers 0.11.0.0 Supported
KIP-84 - SASL SCRAM 0.10.2.0 Supported
KIP-85 - SASL config properties 0.10.2.0 Supported
KIP-86 - Configurable SASL callbacks 2.0.0 Not supported
KIP-88 - AdminAPI: ListGroupOffsets 0.10.2.0 Not supported
KIP-91 - Intuitive timeouts in Producer 2.1.0 Supported
KIP-92 - Per-partition lag metrics in Consumer 0.10.2.0 Supported
KIP-97 - Backwards compatibility with older brokers 0.10.2.0 Supported
KIP-98 - EOS 0.11.0.0 Supported
KIP-102 - Close with timeout in consumer 0.10.2.0 Not supported
KIP-107 - AdminAPI: DeleteRecordsBefore 0.11.0.0 Supported
KIP-110 - ZStd compression 2.1.0 Supported
KIP-117 - AdminClient 0.11.0.0 Supported
KIP-124 - Request rate quotas 0.11.0.0 Partially supported (depending on protocol request)
KIP-126 - Producer ensure proper batch size after compression 0.11.0.0 Supported
KIP-133 - AdminAPI: DescribeConfigs and AlterConfigs 0.11.0.0 Supported
KIP-140 - AdminAPI: ACLs 0.11.0.0 Not supported
KIP-144 - Broker reconnect backoff 0.11.0.0 Supported
KIP-152 - Improved SASL auth error messages 1.0.0 Supported
KIP-192 - Cleaner idempotence semantics 1.0.0 Not supported (superceeded by KIP-360)
KIP-195 - AdminAPI: CreatePartitions 1.0.0 Supported
KIP-204 - AdminAPI: DeleteRecords 1.1.0 Supported
KIP-219 - Client-side throttling 2.0.0 Not supported
KIP-222 - AdminAPI: Consumer group operations 2.0.0 Not supported (but some APIs available outside Admin client)
KIP-223 - Consumer partition lead metric 2.0.0 Not supported
KIP-226 - AdminAPI: Dynamic broker config 1.1.0 Supported
KIP-227 - Consumer Incremental Fetch 1.1.0 Not supported
KIP-229 - AdminAPI: DeleteGroups 1.1.0 Supported
KIP-235 - DNS alias for secure connections 2.1.0 Not supported
KIP-249 - AdminAPI: Deletegation Tokens 2.0.0 Not supported
KIP-255 - SASL OAUTHBEARER 2.0.0 Supported
KIP-266 - Fix indefinite consumer timeouts 2.0.0 Supported (bound by session.timeout.ms and max.poll.interval.ms)
KIP-289 - Consumer group.id default to NULL 2.2.0 Supported
KIP-294 - SSL endpoint verification 2.0.0 Supported
KIP-302 - Use all addresses for resolved broker hostname 2.1.0 Supported
KIP-320 - Consumer: handle log truncation 2.1.0, 2.2.0 Not supported
KIP-322 - DeleteTopics disabled error code 2.1.0 Supported
KIP-339 - AdminAPI: incrementalAlterConfigs 2.3.0 Not supported
KIP-341 - Update Sticky partition assignment data 2.3.0 Not supported (superceeded by KIP-429)
KIP-342 - Custom SASL OAUTHBEARER extensions 2.1.0 Supported
KIP-345 - Consumer: Static membership 2.4.0 Supported
KIP-357 - AdminAPI: list ACLs per principal 2.1.0 Not supported
KIP-359 - Producer: use EpochLeaderId 2.4.0 Not supported
KIP-360 - Improve handling of unknown Idempotent Producer 2.5.0 Supported
KIP-361 - Consumer: add config to disable auto topic creation 2.3.0 Supported
KIP-368 - SASL periodic reauth 2.2.0 Not supported
KIP-369 - Always roundRobin partitioner 2.4.0 Not supported
KIP-389 - Consumer group max size 2.2.0 Supported (error is propagated to application, but the consumer does not raise a fatal error)
KIP-392 - Allow consumers to fetch from closest replica 2.4.0 Supported
KIP-394 - Consumer: require member.id in JoinGroupRequest 2.2.0 Supported
KIP-396 - AdminAPI: commit/list offsets 2.4.0 Not supported (but some APIs available outside Admin client)
KIP-412 - AdminAPI: adjust log levels 2.4.0 Not supported
KIP-421 - Variables in client config files 2.3.0 Not applicable (librdkafka, et.al, does not provide a config file interface, and shouldn't)
KIP-429 - Consumer: incremental rebalance protocol 2.4.0 Supported
KIP-430 - AdminAPI: return authorized operations in Describe.. responses 2.3.0 Not supported
KIP-436 - Start time in stats 2.3.0 Supported
KIP-447 - Producer scalability for EOS 2.5.0 Supported
KIP-455 - AdminAPI: Replica assignment 2.4.0 (WIP) Not supported
KIP-460 - AdminAPI: electPreferredLeader 2.4.0 Not supported
KIP-464 - AdminAPI: defaults for createTopics 2.4.0 Supported
KIP-467 - Per-message (sort of) error codes in ProduceResponse 2.4.0 (WIP) Not supported
KIP-480 - Sticky partitioner 2.4.0 Supported
KIP-482 - Optional fields in Kafka protocol 2.4.0 Partially supported (ApiVersionRequest)
KIP-496 - AdminAPI: delete offsets 2.4.0 Supported
KIP-511 - Collect Client's Name and Version 2.4.0 Supported
KIP-514 - Bounded flush() 2.4.0 Supported
KIP-517 - Consumer poll() metrics 2.4.0 Not supported
KIP-518 - Allow listing consumer groups per state 2.6.0 Not supported
KIP-519 - Make SSL engine configurable 2.6.0 Supported
KIP-525 - Return topic metadata and configs in CreateTopics response 2.4.0 Not supported
KIP-526 - Reduce Producer Metadata Lookups for Large Number of Topics 2.5.0 Not supported
KIP-533 - Add default API timeout to AdminClient 2.5.0 Not supported
KIP-546 - Add Client Quota APIs to AdminClient 2.6.0 Not supported
KIP-559 - Make the Kafka Protocol Friendlier with L7 Proxies 2.5.0 Not supported
KIP-568 - Explicit rebalance triggering on the Consumer 2.6.0 Not supported
KIP-659 - Add metadata to DescribeConfigsResponse 2.6.0 Not supported
KIP-580 - Exponential backoff for Kafka clients WIP Partially supported
KIP-584 - Versioning scheme for features WIP Not supported
KIP-588 - Allow producers to recover gracefully from txn timeouts 2.8.0 (WIP) Not supported
KIP-602 - Use all resolved addresses by default 2.6.0 Supported
KIP-651 - Support PEM format for SSL certs and keys 2.7.0 Supported
KIP-654 - Aborted txns with non-flushed msgs should not be fatal 2.7.0 Supported
KIP-735 - Increase default consumer session timeout 3.0.0 Supported
KIP-768 - SASL/OAUTHBEARER OIDC support WIP Not supported

Supported protocol versions

"Kafka max" is the maximum ApiVersion supported in Apache Kafka 2.4.0, while "librdkafka max" is the maximum ApiVersion supported in the latest release of librdkafka.

ApiKey Request name Kafka max librdkafka max
0 Produce 7 7
1 Fetch 11 11
2 ListOffsets 5 1
3 Metadata 8 2
8 OffsetCommit 7 7
9 OffsetFetch 5 1
10 FindCoordinator 2 2
11 JoinGroup 5 5
12 Heartbeat 3 3
13 LeaveGroup 3 1
14 SyncGroup 3 3
15 DescribeGroups 4 0
16 ListGroups 2 0
17 SaslHandshake 1 1
18 ApiVersions 3 3
19 CreateTopics 5 4
20 DeleteTopics 3 1
21 DeleteRecords 2 1
22 InitProducerId 4 4
24 AddPartitionsToTxn 1 0
25 AddOffsetsToTxn 1 0
26 EndTxn 1 1
28 TxnOffsetCommit 2 0
32 DescribeConfigs 2 1
33 AlterConfigs 1 0
36 SaslAuthenticate 1 0
37 CreatePartitions 1 0
42 DeleteGroups 2 1
47 OffsetDelete 0 0

Recommendations for language binding developers

These recommendations are targeted for developers that wrap librdkafka with their high-level languages, such as confluent-kafka-go or node-rdkafka.

Expose the configuration interface pass-thru

librdkafka's string-based key=value configuration property interface controls most runtime behaviour and evolves over time. Most features are also only configuration-based, meaning they do not require a new API (SSL and SASL are two good examples which are purely enabled through configuration properties) and thus no changes needed to the binding/application code.

If your language binding/applications allows configuration properties to be set in a pass-through fashion without any pre-checking done by your binding code it means that a simple upgrade of the underlying librdkafka library (but not your bindings) will provide new features to the user.

Error constants

The error constants, both the official (value >= 0) errors as well as the internal (value < 0) errors, evolve constantly. To avoid hard-coding them to expose to your users, librdkafka provides an API to extract the full list programmatically during runtime or for code generation, see rd_kafka_get_err_descs().

Reporting client software name and version to broker

KIP-511 introduces a means for a Kafka client to report its implementation name and version to the broker, the broker then exposes this as metrics (e.g., through JMX) to help Kafka operators troubleshoot problematic clients, understand the impact of broker and client upgrades, etc. This requires broker version 2.4.0 or later (metrics added in 2.5.0).

librdkafka will send its name (librdkafka) and version (e.g., v1.3.0) upon connect to a supporting broker. To help distinguish high-level client bindings on top of librdkafka, a client binding should configure the following two properties:

  • client.software.name - set to the binding name, e.g, confluent-kafka-go or node-rdkafka.
  • client.software.version - the version of the binding and the version of librdkafka, e.g., v1.3.0-librdkafka-v1.3.0 or 1.2.0-librdkafka-v1.3.0. It is highly recommended to include the librdkafka version in this version string.

These configuration properties are hidden (from CONFIGURATION.md et.al.) as they should typically not be modified by the user.

Documentation reuse

You are free to reuse the librdkafka API and CONFIGURATION documentation in your project, but please do return any documentation improvements back to librdkafka (file a github pull request).

Community support

You are welcome to direct your users to librdkafka's Gitter chat room as long as you monitor the conversions in there to pick up questions specific to your bindings. But for the most part user questions are usually generic enough to apply to all librdkafka bindings.