Configure the Kafka output

The Kafka output sends events to Apache Kafka.

To use this output, edit the APM Server configuration file to disable the Elasticsearch output by commenting it out, and enable the Kafka output by uncommenting the Kafka section.

Example configuration:

output.kafka:
  # initial brokers for reading cluster metadata
  hosts: ["kafka1:9092", "kafka2:9092", "kafka3:9092"]

  # message topic selection + partitioning
  topic: '%{[fields.log_topic]}'
  partition.round_robin:
    reachable_only: false

  required_acks: 1
  compression: gzip
  max_message_bytes: 1000000

APM Server uses the APM integration to set up and manage APM templates, policies, and pipelines. To confirm the integration is installed, APM Server polls either Elasticsearch or Kibana on startup. When using a non-Elasticsearch output, APM Server requires access to Kibana via the Kibana endpoint.

Example configuration:

apm-server:
  kibana:
    enabled: true
    host: "https://..."
    username: "elastic"
    password: "xxx"

This output works with all Kafka versions in between 0.11 and 2.2.2. Older versions might work as well, but are not supported.

You can specify the following options in the kafka section of the apm-server.yml config file:

The enabled config is a boolean setting to enable or disable the output. If set to false, the output is disabled.

The default value is false.

The list of Kafka broker addresses from where to fetch the cluster metadata. The cluster metadata contain the actual Kafka brokers events are published to.

Kafka version apm-server is assumed to run against. Defaults to 1.0.0.

Event timestamps will be added, if version 0.10.0.0+ is enabled.

Valid values are all Kafka releases in between 0.8.2.0 and 2.0.0.

See Compatibility for information on supported versions.

The username for connecting to Kafka. If username is configured, the password must be configured as well.

The password for connecting to Kafka.

The SASL mechanism to use when connecting to Kafka. It can be one of:

If sasl.mechanism is not set, PLAIN is used if username and password are provided. Otherwise, SASL authentication is disabled.

The Kafka topic used for produced events.

You can set the topic dynamically by using a format string to access any event field. For example, this configuration uses a custom field, fields.log_topic, to set the topic for each event:

topic: '%{[fields.log_topic]}'

See the topics setting for other ways to set the topic dynamically.

An array of topic selector rules. Each rule specifies the topic to use for events that match the rule. During publishing, APM Server sets the topic for each event based on the first matching rule in the array. Rules can contain conditionals, format string-based fields, and name mappings. If the topics setting is missing or no rule matches, the topic field is used.

Rule settings:

The following example sets the topic based on whether the message field contains the specified string:

output.kafka:
  hosts: ["localhost:9092"]
  topic: "logs-%{[agent.version]}"
  topics:
    - topic: "critical-%{[agent.version]}"
      when.contains:
        message: "CRITICAL"
    - topic: "error-%{[agent.version]}"
      when.contains:
        message: "ERR"

This configuration results in topics named critical-8.13.4, error-8.13.4, and logs-8.13.4.

Optional formatted string specifying the Kafka event key. If configured, the event key can be extracted from the event using a format string.

See the Kafka documentation for the implications of a particular choice of key; by default, the key is chosen by the Kafka cluster.

Kafka output broker event partitioning strategy. Must be one of random, round_robin, or hash. By default the hash partitioner is used.

random.group_events: Sets the number of events to be published to the same partition, before the partitioner selects a new partition by random. The default value is 1 meaning after each event a new partition is picked randomly.

round_robin.group_events: Sets the number of events to be published to the same partition, before the partitioner selects the next partition. The default value is 1 meaning after each event the next partition will be selected.

hash.hash: List of fields used to compute the partitioning hash value from. If no field is configured, the events key value will be used.

hash.random: Randomly distribute events if no hash or key value can be computed.

All partitioners will try to publish events to all partitions by default. If a partition’s leader becomes unreachable for the beat, the output might block. All partitioners support setting reachable_only to overwrite this behavior. If reachable_only is set to true, events will be published to available partitions only.

The configurable client ID used for logging, debugging, and auditing purposes. The default is "beats".

The number of concurrent load-balanced Kafka output workers.

Output codec configuration. If the codec section is missing, events will be JSON encoded.

See Change the output codec for more information.

Kafka metadata update settings. The metadata do contain information about brokers, topics, partition, and active leaders to use for publishing.

The number of times to retry publishing an event after a publishing failure. After the specified number of retries, the events are typically dropped.

Set max_retries to a value less than 0 to retry until all events are published.

The default is 3.

The number of seconds to wait before trying to republish to Kafka after a network error. After waiting backoff.init seconds, APM Server tries to republish. If the attempt fails, the backoff timer is increased exponentially up to backoff.max. After a successful publish, the backoff timer is reset. The default is 1s.

The maximum number of seconds to wait before attempting to republish to Kafka after a network error. The default is 60s.

The maximum number of events to bulk in a single Kafka request. The default is 2048.

Duration to wait before sending bulk Kafka request. 0 is no delay. The default is 0.

The number of seconds to wait for responses from the Kafka brokers before timing out. The default is 30 (seconds).

The maximum duration a broker will wait for number of required ACKs. The default is 10s.

Per Kafka broker number of messages buffered in output pipeline. The default is 256.

The keep-alive period for an active network connection. If 0s, keep-alives are disabled. The default is 0s.

Sets the output compression codec. Must be one of none, snappy, lz4 and gzip. The default is gzip.

Sets the compression level used by gzip. Setting this value to 0 disables compression. The compression level must be in the range of 1 (best speed) to 9 (best compression).

Increasing the compression level will reduce the network usage but will increase the CPU usage.

The default value is 4.

The maximum permitted size of JSON-encoded messages. Bigger messages will be dropped. The default value is 1000000 (bytes). This value should be equal to or less than the broker’s message.max.bytes.

The ACK reliability level required from broker. 0=no response, 1=wait for local commit, -1=wait for all replicas to commit. The default is 1.

Note: If set to 0, no ACKs are returned by Kafka. Messages might be lost silently on error.

Enable Kerberos FAST authentication. This may conflict with some Active Directory installations. It is separate from the standard Kerberos settings because this flag only applies to the Kafka output. The default is false.

Configuration options for SSL parameters like the root CA for Kafka connections. The Kafka host keystore should be created with the -keyalg RSA argument to ensure it uses a cipher supported by Filebeat’s Kafka library. See SSL/TLS output settings for more information.