Configure the Kafka outputedit

The Kafka output sends the events to Apache Kafka.

Example configuration:

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

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

  required_acks: 1
  compression: gzip
  max_message_bytes: 1000000

Events bigger than max_message_bytes will be dropped. To avoid this problem, make sure APM Server does not generate events bigger than max_message_bytes.


This output works with all Kafka in between 0.11 and 1.1.1. Older versions might work as well, but are not supported.

Configuration optionsedit

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 true.


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 is enabled.

Valid values are all kafka releases in between and 1.1.1.


The username for connecting to Kafka. If username is configured, the password must be configured as well. Only SASL/PLAIN is supported.


The password for connecting to Kafka.


The Kafka topic used for produced events. The setting can be a format string using any event field. For example, you can use the fields configuration option to add a custom field called log_topic to the event, and then set topic to the value of the custom field:

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


Array of topic selector rules supporting conditionals, format string based field access and name mappings. The first rule matching will be used to set the topic for the event to be published. If topics is missing or no rule matches, the topic field will be used.

Rule settings:

topic: The topic format string to use. If the fields used are missing, the rule fails.

mapping: Dictionary mapping index names to new names

default: Default string value if mapping does not find a match.

when: Condition which must succeed in order to execute the current rule.


Optional Kafka event key. If configured, the event key must be unique and can be extracted from the event using a format string.


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.

Publishing to a subset of available partitions potentially increases resource usage because events may become unevenly distributed.


The configurable ClientID 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.

Metadata refresh interval. Defaults to 10 minutes.
Total number of metadata update retries when cluster is in middle of leader election. The default is 3.
Waiting time between retries during leader elections. Default is 250ms.


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


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 0 seconds.


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.


Configuration options for SSL parameters like the root CA for Kafka connections. See Specify SSL settings for more information.