Kafka (om_kafka)

This module implements an Apache Kafka producer for publishing event records to a Kafka topic. See also the im_kafka module.

The om_kafka module is not supported as the underlying librdkafka library is unstable on AIX. Use it on IBM AIX at your own risk.

The module uses an internal persistent queue to back up event records that should be pushed to a Kafka broker. Once the module receives an acknowledgement from the Kafka server that the message has been delivered successfully, the module removes the corresponding message from the internal queue. If the module is unable to deliver a message to a Kafka broker (for example, due to connectivity issues or the Kafka server being down), this message is retained in the internal queue (including cases when NXLog restarts) and the module will attempt to re-deliver the message again.

The number of re-delivery attempts can be specified by passing the message.send.max.retries property via the Option directive (for example, Option message.send.max.retries 5). By default, the number of retries is set to 2 and the time interval between two subsequent retries is 5 minutes. Thus, by altering the number of retries, it is possible to control the total time for a message to remain in the internal queue. If a message cannot be delivered within the allowed retry attempts, the message is dropped. The maximum size of the internal queue defaults to 100 messages. To increase the size of the internal queue, you can use the LogqueueSize directive.


The om_kafka module accepts the following directives in addition to the common module directives. The BrokerList and Topic directives are required.


This mandatory directive specifies the list of Kafka brokers to connect to for publishing logs. The list should include ports and be comma-delimited (for example, localhost:9092,


This mandatory directive specifies the Kafka topic to publish records to.


This specifies the path of the certificate authority (CA) certificate, which will be used to check the certificate of the remote brokers. CAFile is required if Protocol is set to ssl or sasl_ssl. To trust a self-signed certificate presented by the remote (which is not signed by a CA), provide that certificate instead.


This optional directive specifies the thumbprint of the certificate authority (CA) certificate that will be used to verify the certificate presented by the remote server. The hexadecimal fingerprint string can be copied from Windows Certificate Manager (certmgr.msc). Whitespaces are automatically removed. The certificate must be added to a Windows certificate store that is accessible by NXLog. This directive is only supported on Windows and is mutually exclusive with the CADir and CAFile directives.


This specifies the path of the certificate file to be used for the SSL handshake.


This specifies the path of the certificate key file to be used for the SSL handshake.


This directive specifies the compression types to use during transfer. Available types depend on the Kafka library, and should include none (the default), gzip, snappy, and lz4.


With this directive, a password can be supplied for the certificate key file defined in CertKeyFile. This directive is not needed for passwordless private keys.


This directive can be used to pass a custom configuration property to the Kafka library (librdkafka). For example, the group ID string can be set with Option group.id mygroup. This directive may be used more than once to specify multiple options. For a list of configuration properties, see the librdkafka CONFIGURATION.md file.

Passing librdkafka configuration properties via the Option directive should be done with care since these properties are used for the fine-tuning of the librdkafka performance and may result in various side effects.

This optional integer directive specifies the topic partition to write to. If this directive is not given, messages are sent without a partition specified.


This optional directive specifies the protocol to use for connecting to the Kafka brokers. Accepted values include plaintext (the default), ssl, sasl_plaintext and sasl_ssl. If Protocol is set to ssl or sasl_ssl, then the CAFile directive must also be provided.


This directive specifies the Kerberos service name to be used for SASL authentication. The service name is required for the sasl_plaintext and sasl_ssl protocols.


This specifies the client’s Kerberos principal name for the sasl_plaintext and sasl_ssl protocols. This directive is only available and mandatory on Linux/UNIX. See note below.


Specifies the path to the kerberos keytab file which contains the client’s allocated principal name. This directive is only available and mandatory on Linux/UNIX.

The SASLKerberosServiceName and SASLKerberosPrincipal directives are only available on Linux/UNIX. On Windows, the login user’s principal name and credentials are used for SASL/Kerberos authentication.

For details about configuring Apache Kafka brokers to accept SASL/Kerberos authentication from clients, please follow the instructions provided by the librdkafka project:


Example 1. Using the om_kafka Module

This configuration sends events to a Kafka cluster using the brokers specified. Events are published to the first partition of the nxlog topic.

<Output out>
    Module          om_kafka
    BrokerList      localhost:9092,
    Topic           nxlog
    LogqueueSize    100000
    Partition       0
    Protocol        ssl
    CAFile          /root/ssl/ca-cert
    CertFile        /root/ssl/client_debian-8.pem
    CertKeyFile     /root/ssl/client_debian-8.key
    KeyPass         thisisasecret

The librdkafka library can produce its performance statistics and format it in JSON. All fields from the JSON structure are explained on the Statistics page of the librdkafka project on the GitHub website. NXLog can be configured to poll this data at a specified fixed interval. The result can be saved to the internal logger.

Example 2. Collecting Internal Statistics

To read statistical data of the librdkafka library, the millisecond polling interval needs to be specified against the Option directive using the statistics.interval.ms option.

The Schedule block sets the interval to run the code of the nested Exec block. Inside the Exec block, the log_info() procedure is called with the kafka_in->get_stats() parameter passed.

To get the librdkafka statistics produced and delivered synchronously, the statistics.interval.ms option and the Schedule block should specify the same interval amount.

nxlog.conf, Writing to the Internal Logger
<Input to_kafka>
    Module        im_kafka
    Topic         nxlog
    BrokerList    localhost:9092
    Option        statistics.interval.ms 10000
        Every     10 sec
        Exec      log_info(to_kafka->get_stats());