Skip to content

jhc-systems/brave-kafka-interceptor

 
 

Repository files navigation

Brave: Kafka Interceptor

Gitter chat Build Status Maven Central

Kafka Consumer and Producer Interceptors for tracing and report to Zipkin.

These interceptors could be plugged into Kafka applications via classpath configuration.

The purpose of this instrumentation is to enable tracing for Kafka Connect and other off-the-shelf components like Kafka REST Proxy, ksqlDB, etc. For more complete tracing support, check Brave instrumentation for Kafka Clients and Kafka Streams.

Installation

Producer Interceptor

Producer Interceptor create spans when sending records. This span will only represent the time it took to execute the on_send method provided by the API, not how long to send the actual record, or any other latency.

Kafka Clients

Add Interceptor to Producer Configuration:

    producerConfig.put(ProducerConfig.INTERCEPTOR_CLASSES_CONFIG, Collections.singletonList(TracingProducerInterceptor.class));
    //or
    producerConfig.put("interceptor.classes", "brave.kafka.interceptor.TracingProducerInterceptor");

Consumer Interceptor

Consumer Interceptor create spans on consumption of records. This span will only represent the time it took execute the on_consume method provided by the API, not how long it took to commit, or any other latency.

Kafka Clients

    consumerConfig.put(ConsumerConfig.INTERCEPTOR_CLASSES_CONFIG, Collections.singletonList(TracingConsumerInterceptor.class));
    //or
    consumerConfig.put("interceptor.classes", "brave.kafka.interceptor.TracingConsumerInterceptor");

Configuration

Key Value
zipkin.sender.type Sender type: NONE(default), KAFKA, HTTP
zipkin.encoding Zipkin encoding: JSON(default), PROTO3.
zipkin.http.endpoint Zipkin HTTP Endpoint sender.
zipkin.kafka.bootstrap.servers Bootstrap Servers list to send Spans. if not present, bootstrap.servers (Kafka Client property) is used.
zipkin.local.service.name Application Service name used to tag span. Default: kafka-client.
zipkin.trace.id.128bit.enabled Trace ID 128 bit enabled, default: true
zipkin.sampler.rate Rate to sample spans. Default: 1.0

How to test it

Required software available:

  • Docker and Docker Compose available.
  • JDK 11+

Start by building libraries and run Docker Compose:

make docker-up # will build jars and start containers
# or
./mvnw clean package
docker-compose up -d

Create database tables:

make pg-table

Once table is created deploy source and sink connectors:

make kafka-connectors

Create a KSQL table and select to wait for results:

make ksql-stream
make ksql-select

Insert new rows:

make pg-row

Go to http://localhost:9411 Check the traces.

Traces should look like this:

Search:

Trace view:

Dependencies:

Artifacts

All artifacts publish to the group ID "io.zipkin.contrib.brave-kafka-interceptor". We use a common release version for all components.

Library Releases

Releases are at Sonatype and Maven Central

Library Snapshots

Snapshots are uploaded to Sonatype after commits to master.

Packages

No packages published

Languages

  • Java 74.3%
  • Shell 22.9%
  • Makefile 2.8%