The kafka-connect-wrap-smt is a single message transform (SMT) that wraps key and record of kafka messages into a single struct. This ensures, e.g., that data contained in complex keys is not lost when ingesting data from kafka in a sink such as elasticsearch. Additionally, it supports exporting meta-data including partition, offset, timestamp, topic name and kafka headers.
Note that kafka-connect-wrap-smt does only support sink connectors, as it wraps kafka specific meta-data that is not available for all source connectors.
To install the latest release, you can download the plugin binaries directly from github or build them from source (see section Build below):
curl -sLJO https://github.com/f0xdx/kafka-connect-wrap-smt/releases/download/v0.2.0/kafka-connect-wrap-smt-0.2.0.jar
cp kafka-connect-wrap-smt-0.2.0.jar connect/plugin/folder
Make sure that the plugin folder is picked up by kafka connect by verifying its logs. For instance,
with docker-compose
, you could run docker-compose logs connect | grep Wrap
which should show
relevant logs, e.g.,
connect | [2020-03-25 12:48:00,429] INFO Added plugin 'com.github.f0xdx.Wrap' (org.apache.kafka.connect.runtime.isolation.DelegatingClassLoader)
connect | [2020-03-25 12:48:01,463] INFO Added alias 'Wrap' to plugin 'com.github.f0xdx.Wrap' (org.apache.kafka.connect.runtime.isolation.DelegatingClassLoader)
After installing the plugin, you can configure your connector to apply the SMT, e.g.:
{
"transforms": "wrap",
"transforms.wrap.type": "Wrap",
"transforms.wrap.include.headers": false
}
As stated above, this SMT can only be used in conjunction with sink connectors.
To build this project locally simply run:
git clone git@github.com:f0xdx/kafka-connect-wrap-smt.git
./gradlew build
After building, you can deploy the build/libs/kafka-connect-wrap-smt-0.1-SNAPSHOT.jar
into the
plugins folder of your kafka connect instance, e.g.:
cp build/libs/kafka-connect-wrap-smt-0.2.0.jar connect/plugin/folder
Upcoming features are:
- CI/CD and automated publishing using github packages
- install script + docker image for ready made connect
- schema based export with schema-less keys (if key schema can be derived)