Based on mm2-lag-exporter.

Mirrormaker 2

Apache Kafka MirrorMaker2 replicates data across two Kafka clusters, within or across data centers. MirrorMaker takes messages from a source Kafka cluster and writes them to a target Kafka cluster.

MM2 Lag Exporter

This application monitors the mirror maker 2 replication process and calculates the mirroring lag between source and target cluster topics.These lag details are exposed as REST endpoints and as well as prometheus metrics which can be visualized in grafana dashboard.

Clusters configuration

All configuration should be stored in a file clusters.yaml

Check cluster-demo.yaml. It contains 2 main sections.

• clusters - describe each cluster. You should set cluster alias as a name of the property of clusters object. Put all kafka properties as a fields of your alias.
• traffic - an array that defines how data is copied between the clusters. For each flow set source cluster alias in from and target cluster alias in target

connectorName should be the name of the connector with "connector.class": "org.apache.kafka.connect.mirror.MirrorSourceConnector". This property will be used to collect offsets from the mm2 offsets topic.

mm2OffsetsTopic - the name of the topic, used by mm2 to store offsets

Run application

Checkout the sources:

git clone https://github.com/papirosko/mm2-lag.git
cd mm2-lag

Create clusters.yaml.

Compile and run the application:

sbt clean stage
./target/universal/stage/bin/mm2-lag

Open http://localhost:8080. Check metrics endpoint, take a look at counters section, you should see SourceConsumer and TargetConsumer counters increasing for each cluster.

Also you can build archive with the application:

sbt clean universal:packageBin

You can override the cluster file location with environment variable CLUSTERS_FILE:

CLUSTERS_FILE="/etc/clusters.yaml" ./target/universal/stage/bin/mm2-lag

Lag values

Use /api/lag_per_topic endpoint to see lag values per each topic in text format.

Use /api/prometheus_lag to see all lags per partition for all clusters. This endpoint can be used for prometheus to grab the metrics.

Docker

Build and run docker image locally:

Create clusters.yaml.

sbt docker:publishLocal
docker run --mount type=bind,source=$(pwd)/clusters.yaml,target=/opt/docker/clusters.yaml mm2-lag

You can mount the required files in the same way, e.g. *.jks files.

Use this to push image to remote registry:

DOCKER_REGISTRY="gitlab.server:5050" sbt docker:publish

You can also use docker image from docker.io: https://hub.docker.com/r/papirosko/mm2-lag

Development

Build and publish (any branch):

DOCKER_REGISTRY="docker.io" CI_REGISTRY_IMAGE="docker.io/papirosko/mm2-lag" sbt clean docker:publish

For main branch it will deploy docker image with tag equal to the app version, so you need to increase app version before publishing. Also it will update latest tag.

For any other branch it will deploy docker image with the branch name as a tag.