[Guide] RedPanda

[Paultagoras]

We should have a detailed guide for setting up the connector with a RedPanda environment.

[jirislav]

I'm currently trying to use this connector in a Self-Hosted Redpanda Kubernetes Setup.

I'm installing the connector by extracting the released ZIP file within an init container that I added to the native Deployment/redpanda-connectors.

Using this approach, Redpanda's Connectors workload recognizes the ClickHouse Connector:

But there are problems with missing classes when trying to configure it using Avro:

Class io.confluent.connect.avro.AvroConverter could not be found

When provided necessary JARs, I was able to instantiate the connector. But in runtime, I got:

org.apache.kafka.connect.errors.ConnectException: Instantiation error
	at org.apache.kafka.connect.runtime.isolation.Plugins.newPlugin(Plugins.java:85)
	at org.apache.kafka.connect.runtime.isolation.Plugins.newConverter(Plugins.java:327)
	at org.apache.kafka.connect.runtime.Worker.startTask(Worker.java:622)
	at org.apache.kafka.connect.runtime.Worker.startSinkTask(Worker.java:525)
	at org.apache.kafka.connect.runtime.distributed.DistributedHerder.startTask(DistributedHerder.java:1800)
	at org.apache.kafka.connect.runtime.distributed.DistributedHerder.lambda$getTaskStartingCallable$32(DistributedHerder.java:1850)
	at java.base/java.util.concurrent.FutureTask.run(FutureTask.java:264)
	at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1136)
	at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:635)
	at java.base/java.lang.Thread.run(Thread.java:833)
Caused by: java.lang.NoClassDefFoundError: io/confluent/kafka/schemaregistry/client/SchemaRegistryClient
	at java.base/java.lang.Class.getDeclaredConstructors0(Native Method)
	at java.base/java.lang.Class.privateGetDeclaredConstructors(Class.java:3373)
	at java.base/java.lang.Class.getConstructor0(Class.java:3578)
	at java.base/java.lang.Class.getDeclaredConstructor(Class.java:2754)
	at org.apache.kafka.common.utils.Utils.newInstance(Utils.java:396)
	at org.apache.kafka.connect.runtime.isolation.Plugins.newPlugin(Plugins.java:83)
	... 9 more
Caused by: java.lang.ClassNotFoundException: io.confluent.kafka.schemaregistry.client.SchemaRegistryClient
	at java.base/java.net.URLClassLoader.findClass(URLClassLoader.java:445)
	at java.base/java.lang.ClassLoader.loadClass(ClassLoader.java:592)
	at org.apache.kafka.connect.runtime.isolation.PluginClassLoader.loadClass(PluginClassLoader.java:136)
	at java.base/java.lang.ClassLoader.loadClass(ClassLoader.java:525)
	... 15 more

Would you consider also publishing a fat jar into releases, that would contain all the needed classes for this to work without users needing to fetch multiple modules separately?

[Paultagoras]

Would you consider also publishing a fat jar into releases, that would contain all the needed classes for this to work without users needing to fetch multiple modules separately?

We'll certainly consider it - we're not using something specific to Confluent though in the connector itself, that referenced jar file is meant to work with the Confluent Schema Registry.

What I've seen work before is to include something more like https://mvnrepository.com/artifact/io.confluent/kafka-avro-serializer/7.6.0 (though I forget the specific dependency) - either way, we'll discuss it internally!

[jirislav]

Just to let you know, the JARs package that I've linked previously is all that you need apart from the pre-built release.

The problem with the exception I have shared with you was that the Confluent's JARs need to be added to the classpath. Solely putting them to the CONNECT_PLUGIN_PATH wasn't enough.

If you wanted to experiment with setting up Redpanda Connectors in Kubernetes, here's my approach using Flux's HelmRelease's Post Renderers (to don't have to think about backporting my changes to new Redpanda releases by installing it through initContainer):

apiVersion: source.toolkit.fluxcd.io/v1beta2
kind: HelmRepository
metadata:
  name: redpanda
spec:
  interval: 1h
  url: https://charts.redpanda.com
---
apiVersion: helm.toolkit.fluxcd.io/v2beta2
kind: HelmRelease
metadata:
  name: redpanda
  namespace: redpanda
spec:
  chart:
    spec:
      chart: redpanda
      interval: 12h
      reconcileStrategy: ChartVersion
      sourceRef:
        kind: HelmRepository
        name: redpanda
        namespace: redpanda
      version: '*'
  interval: 30s
  timeout: 10m0s
  values:
    # Values ref: https://github.com/redpanda-data/helm-charts/blob/main/charts/redpanda/values.yaml
    connectors:
      enabled: true
      deployment:
        replicas: 2
  postRenderers:
    - kustomize:
        patches:
          - patch: |
              - op: add
                path: /spec/template/spec/volumes/-
                value:
                  - name: custom-connectors
                    emptyDir: {}
              - op: add
                path: /spec/template/spec/containers/0/volumeMounts/-
                value:
                  - name: custom-connectors
                    mountPath: /custom-connectors
              - op: add
                path: /spec/template/spec/containers/0/env/-
                value:
                  - name: CONNECT_PLUGIN_PATH
                    value: /custom-connectors
              - op: add
                path: /spec/template/spec/containers/0/env/-
                value:
                  - name: CLASSPATH
                    value: /custom-connectors/*
              - op: add
                path: /spec/template/spec/initContainers
                value:
                  - name: custom-connectors
                    image: alpine
                    env:
                      - name: CH_VERSION
                        value: v1.0.16
                      - name: AVRO_CONVERTER_VERSION
                        value: 7.6.0
                    command:
                      - sh
                      - '-c'
                      - |
                        cd /tmp
                        wget https://github.com/ClickHouse/clickhouse-kafka-connect/releases/download/${CH_VERSION}/clickhouse-kafka-connect-${CH_VERSION}.zip
                        unzip -j clickhouse-kafka-connect-${CH_VERSION}.zip '*.jar' -d /custom-connectors

                        wget https://d1i4a15mxbxib1.cloudfront.net/api/plugins/confluentinc/kafka-connect-avro-converter/versions/${AVRO_CONVERTER_VERSION}/confluentinc-kafka-connect-avro-converter-${AVRO_CONVERTER_VERSION}.zip
                        unzip -j confluentinc-kafka-connect-avro-converter-${AVRO_CONVERTER_VERSION}.zip '*.jar' -d /custom-connectors
                    volumeMounts:
                      - name: custom-connectors
                        mountPath: /custom-connectors
            target:
              kind: Deployment
              name: redpanda-connectors

Alternative approach could be building a custom Docker image that would include all the necessary JARs.