pubsub VS jmx_exporter

In this example, we’re reading from the Kafka topic demo-event-stream-export and writing to the GCP PubSub topic mux-demo-event-stream-export in the project external-mux using the GCP PubSub sink implementation. The connector will run 6 Kafka Connector tasks; each task works as a Kafka consumer, so it makes sense for the number of tasks to not exceed the number of partitions on the source topic (e.g. demo-event-stream-export). The connector sink is identified by the connector.class setting, which references the GCP sink implementation.

not quite sure if this is where I should be asking for help but i'm kind of befuddled about this whole situation. I'm trying to set up https://github.com/prometheus/jmx_exporter for our containerized Java application on our cluster. Specifically the JavaAgent as we are interested in getting the CPU and memory metrics especially. However, getting it initialized I am faced with this:

Prometheus doesn't deal with logs at all - only metrics. Even just for metrics, you have several different choices depending on how your Java application is running. You may be able to use the JMX Exporter to expose the metrics that you would then scrape with Prometheus. For the logs, you'll need something like Fluent Bit as an agent to grab ('tail' in Fluent Bit input terms) the logs and send them to Elasticsearch or Grafana Loki.

- https://github.com/prometheus/jmx_exporter - https://github.com/lightbend/kafka-lag-exporter (Export Kafka Topics Lag Metrics) - https://github.com/danielqsj/kafka_exporter (Export Kafka Metrics) - https://github.com/obsidiandynamics/kafdrop (Topic Administration tool - create/edit/delete topics & view messages)

Usually I use a javaagent that transforms the JMX metrics into prometheus style metrics, for example: https://github.com/prometheus/jmx_exporter/blob/116a1c4b3fafcfbb9e566b34e130ccc805f331fc/example_configs/kafka-0-8-2.yml

Another good one is to add the jmx_exporter agent plugin. This allows easy collection of real-time performance metrics.

In our case, we add sink connectors for AWS Kinesis and Google Cloud PubSub. We also add a Prometheus exporter JAR that scrapes the Kafka Connect JMX metrics and exposes them as Prometheus metrics.