If you have many(?) Kafka messages that you need to take a full snapshot of and store in another database, this case study may be a good starting point for you.
When you start the application, it will produce 5M messages to Kafka with 5 partitions.
After that, it will start a consumer group that store messages into the buffer and flush them to be bulk indexed in Elasticsearch.
You can fine-tune the current batch size (20K) and make your own performance tests.
Here is an example consumer log with 3K batch size;
