When a Kafka Connect connector handles more data, tasks.max sets the upper limit on how many independent tasks may divide its workload. Raising the value enables parallelization only when the connector and its data can support it. section 11.6.1 section 11.3
One file, one task
A FileStreamSource connector reads /tmp/customers.txt and writes to the customers topic. Because one file cannot be scaled across multiple tasks, its configuration sets "tasks.max": "1". The connector can then be submitted to Kafka Connect with a REST POST request. section 11.2.2
Twelve tables across four tasks
If a database connector imports 12 tables with tasks.max set to 4, Kafka Connect can assign three tables to each task and distribute those tasks across available workers. Adding workers increases placement capacity, but does not create more tasks than the connector configuration and source structure allow. section 11.3
tasks.max is a ceiling, not a promise. The connector may create fewer tasks because of its own implementation or because the data offers limited parallelism, such as too few source-topic partitions. Setting the value above that limit adds no parallelization. section 11.6.1
