Kafka can increase the partition count of an existing topic. It cannot decrease that count in place. Existing records remain in their original partitions; Kafka does not automatically redistribute the historical log. Adding partitions can change where a keyed producer sends future records because the partition calculation uses the partition count. That can split the history of one key across old and new partitions. Read the ordering requirement before resizing a production topic. These exercises ask for an expansion that preserves the existing records, not a delete-and-recreate shortcut. Inspect the original log afterward. More partitions also mean more metadata, open files and recovery work.
Inspect the topics, partition logs and group offsets before changing the cluster.