Tuning Apache Kafka Consumers to maximize throughput and reduce costs
Briefly

If the average fetch size (fetch-size-avg) is consistently around 1 MiB * number of partitions per fetch request, your consumer may be hitting the max.partition.fetch.bytes limit, impacting throughput due to more requests and round trip time constraints.
A reasonable fetch rate is around 1 to 5 requests per second per broker, exceeding this along with constrained fetch size can lead to reduced application throughput and increased Kafka Broker costs.
Read at New Relic
[
add
]
[
|
|
]