Is there any other way to analyse internal Kafka metrics?

Question

In my case when i increasing no of consumer groups latency increasing in producer,consumer.
With JMX metrics and grafana i have analyzed the network thread, I/O threads
The network thread is idle most of the time, and the request and response queue not even filling up more than 10 where the size is 500.
With 10 I/O threads there is no delay in picking up request from request queue.

request_local_time_ms(disk write time),request_queue_time(Request waiting time in queue) are very low,
request_remote_time_ms(request waiting in network buffer) is equal to total time for produce request.

With that we can see some latency in network end.
But the network thread is most of the time idle.!
Any other help full metrics need to be monitor to find the bottleneck?