[KAFKA-9794] JMX metrics produce higher memory and CPU consumption in Kafka docker

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 2.2.0
• Component/s: metrics
• Labels: performance
• Environment: kafka version: kafka_2.12-2.2.0.tgz using docker
• Assignee: Unassigned
• Reporter: Hariprasad
• Created: 01/Apr/20 06:10
• Updated: 15/Jul/20 11:47

Description

We have implemented the Kafka metrics using this github: https://github.com/oded-dd/prometheus-jmx-kafka

we have enabled the metrics using KAFKA_OPTS. export KAFKA_OPTS='-javaagent:/etc/kafka/jmx_prometheus_javaagent-0.3.1.jar=9097:/etc/kafka/kafka-jmx-metrics.yaml'

After starting the Kafka with this, we have seen high usage in CPU and memory shoot up in the docker

CPU cycles goes some times as up as 300% , we are running with 4 cpu's  Kindly suggest some solution how to fix this.