Here is a result of the impact of the new setting `buffer.memory` in this other
scenario:
Kafka cluster with 6 brokers
2 topics with 6 partitions each, one per broker
both topics well preloaded with messages of size ~100k
Demo app with 1 consumer subscribed to both topics running a loop like
```
while(true) {
consumer.poll(100L)
Thread.sleep(1000L)
}
```
and consumer config
```
"fetch.max.bytes"=10000000
"max.partition.fetch.bytes"=5000000
```
on Java 1.8.0_181 on MacOS
1) running this single consumer app without the memory pool (i.e. no
`buffer.memory` set)
required -Xmx200M -Xms200M to avoid an OOM error.
2) running this consumer with a memory pool defined by
"buffer.memory"=30000000
I could limit the JVM to -Xmx120M -Xms120M
[ Full content available at: https://github.com/apache/kafka/pull/4934 ]
This message was relayed via gitbox.apache.org for [email protected]
