I installed a preview version of Kafka 3.0 with log4j2 support (http://home.apache.org/~dongjin/post/apache-kafka-log4j2-support/) on a RHEL 8 server. Kafka and Zookeeper are running successfully as systemd user services. I set the environment variable KAFKA_LOG4J_OPTS="-Dlog4j.configurationFile=file:/home/username/kafka/bin/../config/log4j2.properties" in the systemd unit file so that log4j 2.17 ist used.
However, there is one strange thing: all log4j logs are stored in a directory literally called ${kafka.logs.dir} in the home directory. The garbage collector successfully stores logs in the correct log location ~/kafka/logs. So the directory ~/kafka/logs contains files like kafkaServer-gc.log.0.current, while the hilariously strange directory ~/${kafka.logs.dir} contains server.log, controller.log, etc.
When looking at the processes for kafka and zookeeper, I can see that they both have the argument -Dkafka.logs.dir=/home/username/kafka/bin/../logs. Shouldn't this define the environment variable kafka.logs.dir, which is used in the file config/log4j2.properties? How come that Zookeeper and Kafka apparently don't have access to this $kafka.logs.dir environment variable?