expose zookeeper configuration and kafka zookeeper.connect for proper clustering

[erulabs]

Hello!

Love the image for development, but would like to expand it into a full blown production-ready image! I know we're a bit of distance away from that, but figured I'd rather contribute with you fine folks than to set out alone.

What I've done is expose the zookeeper configuration file by wrapping it in a start-zookeeper.sh (much like your start-kafka.sh) which reads in environment values and sets the configuration. It also allows the "zookeeper.connect" flag to be set in Kafka's configuration. I also moved the KAFKA_HOME environment variable into that script and merged the ENV docker layers.

I'll be continuing to work on this branch, so feel free to leave unmerged until I can add some documentation and actually test it across multiple hosts :)

Any thoughts about how I've handled ZK's configuration are welcome - I hesitate to do differently than the start-kafka.sh style, but I didn't want to write out every single possible ZK config option either. This feels like an OK middle ground to me, especially considering the very small number of default options ZK has.

Anyways, thanks again!

[zoidyzoidzoid]

This should be sed -r -i "s/(zookeeper.connect)=(.*)/\1=${ZOOKEEPER_SERVERS}/g" $KAFKA_HOME/config/server.properties

Or it is replacing the full match, not just the second group.

[gg7]

I hesitate to do differently than the start-kafka.sh style, but I didn't want to write out every single possible ZK config option either.

Relevant pull request: #65.