This small repo helps you to execute custom ZooKeeper version on docker, by mounting your zookeeper local build to standard openjdk images. The current implementation will extract and use the zookeeper-assembly/target/*-bin.tar.gz file from your local git repository when starting the container.

The clusters are specified in compose files. You can customize them, these are the ones I created:

• 3_nodes_zk.yml: three zookeeper nodes, using a single virtual network
• 3_nodes_2_networks_zk.yml: three zookeeper nodes, using two virtual networks (created for testing ZOOKEEPER-3188)
• 4_nodes_2_networks_zk.yml: four zookeeper nodes, using two virtual networks (created for testing ZOOKEEPER-3188)
• 3_nodes_digets_quorum_auth_zk.yml: three zookeeper nodes, using a single virtual network and digest SASL authentication
• 3_nodes_zk_no_wildcard_addr.yml: three zookeeper nodes, having proper hosts in the server configs (not using 0.0.0.0 anywhere)
• 3_nodes_zk_jdk_12.yml: three zookeeper nodes on OpenJDK 12.0.2 (to reproduce issue in ZOOKEEPER-3769)
• 3_nodes_zk_dynamic_config.yml: three zookeeper nodes using separate dynamic config file supported in 3.6.0+ (to reproduce issue in ZOOKEEPER-3776)
• 3_nodes_zk_mounted_data_folder.yml: three zookeeper nodes storing all their config and data in a mounted volume

Some ports are also exposed on localhost, so you can connect to your cluster. My port configs (for server X):

• REST api port: 808(X) (e.g. for server 1 use: http://localhost:8081/commands/leader)
• Client port: 218(X) (e.g. for server 3 use: localhost:2183)

# the docker containers will mount both the zookeeper source code folder
# and the folder of the docker helper scripts, we need to define these folders first
export ZOOKEEPER_GIT_REPO=~/git/zookeeper-apache
export ZOOKEEPER_DOCKER_TEST_GIT_REPO=~/git/zookeeper-docker-test

# you always need to do a maven install to have the assembly tar.gz file updated!
cd $ZOOKEEPER_GIT_REPO
mvn clean install -DskipTests

cd $ZOOKEEPER_DOCKER_TEST_GIT_REPO
docker-compose --project-name zookeeper --file 3_nodes_zk.yml up

While the cluster is running, you can use the following commands on a different shell:

• list networks: docker network ls
• disconnect the 3rd container from network 1: docker network disconnect zookeeper_net_1 zookeeper_zoo3_1
• reconnect the same container later with the same IP: docker network connect --ip 172.16.101.33 zookeeper_net_1 zookeeper_zoo3_1
• restarting a single ZooKeeper server: docker-compose --file 3_nodes_zk.yml --project-name zookeeper restart zoo2
• stopping a single ZooKeeper server: docker-compose --file 3_nodes_zk.yml --project-name zookeeper stop zoo3
• starting a single ZooKeeper server: docker-compose --file 3_nodes_zk.yml --project-name zookeeper start zoo3
• use 4lw command 'stat' on the second ZK server (given it's ZK port is exposed on the port 2182 of the host machine): echo "stat" | nc localhost 2182

• first press ctrl-C on the shell where the zookeeper cluster is running
• then remove the networks / leftovers: docker-compose --file 3_nodes_zk.yml --project-name zookeeper down

To simulate the case of changing a hostname of a server with rolling restarts (without using dynamic reconfig, ZOOKEEPER-3814):

• clean the data folder
• start first all the 3 nodes of 3_nodes_zk_mounted_data_folder.yml
• then perform rolling restart by stopping nodes of 3_nodes_zk_mounted_data_folder.yml and starting nodes of 3_nodes_zk_server_hostname_changed.yml

Similarly, to expand a cluster with rolling restart (ZOOKEEPER-3829):

• clean the data folder
• start first all the 3 nodes of 3_nodes_zk_mounted_data_folder.yml
• then perform rolling restart by stopping nodes of 3_nodes_zk_mounted_data_folder.yml and starting nodes of 4_nodes_zk_mounted_data_folder.yml

In all rolling restart case, use docker-compose stop to terminate the old instances and use up to bring up the instances with new config.

# you can see the running containers:
docker ps

CONTAINER ID        IMAGE               COMMAND                  CREATED              STATUS              PORTS                                            NAMES
debdc72d864c        openjdk:8u222-jre   "/bin/bash /scripts/…"   About a minute ago   Up About a minute   0.0.0.0:2181->2181/tcp, 0.0.0.0:8081->8080/tcp   zookeeper_zoo1_1
f1702d83721c        openjdk:8u222-jre   "/bin/bash /scripts/…"   About a minute ago   Up About a minute   0.0.0.0:2183->2181/tcp, 0.0.0.0:8083->8080/tcp   zookeeper_zoo3_1
15dea744231d        openjdk:8u222-jre   "/bin/bash /scripts/…"   About a minute ago   Up About a minute   0.0.0.0:2182->2181/tcp, 0.0.0.0:8082->8080/tcp   zookeeper_zoo2_1


# connect to server 1:
docker exec -it zookeeper_zoo1_1 /bin/bash

root@zoo1:/# cd /zookeeper/bin/
root@zoo1:/zookeeper/bin# ./zkCli.sh

# or to start the zookeeper client on server 2 using a single command:
docker exec -it zookeeper_zoo1_1 /bin/bash /zookeeper/bin/zkCli.sh

Assuming you used 3_nodes_digest_quorum_auth.yml to start your cluster, you can start a cli by:

docker exec -e CLIENT_JVMFLAGS="-Djava.security.auth.login.config=/scripts/conf/digest_jaas_client.conf" -it zookeeper_zoo1_1 /bin/bash /zookeeper/bin/zkCli.sh

The containers will log with INFO threshold to the console and DEBUG logs will be produced into log files under the logs folder. You can change the log levels in the start_zookeeper.sh file.

Good old Apache License 2.0