Fluo 1.2.0 documentation  >>  Administration  >>  Run Fluo in Docker

Obtain the Docker image

To obtain the docker image created by this project, you can either pull it from DockerHub at apache/fluo or build it yourself. To pull the image from DockerHub, run the command below:

docker pull apache/fluo

While it is easier to pull from DockerHub, the image will default to the software versions below:

If these versions do not match what is running on your cluster, you should consider building your own image with matching versions.

Image basics

The entrypoint for the Fluo docker image is the fluo script. While the primary use case for this image is to start an oracle or worker, you can run other commands in the fluo script to test out the image:

# No arguments prints Fluo command usage
docker run apache/fluo
# Print Fluo version
docker run apache/fluo version
# Print Fluo classpath
docker run apache/fluo classpath

This Docker image only runs Fluo, it does not run Accumulo, Hadoop, or ZooKeeper processes. These systems must be run separately and the Docker image configured to use them.

Initialize application

Before starting a Fluo oracle and worker using the Fluo Docker image, initialize your Fluo application. Initialization can also be done using Docker. To do this, the applications jars and config file need to be mapped into the docker container. To understand how to do this, assume the following setup for local files outside of Docker.

• Application jars in /home/user1/myapp/lib
• fluo-app.properties in /home/user1/myapp/conf
• fluo.observer.init.dir=/opt/myapp/lib is set in fluo-app.properties along with any other properties needed

With this setup, the following Docker command will initialize myapp. Fluo init runs in docker as root, so HADOOP_USER_NAME is set in order to avoid copying into HDFS as root. The Docker run option -v /home/user1/myapp:/opt/myapp maps the local directory /home/user1/myapp into the container at /opt/myapp.

docker run --network="host" -v /home/user1/myapp:/opt/myapp \
           -e HADOOP_USER_NAME=`whoami` \
           apache/fluo init -a myapp -f -p /opt/myapp/conf/fluo-app.properties \
                            -o fluo.connection.zookeepers=zkhost/fluo

Run application in Docker

After initializing, choose a method below to run the oracle and worker(s) of your Fluo application. In the examples below, the Fluo application is named myapp and was initialized using a Zookeeper node on zkhost.

Docker engine

Use the docker command to start local docker containers.

1. Start a Fluo oracle for myapp application connecting to a Zookeeper at zkhost.

    docker run -d --network="host" apache/fluo oracle -a myapp -o fluo.connection.zookeepers=zkhost/fluo
   

   The command above uses -d to run the container in the background and --network="host" to run the container on the host’s network stack.

2. Start Fluo worker(s). Execute this command multiple times to start multiple workers.

    docker run -d --network="host" apache/fluo worker -a myapp -o fluo.connection.zookeepers=zkhost/fluo
   

Marathon

Using the Marathon UI, you can create applications using JSON configuration.

The JSON below can be used to start a Fluo oracle.

{
  "id": "myapp-fluo-oracle",
  "cmd": "fluo oracle -a myapp -o fluo.connection.zookeepers=zkhost/fluo",
  "cpus": 1,
  "mem": 256,
  "disk": 0,
  "instances": 1,
  "container": {
    "docker": {
      "image": "apache/fluo",
      "network": "HOST"
    },
    "type": "DOCKER"
  }
}

The JSON below can be used to start Fluo worker(s). Modify instances to start multiple workers.

{
  "id": "myapp-fluo-worker",
  "cmd": "fluo worker -a myapp -o fluo.connection.zookeepers=zkhost/fluo",
  "cpus": 1,
  "mem": 512,
  "disk": 0,
  "instances": 1,
  "container": {
    "docker": {
      "image": "apache/fluo",
      "network": "HOST"
    },
    "type": "DOCKER"
  }
}

Next Steps

Learn how to manage Fluo applications.