Docker combines an easy-to-use interface to Linux containers with easy-to-construct image files for those containers. In short, Docker launches very light weight virtual machines.
The Docker Container Executor (DCE) allows the YARN NodeManager to launch YARN containers into Docker containers. Users can specify the Docker images they want for their YARN containers. These containers provide a custom software environment in which the user’s code runs, isolated from the software environment of the NodeManager. These containers can include special libraries needed by the application, and they can have different versions of Perl, Python, and even Java than what is installed on the NodeManager. Indeed, these containers can run a different flavor of Linux than what is running on the NodeManager – although the YARN container must define all the environments and libraries needed to run the job, nothing will be shared with the NodeManager.
Docker for YARN provides both consistency (all YARN containers will have the same software environment) and isolation (no interference with whatever is installed on the physical machine).
Docker Container Executor runs in non-secure mode of HDFS and YARN. It will not run in secure mode, and will exit if it detects secure mode.
The DockerContainerExecutor requires Docker daemon to be running on the NodeManagers, and the Docker client installed and able to start Docker containers. To prevent timeouts while starting jobs, the Docker images to be used by a job should already be downloaded in the NodeManagers. Here’s an example of how this can be done:
sudo docker pull sequenceiq/hadoop-docker:2.4.1
This should be done as part of the NodeManager startup.
The following properties must be set in yarn-site.xml:
<property>
<name>yarn.nodemanager.docker-container-executor.exec-name</name>
<value>/usr/bin/docker</value>
<description>
Name or path to the Docker client. This is a required parameter. If this is empty,
user must pass an image name as part of the job invocation(see below).
</description>
</property>
<property>
<name>yarn.nodemanager.container-executor.class</name>
<value>org.apache.hadoop.yarn.server.nodemanager.DockerContainerExecutor</value>
<description>
This is the container executor setting that ensures that all
jobs are started with the DockerContainerExecutor.
</description>
</property>
Administrators should be aware that DCE doesn’t currently provide user name-space isolation. This means, in particular, that software running as root in the YARN container will have root privileges in the underlying NodeManager. Put differently, DCE currently provides no better security guarantees than YARN’s Default Container Executor. In fact, DockerContainerExecutor will exit if it detects secure yarn.
By default, docker images are pulled from the docker public repository. The format of a docker image url is: username/image_name. For example, sequenceiq/hadoop-docker:2.4.1 is an image in docker public repository that contains java and hadoop.
If you want your own private repository, you provide the repository url instead of your username. Therefore, the image url becomes: private_repo_url/image_name. For example, if your repository is on localhost:8080, your images would be like: localhost:8080/hadoop-docker
To connect to a secure docker repository, you can use the following invocation:
docker login [OPTIONS] [SERVER]
Register or log in to a Docker registry server, if no server is specified
"https://index.docker.io/v1/" is the default.
-e, --email="" Email
-p, --password="" Password
-u, --username="" Username
If you want to login to a self-hosted registry you can specify this by adding the server name.
docker login <private_repo_url>
This needs to be run as part of the NodeManager startup, or as a cron job if the login session expires periodically. You can login to multiple docker repositories from the same NodeManager, but all your users will have access to all your repositories, as at present the DockerContainerExecutor does not support per-job docker login.
Currently you cannot configure any of the Docker settings with the job configuration. You can provide Mapper, Reducer, and ApplicationMaster environment overrides for the docker images, using the following 3 JVM properties respectively(only for MR jobs):
mapreduce.map.env: You can override the mapper’s image by passing yarn.nodemanager.docker-container-executor.image-name=your_image_name to this JVM property.
mapreduce.reduce.env: You can override the reducer’s image by passing yarn.nodemanager.docker-container-executor.image-name=your_image_name to this JVM property.
yarn.app.mapreduce.am.env: You can override the ApplicationMaster’s image by passing yarn.nodemanager.docker-container-executor.image-name=your_image_name to this JVM property.
The Docker Images used for YARN containers must meet the following requirements:
The distro and version of Linux in your Docker Image can be quite different from that of your NodeManager. (Docker does have a few limitations in this regard, but you’re not likely to hit them.) However, if you’re using the MapReduce framework, then your image will need to be configured for running Hadoop. Java must be installed in the container, and the following environment variables must be defined in the image: JAVA_HOME, HADOOP_COMMON_PATH, HADOOP_HDFS_HOME, HADOOP_MAPRED_HOME, HADOOP_YARN_HOME, and HADOOP_CONF_DIR
The following example shows how to run teragen using DockerContainerExecutor.
Step 1. First ensure that YARN is properly configured with DockerContainerExecutor(see above).
<property>
<name>yarn.nodemanager.docker-container-executor.exec-name</name>
<value>docker -H=tcp://0.0.0.0:4243</value>
<description>
Name or path to the Docker client. The tcp socket must be
where docker daemon is listening.
</description>
</property>
<property>
<name>yarn.nodemanager.container-executor.class</name>
<value>org.apache.hadoop.yarn.server.nodemanager.DockerContainerExecutor</value>
<description>
This is the container executor setting that ensures that all
jobs are started with the DockerContainerExecutor.
</description>
</property>
Step 2. Pick a custom Docker image if you want. In this example, we’ll use sequenceiq/hadoop-docker:2.4.1 from the docker hub repository. It has jdk, hadoop, and all the previously mentioned environment variables configured.
Step 3. Run.
hadoop jar $HADOOP_PREFIX/share/hadoop/mapreduce/hadoop-mapreduce-examples-2.8.4.jar \
teragen \
-Dmapreduce.map.env="yarn.nodemanager.docker-container-executor.image-name=sequenceiq/hadoop-docker:2.4.1" \
-Dyarn.app.mapreduce.am.env="yarn.nodemanager.docker-container-executor.image-name=sequenceiq/hadoop-docker:2.4.1" \
1000 \
teragen_out_dir
Once it succeeds, you can check the yarn debug logs to verify that docker indeed has launched containers.