CDH 5 pseudo-distributed cluster Docker image

Do you develop Hadoop mapreduce applications on top of Cloudera distribution? This docker image can help you. It contains basic CDH 5 setup with YARN. You can use it for developmeent and verification of your code in local environment without messing up your system with Hadoop instalation.

Docker image was prepared according to Installing CDH 5 with YARN on a Single Linux Node in Pseudo-distributed mode with a few adjustments for Docker environment.

Installed services

• HDFS
• YARN
• JobHistoryServer
• Oozie
• Hue
• Spark (installation for execution on top of YARN)

Execution

Get docker image

docker pull chalimartines/cdh5-pseudo-distributed

Run image with specified port mapping

docker run --name cdh -d -p 8020:8020 -p 50070:50070 -p 50010:50010 -p 50020:50020 -p 50075:50075 -p 8030:8030 -p 8031:8031 -p 8032:8032 -p 8033:8033 -p 8088:8088 -p 8040:8040 -p 8042:8042 -p 10020:10020 -p 19888:19888 -p 11000:11000 -p 8888:8888 -p 18080:18080 -p 9999:9999 chalimartines/cdh5-pseudo-distributed

Or you can use docker-compose configuration from here

If you are Mac OS user with docker machine with virtualbox driver and you would like to get from your local system to a cdh container add these port forwardings. Name of virtual machine is equal to your docker machine name (here it is "dev").

VBoxManage modifyvm "dev" --natpf1 "tcp-port8020,tcp,,8020,,8020"
VBoxManage modifyvm "dev" --natpf1 "tcp-port50070,tcp,,50070,,50070"
VBoxManage modifyvm "dev" --natpf1 "tcp-port50010,tcp,,50010,,50010"
VBoxManage modifyvm "dev" --natpf1 "tcp-port50020,tcp,,50020,,50020"
VBoxManage modifyvm "dev" --natpf1 "tcp-port50075,tcp,,50075,,50075"
VBoxManage modifyvm "dev" --natpf1 "tcp-port8030,tcp,,8030,,8030"
VBoxManage modifyvm "dev" --natpf1 "tcp-port8031,tcp,,8031,,8031"
VBoxManage modifyvm "dev" --natpf1 "tcp-port8032,tcp,,8032,,8032"
VBoxManage modifyvm "dev" --natpf1 "tcp-port8033,tcp,,8033,,8033"
VBoxManage modifyvm "dev" --natpf1 "tcp-port8088,tcp,,8088,,8088"
VBoxManage modifyvm "dev" --natpf1 "tcp-port8040,tcp,,8040,,8040"
VBoxManage modifyvm "dev" --natpf1 "tcp-port8042,tcp,,8042,,8042"
VBoxManage modifyvm "dev" --natpf1 "tcp-port10020,tcp,,10020,,10020"
VBoxManage modifyvm "dev" --natpf1 "tcp-port19888,tcp,,19888,,19888"
VBoxManage modifyvm "dev" --natpf1 "tcp-port11000,tcp,,11000,,11000"
VBoxManage modifyvm "dev" --natpf1 "tcp-port8888,tcp,,8888,,8888"
VBoxManage modifyvm "dev" --natpf1 "tcp-port9999,tcp,,9999,,9999"
VBoxManage modifyvm "dev" --natpf1 "tcp-port18080,tcp,,18080,,18080"

UI entry points

Those urls consider port forwarding from localhost.

• name node - http://localhost:50070
• resource manager - http://localhost:8088
• job history server - http://localhost:19888
• oozie console - http://localhost:11000
• hue - http://localhost:8888
• spark history server - http://localhost:18080

Hue login

You will be asked to create account during the first login. You can pick your prefered username and password. It will create home folder on HDFS and it can be used as hadoop user.

Mac OS and Docker in Virtualbox resource setting.

This docker image doesn't completely follow philosophy (one process = one image), but it prefers developer convenience to easily set up the whole hadoop dev stack. It comes with price and I recommend add resources to your virtualbox machine. I use 4 cores and 8GB RAM.

Custom port for your usecases

This image has exposed one port (9999). It is not used by any currently running service. It can be used by you for example when you need to attach debugger to running mapreduce job. So your mapreduce job can start debugging server on this port.