Dockerfile for Greenplum SNE 4.2.6.1

Dockerfile

FROM centos:6.6
MAINTAINER "Francisco Lopez" teraflopx@gmail.com

# update & upgrade packages
RUN yum update -y & yum upgrade -y

# install extra packages and basic tools (included some required for Greenplum installer)
RUN yum install -y epel-release git unzip which tar sed wget curl nano expect

# cleanup packages
RUN yum clean -all

# set password to SSH with root
RUN echo root:docker | chpasswd

# create admin user group and admin user 
RUN groupadd -g 8000 gpadmin
RUN useradd -m -s /bin/bash -d /home/gpadmin -g gpadmin -u 8000 gpadmin
RUN echo "greenplum" | passwd gpadmin --stdin
 
# create data directories (master and 2 segments)
RUN mkdir -p /data/gpmaster /data/gpseg1 /data/gpseg2 /data/greenplum
RUN chown -R gpadmin:gpadmin /data

# install Greenplum
COPY bin/greenplum-db-4.2.6.1-build-1-RHEL5-x86_64.zip greenplum-db-4.2.6.1-build-1-RHEL5-x86_64.zip
RUN unzip greenplum-db-4.3.2.0-build-1-RHEL5-x86_64.zip -d /

COPY bin/greenplum.sh greenplum.sh
RUN sh greenplum.sh

# housekeeping (remove non necessary files)
# TO BE COMPLETED

# configure system
COPY config/sysctl.conf /tmp/sysctl.conf
RUN cat /tmp/sysctl.conf >> /etc/sysctl.conf
# next line is not working; it seems the only option is tuning the kernel outside this Dockerfile (using docker run -w)
RUN sysctl -p /etc/sysctl.conf

COPY config/limits.conf /tmp/limits.conf
RUN cat /tmp/limits.conf >> /etc/security/limits.conf

USER gpadmin

# configure Greenplum environment
RUN echo "source /data/greenplum/greenplum_path.sh" > /home/gpadmin/.bash_profile
RUN echo "export MASTER_DATA_DIRECTORY=/data/gpmaster/gpsne-1" >> /home/gpadmin/.bash_profile 
#RUN source /home/gpadmin/.bash_profile

RUN echo "$(hostname)" > /home/gpadmin/sne_hostlist
# next line is not working; why?
RUN gpssh-exkeys -f /home/gpadmin/sne_hostlist
#RUN gpcheckos -f /home/gpadmin/sne_hostlist

COPY config/sne_gpinit /home/gpadmin/sne_gpinit
RUN gpinitsystem -c /home/gpadmin/sne_gpinit

# expose ports and data folder
EXPOSE 22 5432
VOLUME ["/data"]

# start Greenplum
ENTRYPOINT ["/bin/bash"]
CMD ["gpstart"]

Note: Greenplum is not open source but the installer can be downloaded from https://network.pivotal.io/products/pivotal-gpdb for testing purposes. It requires a user account to be created in Pivotal website.

At the moment I am stuck with two problems:

• Kernel tuning (should I use 'docker run -w'? should I tune this in the Vagrant host instead?)
• SSH key exchange (not sure what´s going on here yet)

Hi teraflopx,

you seem to be the only one (over the internet) who tried to create a docker image for greenplum.
Have you succeed? Or this is the final version of the dockerfile that you have?

Greenplum is now open source, Maybe this is best time to dockerize it?

Hi guys,

Hey, I have been able to "dockerize" Greenplum database in "singlenode" mode, just for testing purposes. Thanks for your Dockerfile, it was a light in the darkness to me, when I started last week.

Here I have placed the "Dockerfile build" repo on Github:
https://github.com/kevinmtrowbridge/greenplumdb_singlenode_docker

You'll notice some gross hacks in my version (for example, this choice bit: su gpadmin -l -c "gpstart -a --verbose" && sleep 86400 # HACK: it's difficult to get Docker to attach to the GPDB process(es) ... so, instead attach to process "sleep for 1 day") ... if anyone manages to figure out more elegant ways to do these things, I would love to hear about it! (I initially tried using "supervisord" but, it was just too much work for what I needed this for ...).

Best,
Kevin Trowbridge