Travis

Concourse

---

The Greenplum Database (GPDB) is an advanced, fully featured, open source data warehouse. It provides powerful and rapid analytics on petabyte scale data volumes. Uniquely geared toward big data analytics, Greenplum Database is powered by the world’s most advanced cost-based query optimizer delivering high analytical query performance on large data volumes.

The Greenplum project is released under the Apache 2 license. We want to thank all our current community contributors and are really interested in all new potential contributions. For the Greenplum Database community no contribution is too small, we encourage all types of contributions.

Overview

A Greenplum cluster consists of a master server, and multiple segment servers. All user data resides in the segments, the master contains only metadata. The master server, and all the segments, share the same schema.

Users always connect to the master server, which divides up the query into fragments that are executed in the segments, sends the fragments to the segments, and collects the results.

Building Greenplum Database with GPORCA

Currently GPDB assumes ORCA libraries and headers are available in the targeted system and tries to build with ORCA by default. For your convenience, here are the steps of how to build the optimizer. For the most up-to-date way of building, see the README at the following repositories:

1. https://github.com/greenplum-db/gp-xerces
2. https://github.com/greenplum-db/gporca

Build the optimizer

1. Install our patched version of Xerces-C

   git clone https://github.com/greenplum-db/gp-xerces
   mkdir gp-xerces/build
   cd gp-xerces/build
   ../configure
   make install
   

2. ORCA requires CMake and Ninja, make sure you have them installed. Installation instructions vary, please check the CMake and Ninja websites.

3. Install ORCA, the query optimizer:

   git clone https://github.com/greenplum-db/gporca
   mkdir gporca/build
   cd gporca/build
   cmake -GNinja ..
   ninja install
   

   Note: Get the latest ORCA git pull --ff-only if you see an error message like below:

   checking Checking ORCA version... configure: error: Your ORCA version is expected to be 2.33.XXX
   

Install needed python modules

Add the following Python modules (2.7 & 2.6 are supported)

• psutil
• lockfile (>= 0.9.1)
• paramiko
• setuptools

If necessary, upgrade modules using "pip install --upgrade". pip should be at least version 7.x.x.

Build the database

# Clean environment
make distclean

# Configure build environment to install at /usr/local/gpdb
./configure --with-perl --with-python --with-libxml --prefix=/usr/local/gpdb

# Compile and install
make
make install

# Bring in greenplum environment into your running shell
source /usr/local/gpdb/greenplum_path.sh

# Start demo cluster (gpdemo-env.sh is created which contain
# __PGPORT__ and __MASTER_DATA_DIRECTORY__ values)
cd gpAux/gpdemo
make create-demo-cluster
source gpdemo-env.sh

Compilation can be sped up with parallelization. Instead of make, consider:

make -j8

The directory and the TCP ports for the demo cluster can be changed on the fly. Instead of make cluster, consider:

DATADIRS=/tmp/gpdb-cluster MASTER_PORT=15432 PORT_BASE=25432 make cluster

The TCP port for the regression test can be changed on the fly:

PGPORT=15432 make installcheck-world

Once build and started, run psql and check the GPOPT (e.g. GPORCA) version:

select gp_opt_version();

To turn ORCA off and use legacy planner for query optimization:

set optimizer=off;

Running tests

• The default regression tests

make installcheck-world

• The top-level target installcheck-world will run all regression tests in GPDB against the running cluster. For testing individual parts, the respective targets can be run separately.

• The PostgreSQL check target does not work. Setting up a Greenplum cluster is more complicated than a single-node PostgreSQL installation, and no-one's done the work to have make check create a cluster. Create a cluster manually or use gpAux/gpdemo/ (example below) and run the toplevel make installcheck-world against that. Patches are welcome!

• The PostgreSQL installcheck target does not work either, because some tests are known to fail with Greenplum. The installcheck-good schedule in src/test/regress excludes those tests.

• When adding a new test, please add it to one of the GPDB-specific tests, in greenplum_schedule, rather than the PostgreSQL tests inherited from the upstream. We try to keep the upstream tests identical to the upstream versions, to make merging with newer PostgreSQL releases easier.

Alternative Configurations

Building GPDB without GPORCA

Currently, GPDB is built with ORCA by default so latest ORCA libraries and headers need to be available in the environment. Build and Install the latest ORCA.

If you want to build GPDB without ORCA, configure requires --disable-orca flag to be set.

# Clean environment
make distclean

# Configure build environment to install at /usr/local/gpdb
./configure --disable-orca --with-perl --with-python --with-libxml --prefix=/usr/local/gpdb

Building GPDB with code generation enabled

To build GPDB with code generation (codegen) enabled, you will need cmake 2.8 or higher and a recent version of llvm and clang (include headers and developer libraries). Codegen utils is currently developed against the LLVM 3.7.X release series. You can find more details about the codegen feature, including details about obtaining the prerequisites, building and testing GPDB with codegen in the Codegen README.

In short, you can change the configure with additional option --enable-codegen, optionally giving the path to llvm and clang libraries on your system.

# Configure build environment to install at /usr/local/gpdb
# Enable CODEGEN
./configure --with-perl --with-python --with-libxml --enable-codegen --prefix=/usr/local/gpdb --with-codegen-prefix="/path/to/llvm;/path/to/clang"

Building GPDB with gpperfmon enabled

gpperfmon tracks a variety of queries, statistics, system properties, and metrics. To build with it enabled, change your configure to have an additional option --enable-gpperfmon

See more information about gpperfmon here

gpperfmon is dependent on several libraries like apr, apu, and libsigar

Development with Docker

We provide a docker image with all dependencies required to compile and test GPDB. You can view the dependency dockerfile at ./src/tools/docker/base/Dockerfile. The image is hosted on docker hub at pivotaldata/gpdb-devel. This docker image is currently under heavy development.

A quickstart guide to Docker can be found on the Pivotal Engineering Journal.

Known issues:

• The installcheck-world make target has at least 4 failures, some of which are non-deterministic

Running regression tests with Docker

1. Create a docker host with 8gb RAM and 4 cores

   docker-machine create -d virtualbox --virtualbox-cpu-count 4 --virtualbox-disk-size 50000 --virtualbox-memory 8192 gpdb
   eval $(docker-machine env gpdb)

2. Build your code on gpdb-devel rootfs

   cd [path/to/gpdb]
   docker build .
   # image beefc4f3 built

   The top level Dockerfile will automatically sync your current working directory into the docker image. This means that any code you are working on will automatically be built and ready for testing in the docker context

3. Log into docker image

   docker run -it beefc4f3

4. As gpadmin user run installcheck-world

   su gpadmin
   cd /workspace/gpdb
   make installcheck-world

Caveats

• No Space Left On Device On macOS the docker-machine vm can periodically become full with unused images. You can clear these images with a combination of docker commands.

  # assuming no currently running containers
  # remove all stopped containers from cache
  docker ps -aq | xargs -n 1 docker rm
  # remove all untagged images
  docker images -aq --filter dangling=true | xargs -n 1 docker rmi

• The Native macOS docker client available with docker 1.12+ (beta) or Community Edition 17+ may also work

Code layout

The directory layout of the repository follows the same general layout as upstream PostgreSQL. There are changes compared to PostgreSQL throughout the codebase, but a few larger additions worth noting:

• gpMgmt/

  Contains Greenplum-specific command-line tools for managing the cluster. Scripts like gpinit, gpstart, gpstop live here. They are mostly written in Python.

• gpAux/

  Contains Greenplum-specific extensions such as gpfdist and gpmapreduce. Some additional directories are submodules and will be made available over time.

• doc/

  In PostgreSQL, the user manual lives here. In Greenplum, the user manual is maintained separately and only the reference pages used to build man pages are here.

• gpdb-doc/

  Constains the Greenplum documentation in DITA XML format. Refer to gpdb-doc/README.md for information on how to build, and work with the documentation.

• ci/

  Contains configuration files for the GPDB continuous integration system.

• src/backend/cdb/

  Contains larger Greenplum-specific backend modules. For example, communication between segments, turning plans into parallelizable plans, mirroring, distributed transaction and snapshot management, etc. cdb stands for Cluster Database - it was a workname used in the early days. That name is no longer used, but the cdb prefix remains.

• src/backend/gpopt/

  Contains the so-called translator library, for using the ORCA optimizer with Greenplum. The translator library is written in C++ code, and contains glue code for translating plans and queries between the DXL format used by ORCA, and the PostgreSQL internal representation.

• src/backend/gp_libpq_fe/

  A slightly modified copy of libpq. The master node uses this to connect to segments, and to send fragments of a query plan to segments for execution. It is linked directly into the backend, it is not a shared library like libpq.

• src/backend/fts/

  FTS is a process that runs in the master node, and periodically polls the segments to maintain the status of each segment.

Contributing

Greenplum is maintained by a core team of developers with commit rights to the main gpdb repository on GitHub. At the same time, we are very eager to receive contributions from anybody in the wider Greenplum community. This section covers all you need to know if you want to see your code or documentation changes be added to Greenplum and appear in the future releases.

Getting started

Greenplum is developed on GitHub, and anybody wishing to contribute to it will have to have a GitHub account and be familiar with Git tools and workflow. It is also recommend that you follow the developer's mailing list since some of the contributions may generate more detailed discussions there.

Once you have your GitHub account, fork repository so that you can have your private copy to start hacking on and to use as source of pull requests.

Anybody contributing to Greenplum has to be covered by either the Corporate or the Individual Contributor License Agreement. If you have not previously done so, please fill out and submit the Contributor License Agreement. Note that we do allow for really trivial changes to be contributed without a CLA if they fall under the rubric of obvious fixes. However, since our GitHub workflow checks for CLA by default you may find it easier to submit one instead of claiming an "obvious fix" exception.

Coding guidelines

Your chances of getting feedback and seeing your code merged into the project greatly depend on how granular your changes are. If you happen to have a bigger change in mind, we highly recommend engaging on the developer's mailing list first and sharing your proposal with us before you spend a lot of time writing code. Even when your proposal gets validated by the community, we still recommend doing the actual work as a series of small, self-contained commits. This makes the reviewer's job much easier and increases the timeliness of feedback.

When it comes to C and C++ parts of Greenplum, we try to follow PostgreSQL Coding Conventions. In addition to that we require that:

• All Python code passes Pylint
• All Go code is formatted according to gofmt

We recommend using git diff --color when reviewing your changes so that you don't have any spurious whitespace issues in the code that you submit.

All new functionality that is contributed to Greenplum should be covered by regression tests that are contributed alongside it. If you are uncertain on how to test, or document your work, please raise the question on the gpdb-dev mailinglist and the developer community will do its best to help you.

Testing guidelines

At the very minimum you should always be running make installcheck-world to make sure that you're not breaking anything.

Changes applicable to upstream PostgreSQL

If the change you're working on touches functionality that is common between PostgreSQL and Greenplum, you may be asked to forward-port it to PostgreSQL. This is not only so that we keep reducing the delta between the two projects, but also so that any change that is relevant to PostgreSQL can benefit from a much broader review of the upstream PostgreSQL community. In general, it is a good idea to keep both code bases handy so you can be sure whether your changes may need to be forward-ported.

Submission timing

To improve the odds of the right discussion of your patch or idea happening, pay attention to what the community work cycle is. For example, if you send in a brand new idea in the beta phase, don't be surprised if no one is paying attention because we are focused on release work. Come back when the beta is done, please!

You can read more on Greenplum release policy and timing in the RELEASE.md

Patch submission

Once you are ready to share your work with the Greenplum core team and the rest of the Greenplum community, you should push all the commits to a branch in your own repository forked from the official Greenplum and send us a pull request.

For now, we require all pull requests to be submitted against the main master branch, but over time, once there are many supported open source releases of Greenplum in the wild, you may decide to submit your pull requests against an active release branch if the change is only applicable to a given release.

Validation checks and CI

Once you submit your pull request, you will immediately see a number of validation checks performed by our automated CI pipelines. There also will be a CLA check telling you whether your CLA was recognized. If any of these checks fails, you will need to update your pull request to take care of the issue. Pull requests with failed validation checks are very unlikely to receive any further peer review from the community members.

Keep in mind that the most common reason for a failed CLA check is a mismatch between an email on file and an email recorded in the commits submitted as part of the pull request.

If you can not figure out why a certain validation check failed, feel free to ask on the developer's mailing list, but make sure to include a direct link to a pull request in your email.

Patch review

A submitted pull request with passing validation checks is assumed to be available for peer review. Peer review is the process that ensures that contributions to Greenplum are of high quality and align well with the road map and community expectations. Every member of the Greenplum community is encouraged to review pull requests and provide feedback. Since you don't have to be a core team member to be able to do that, we recommend following a stream of pull reviews to anybody who's interested in becoming a long-term contributor to Greenplum. As Linus would say "given enough eyeballs, all bugs are shallow".

One outcome of the peer review could be a consensus that you need to modify your pull request in certain ways. GitHub allows you to push additional commits into a branch from which a pull request was sent. Those additional commits will be then visible to all of the reviewers.

A peer review converges when it receives at least one +1 and no -1s votes from the participants. At that point you should expect one of the core team members to pull your changes into the project.

Greenplum prides itself on being a collaborative, consensus-driven environment. We do not believe in vetoes and any -1 vote casted as part of the peer review has to have a detailed technical explanation of what's wrong with the change. Should a strong disagreement arise it may be advisable to take the matter onto the mailing list since it allows for a more natural flow of the conversation.

At any time during the patch review, you may experience delays based on the availability of reviewers and core team members. Please be patient. That being said, don't get discouraged either. If you're not getting expected feedback for a few days add a comment asking for updates on the pull request itself or send an email to the mailing list.

Direct commits to the repository

On occasion you will see core team members committing directly to the repository without going through the pull request workflow. This is reserved for small changes only and the rule of thumb we use is this: if the change touches any functionality that may result in a test failure, then it has to go through a pull request workflow. If, on the other hand, the change is in the non-functional part of the code base (such as fixing a typo inside of a comment block) core team members can decide to just commit to the repository directly.

Glossary

• QD

  Query Dispatcher. A synonym for the master server.

• QE

  Query Executor. A synonym for a segment server.

Documentation

For Greenplum Database documentation, please check online docs: http://greenplum.org/docs/

For further information beyond the scope of this README, please see our wiki

There is also a Vagrant-based quickstart guide for developers in src/tools/vagrant/README.md.