JDP

Process and view data from many different sources. Regardless of how reliable those sources are or what algorithms and data structures you wish to use.

JDP can fit into a fully automated pipeline, be used interactively or both.

The broader aim is to make prototyping arbitrary reporting and inter-tool workflows cheap. So that we can perform many experiments with a high risk of failure. While incurring a limited cost for failed experiments and an unlimited payoff for successful ones.

  • JDP processes can be incrementally developed in an interactive, graphical environment (Jupyter).
  • JDP may be used as a library in a larger project.
  • JDP has a distributed, highly-available data cache.
  • JDP is not a polished product for non-technical users, but you can use it to make that.
  • JDP makes data from multiple sources/trackers easily accessible, but it is not a source of truth.
  • JDP can post back to trackers; it can automate workflows other than reporting.
  • JDP is intended to fit into a CI/CD pipeline or take over unusual sections of a pipeline, it is not intended as a replacement for Jenkins, GoCD, GitlabCI, etc.

Initially JDP is targeted at SUSE's QA Kernel & Networking team's requirements, however it is a general purpose tool at its core. It could be used with any type of data for most any type of workflow or reporting.

Video Presentation (external)

Tip

This README is best viewed through the docs site (public mirror). Otherwise admonition blocks like this will be misinterpreted as literal blocks.

Install

I don't recommend using Docker for development or regular use on your workstation (see installing from Git). However it is very useful for deployment to an automated production environment.

Note

SUSE employees and associates should view this at: gitlab.suse.de/rpalethorpe/jdp

Docker

You can install using Docker by doing the following from the directory where you cloned this repo.

docker build -t jdp:latest -f install/Dockerfile .

Or you can substitute the build command for the following which will get a pre-built image from hub.docker.com (it may not be up to date).

docker pull suserichiejp/jdp:latest

Then you can inject the access details for a central data cache server if you have them.

docker build -t jdp:latest -f install/Dockerfile-slave \
             --build-arg REDIS_MASTER_HOST=ip-or-name \
             --build-arg REDIS_MASTER_AUTH=password .
Note

If you pulled from dockerhub (or wherever) then you will need to change the tag name to suserichiejp/jdp:latest (or whatever).

Then run it

docker run -it -p 8889:8889 jdp:latest

With a bit of luck you will see a message from Jupyter describing what to do next.

By default JDP will create a local Redis instance automatically inside the container. Redis will save its data within the Docker volume /home/jdp/data. Unless you mount this volume your data cache is likely to get deleted. It is also possible to configure JDP to connect to an existing Redis server; see conf/data.toml and install/Dockerfile-production.

You can use the Docker image for developing JDP itself by mounting the /home/jdp/src volume.

Git (from source)

Installing from source should be fairly easy, it just requires a few none Julia based dependencies. What you need depends on the trackers you intend to use and whether you want to use Jupyter.

Tip

You can also use install/Dockerfile(-base) as a guide. Also check conf/*.toml.

You must install Redis unless you provide a remote Redis address in conf/data.toml.

Generally speaking, if you want to use Jupyter 'notebooks', then you should install Jupyter notebook and client packages. However there are alternatives to Jupyter client, which may also load and edit Jupyter notebooks.

Also install the latest stable release of Julia. JDP currently bundles an 'upstream' version of Julia in the install directory. You will probably have difficulties using your distribution's Julia version.

If you wish to use the OpenQA integration then install openQA-client. Same goes for mailx.

Finally, if you installed Jupyter or equivalent, run julia install/install.jl to setup IJulia. Otherwise this is not necessary.

Usage

With Jupyter

If you are using the Docker image then just browse to localhost:8889. If not then run jupyter notebook in the JDP directory.

Open either the notebooks/Report-Status-Diff.ipynb or notebooks/Propagate Bug Tags.ipynb Jupyter notebooks which are (hopefully) self documenting. I have only tested them with Jupyter itself, but there are fancier alternatives such as JupyterLab.

Other

You can also use the library from a Julia REPL or another project. For example in a julia REPL you could run

include("src/init.jl")

using JDP: Repository, Trackers.OpenQA, ... etc.

Also the run directory contains scripts which are intended to automate various tasks. These can be executed with Julia in a similar way to julia run/all.jl.

Automation

JDP is automated using SUSE's internal Gitlab CI instance. Which automates building and testing the containers as well as deployment and the execution of various scripts/services. See install/gitlab-ci.*.

There is also a public pipeline on GitLab.com, which publishes the public documentation and reports.

Documentation

Further documentation can be found at palethorpe.gitlab.io/jdp or rpalethorpe.io.suse.de/jdp

You can also find documentation at the Julia REPL by typing ? followed by an identifier or in a notebook you can type @doc identifier in a code cell.

The following image may give you some intuition for what JDP is.

Outer Architecture

Contributors

Created and maintained by Richard Palethorpe (rpalethorpe@suse.com). Sebastian Chlad (schlad@suse.com) is mainly responsible for it being a serious (I hope) project.

Cyril has been asking for a result difference view and matrix for years.

Ideas and feedback

Because it is not obvious who has contributed non-code or documentation changes I will try to make a list. Please let me know if I have missed you out or want to be removed.

  • Sebastian Chlad
  • Cyril Chrubis
  • Yong Sun
  • Anton Smorodskyi
  • Sergio Lindo
  • Petr Vorel
  • Oliver Kurz
  • Clemans Famulla-Conrad
  • Jose Lausuch
  • Petr Cervinka

Code and documentation

See the github/lab stats.

  • 1In the sense that the JDP project comes bundled with some scripts for using it with Jupyter amongst other things.