Demisto's Dockerfiles and Image Build Management

This repository's master branch tracks images pushed to the official Demisto Docker Hub organization at: https://hub.docker.com/u/demisto/. Other branches` images are pushed to devdemisto.

Note: We generate nightly information about packages and os dependencies used in each of Demisto's docker images. Checkout the dockerfiles-info project README for a full listing.

Contributing

Contributions are welcome and appreciated. To contribute follow the Getting Started section and submit a PR.

In order to conntribute, we will need all contributors to sign a contributor license agreement. By signing a contributor license agreement, we ensure that the community is free to use your contributions.

When opening a new pull request, a bot will evaluate whether you have signed the CLA. If required, the bot will comment on the pull request, including a link to accept the agreement. The CLA document is also available for review as a PDF. Visit our Frequently Asked Questions article for CLA related issues.

After opening your docker pull request, and in order for the reviewer to understand the context, make sure to link to the corresponding pull request from the Content repo where this docker image will be used.

Getting Started

Each docker image is managed in its own directory. The directory should be named the same as the image name (without the organization prefix). If needed, we prefer using a dash (-) as a separator in the name. All image directories are located under the docker dir.

The directory should contain one Dockerfile which will be used for building the docker image. Each image when it is built is tagged with the commit hash and version.

The script docker/build_docker.sh is used to build all modified docker images. The script detects modified directories by comparing against origin/master if on a branch or if on master by using the CIRCLE_COMPARE_URL environment variable to obtain the commit range of the current build.

Pre-requisites:

Install python 2 and 3 (so you can create both python 2 and 3 images):
- Mac: use brew to install (more info at: https://docs.brew.sh/Homebrew-and-Python): brew install python3 and then: brew install python@2
Or install pyenv (recommended for managing multiple python versions):
- Mac: brew install pyenv . Make sure to run then: pyenv init and follow instructions to add to either ~/.zshrc or ~/.bash_profile depending on you shell.
- Other: see https://github.com/pyenv/pyenv-installer
Install pipenv globally using: pip install pipenv
Install requests globally: pip install requests

To get up and running fast with a Python/PowerShell image with additional packages use the script: docker/create_new_docker_image.py. Usage:

usage: create_new_docker_image.py [-h] [-t {python,powershell}]
                                  [-p {two,three}] [-l {alpine,debian,ubuntu}]
                                  [--pkg PKG]
                                  name

Create a new docker image

positional arguments:
  name                  The image name to use without the organization prefix.
                        For example: ldap3. We use kebab-case naming
                        convention.

optional arguments:
  -h, --help            show this help message and exit
  -t {python,powershell}, --type {python,powershell}
                        Specify type of image to create (default: python)
  -p {two,three}, --python {two,three}
                        Specify python version to use (default: three)
  -l {alpine,debian,ubuntu}, --linux {alpine,debian,ubuntu}
                        Specify linux distro to use (default: alpine)
  --pkg PKG             Specify a package/module to install. Can be specified
                        multiple times. Each package needs to be specified
                        with --pkg. For example: --pkg google-cloud-storage
                        --pkg oath2client (default: None)

For example to create a new image named ldap using python 3 and with the python package ldap3 run the following:

./docker/create_new_docker_image.py -p three --pkg ldap3 ldap

The above command will create a directory docker/ldap with all relevant files all setup for building a docker image. You can now build the image locally by following: Building Locally a Test Build.

Note: for image names we use kebab-case naming convention.

Building Locally a Test Build

It is possible to run a local build to verify that the build process is working. Requirements:

Local install of docker
Local install of pipenv (if building an image which is managing packages via pipenv - recommended)

If you want to test how the script detects commit changes: Make sure you are working on a branch and the changes are committed. If you haven't committed the changes and want to run a local build you can run the script with a image name (which corresponds to a directory name) to the run the build on. For example:

./docker/build_docker.sh ldap

The above example will then run the build against the ldap directory.

When running locally, the script will then use a docker organization of devtesting and will tag the image with a testing tag and a version which has a timestamp as a revision. If you would like to test with a different organization name set the env variable: DOCKER_ORG. If you would like to test the push functionality set the env variable DOCKERHUB_USER. It is also possible to set DOCKERHUB_PASSWORD to avoid being prompted for the password during the build process.

Additionally, if you are working on multiple folders and would like to test only a specific one the script supports an env var of DOCKER_INCLUDE_GREP which will be used to do an extended grep to choose which directories to process.

Example for running with an org name of mytest and a grep extended expression which will process only the python dir (and not python3 dir):

DOCKER_ORG=mytest DOCKER_INCLUDE_GREP=/python$ docker/build_docker.sh

Build configuration

The build script will check for a build.conf file in the target image directory and will read from it name=value properties. Supported properties:

version: The version to use for tagging. Default: 1.0.0. Note: that additionally, the CircleCI build number is always appended to the version as a revision (for example: 1.0.0.15519) to create a unique version per build.
devonly: If set the image will be pushed only to the devdemisto org in docker hub and will not be pushed to the demisto org. Should be used for images which are for development purposes only (such as the image used in CircleCI to build this project).

Base Python Images

There are 4 base python images which should be used when building a new image which is based upon python:

python: Python 2 image based upon alpine
python3: Python 3 image based upon alpine
python-deb: Python 2 image based upon debian
python3-deb: Python 3 image based upon debian

Which image to choose as a base?

If you are using pure python dependencies then choose the alpine image with the proper python version which fits your needs (two or three). The alpine based images are smaller and recommended for use. If you require installing binaries or pre-compiled binary python dependencies (manylinux), you are probably best choosing the debian based images. See the following link: docker-library/docs#904 .

If you are using the python cryptography package we recommend using demisto/crypto as a base image. This base image takes care of properly installing the cryptography package. There is no need to include in the Pipenv file the cryptography package when using this image as a base.

Adding a `verify.py` script

As part of the build we support running a verify.py script in the created image. This allows you to add logic which tests and checks that the docker image built is matching what you expect.

Simply create a file named: vefify.py. It may contain any python code and all it needs is to exit with status 0 as a sign for success. Once the docker image is built, if the script is present it will be run within the image using the following command:

cat verify.py | docker run --rm -i <image_name> python '-'

Example of docker image with simple verify.py script can be seen here

PowerShell Images

We support building PowerShell Core docker images. To create the Dockerfile for a PowerShell image use the docker/create_new_docker_image.py script with the -t or --type argument set to: powershell. For example:

./docker/create_new_docker_image.py -t powershell --pkg Az pwsh-azure

The above command will create a directory docker/pwsh-azure with all relevant files setup for building a PowerShell docker image which imports the Az PowerShell module. You can now build the image locally by following: Building Locally a Test Build.

Naming Convention: To differentiate PowerShell images, name the images with a prefix of either pwsh- or powershell-.

Base PowerShell Images

There are 3 base PowerShell images which should be used when building a new image which is based upon PowerShell:

powershell: PowerShell image based upon Alpine (default)
powershell-deb: PowerShell image based upon Debian
powershell-ubuntu: PowerShell image based upon Ubuntu

We recommend using the default Alpine based image. The Debian and Ubuntu images are provided mainly for cases that there is need to install additional OS packages.

Adding a `verify.ps1` script

Similar to the the verify.py script for Python images, you can add a verify.ps1 script to test and check the image you created.

Once the docker image is built, if the script is present it will be run within the image using the following command:

cat verify.ps1 | docker run --rm -i <image_name> pwsh -c '-'

Docker Image Deployment

When you first open a PR, a development docker image is built (via CircleCI) under the devdemisto docker organization. So for example if your image is named ldap3 an image with the name devdemisto/ldap3 will be built.

If the PR is on a local branch of the dockerfiles github project (relevant only for members of the project with commit access), the image will be deployed to the devdemisto docker hub organization. A bot will add a comment to the PR stating that the image has been deployed and available. You can then test the image out simply by doing docker pull <image_name> (instructions will be included in the comment added to the PR).

If you are contributing (thank you!!) via an external fork, then the image built will not be deployed to docker hub. It will be available to download from the build artifacts. You can download the image and load it locally by running the docker load command. If you go into the build details in CircleCI you will see also instructions in the end of the Build Docker Images step on how to load it with a one liner bash command. Example contribution build can be seen here.

Once merged into master, CircleCI will run another build and create a production ready docker image which will be deployed at Docker Hub under the demisto organization. A bot will add a comment to the original PR about the production deployment and the image will then be fully available for usage. An example production comment added to a PR can be seen here.

Advanced

Support for Pipenv (Pipfile)

Our recommendation is to use Pipenv to manage python dependencies as it ensures that the build produces a deterministic list of python dependencies.

If a Pipfile is detected and a requirements.txt file is not present, the Pipfile will be used to generate a requirements.txt file before invoking docker build. The file is generated by running: pipenv lock -r. This allows the build process in the Dockerfile to simply install python dependencies via:

RUN pip install --no-cache-dir -r requirements.txt

Note: build will fail if a Pipfile is detected without a corresponding Pipfile.lock file.

Pipenv quick start:

If you want to use pipenv manually make sure you first have the pre-requisites installed as specified in Getting Started. Then follow:

In the relevant folder initialize the pipenv environment:
- python 2: PIPENV_MAX_DEPTH=1 pipenv --two
- python 3: PIPENV_MAX_DEPTH=1 pipenv --three
Install dependencies using: pipenv install <dependency>. For example: pipenv install requests
Make sure to commit both Pipfile and Pipfile.lock files
To see the locked dependencies run: pipenv lock -r

Installing a Common Dependency

If you want to install a new common dependency in all python base images use the script: install_common_python_dep.sh. Usage:

Usage: ./docker/install_common_python_dep.sh [packages]

Install a common python dependency in all docker python base images.
Will use pipenv to install the dependency in each directory.
Base images:
   python
   python3
   python-deb
   python3-deb

For example: ./docker/install_common_python_dep.sh dateparser

Note: By default pipenv will install the specified dependency and also update all other dependencies if possible. If you want to only install a dependency and not update the existing dependencies run the script with the env variable: PIPENV_KEEP_OUTDATED. For example:

PIPENV_KEEP_OUTDATED=true ./docker/install_common_python_dep.sh dateparser

Automatic updates via Dependabot

We use dependabot for automated dependency updates. When a new image is added to the repository there is need to add the proper config to .dependabot/config.yml. If you used the ./docker/create_new_python_image.py to create the docker image, then this config will be added automatically by the script. Otherwise, you will need to add the proper dependabot config. The build will fail without this config. You can add the dependabot config by running the script:

./docker/add_dependabot.sh <folder path to new docker image>

For example:

./docker/add_dependabot.sh docker/nmap

alannix-lw/dockerfiles