Skip to content

pyATS utility: standardizing how everyone builds pyATS docker images containing custom scripts and libraries.

License

Notifications You must be signed in to change notification settings

CiscoTestAutomation/pyats-image-builder

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

pyATS Image Builder

pyATS image builder is a utility package aiming to standardize the building of pyATS test scripts and their corresponding environment dependencies into Docker images.

It does so by abstracting away the need to directly write Dockerfiles, and instead presents the common, boilerplate dependency handling paradigms into a simple to use YAML file.

In addition, this package helps conventional users make their scripts portable by leveraging the power of Docker, without requiring them to understand how the Docker image building process works.

No Docker expertise necessary - though, basic familiarity would help.

General Information

Requirements

Installation

To install this package, simply pip install it onto your server's Python environment.

bash$ pip install pyats-image-builder

Usage

This package does not require pyATS to be installed. It features its own command line interface, pyats-image-build.

However, if you do install this package into an existing pyATS virtual environment the primary pyats command will be automatically updated to include this package's functionality under pyats image build sub-command.

usage: pyats-image-build [-h] [--tag TAG] [--path PATH] [--no-cache]
                         [--keep-context] [--dry-run] [--verbose]
                         file

       pyats image build [-h] [--tag TAG] [--path PATH] [--push] [--no-cache]
                         [--keep-context] [--dry-run] [--verbose]
                         file

Create standard pyATS Docker images

positional arguments:
  file                  YAML build file describing the image build details.

optional arguments:
  -h, --help            show this help message and exit
  --tag TAG, -t TAG     Tag for docker image. Overrides any tag defined in the
                        yaml.
  --path PATH, -p PATH  Specify a path to use as the context directory used
                        for building Docekr image
  --push, -P            Push image to Dockerhub after buiding
  --no-cache, -c        Do not use any caching when building the image
  --keep-context, -k    Prevents the Docker context directory from being
                        deleted once the image is built
  --dry-run, -n         Set up the context directory but do not build the
                        image. Use with --keep-context.
  --verbose, -v         Prints the output of docker build

Basic Concepts

YAML Build File
The input file containing the build instructions and dependencies, in YAML format. See section below on syntax and feature support.
Image Name, Tag
See official documentation for details.
Build Context Directory
Folder that includes all the files necessary for the build process. See official documentation for further details.

YAML Build File

The package abstract away the need to write your own Dockerfile (and having to deal with dockerfile syntax). Instead, it uses a human-readable YAML syntax as input, performs the necessary actions such as:

  • copying files
  • cloning repositories
  • installing Python packages

in a standardized fashion, and ensures all built-images looks & feels similar.

Build File Syntax

tag: "mypyatsimage:latest"      # Docker name/tag for the image

python: 3.6.8                   # Your desired Python version

env:                            # environment variables to be set within the image
  "<name>": "<value>"           # <- format
  MY_VARIABLE: "my-value"       # <- example

files:                          # list of files from various sources to be copied into the image
                                # supports syntax for copying files from:
                                #   - localhost (this server)
                                #   - remote URL
                                # [Optional]

  - /path/to/file1              # copy a localhost file to /pyats/

  - myfile_2: /path/to/file2    # copy a localhost file to /pyats/ and renaming it

  - dirname/file3: /path/to/file3               # copy a localhost file to /pyats/ directory

  - "https://webaddress.com/path/to/file4"      # download remote file to /pyats/

  - myfile_5: "https://webaddress/path/to/file5"   # download remote file to /pyats/, and rename it

packages:                       # List of python packages to be installed into the virtual environment
  - pyats[full]
  - otherpackage1==1.0
  - otherpackage2==2.0

repositories:                   # Git repositories to clone and include in the image
                                # [Optional]

  "<name_of_repo>":             # name of the folder to clone to
    url: "ssh://git@address/path/to/repo.git"       # clone source URL
    commit_id: abcd1234                             # [Optional] Commit-id/branch to checkout after cloning
    ssh_key: "<private_ssh_key>"                    # [Optional] Private ssh key for private repositories

  dirname/repo2name:            # alternatively, you can also specify a sub-folder to clone to
    url: "https://address/path/to/repo2.git"
    credentials:                                    # [Optional] Git credentials for private repositories
        username: "<git_username>"
        password: "<git_password>"

jobfiles:                       # Additional criteria to consider in job discovery in the image
                                # [Optional]

  paths:                        # list of paths to the jobfiles
    - relative/path/to/job.py                       # relative path from /pyats
    - /pyats/absolute/path/to/job.py                # absolute path to job in image
  match:                        # list of regex expressions to match python jobfiles
    - .*job.py

proxy:                          # proxy variables - use this if your host server is behind a proxy
                                # (needed for pip installations using public PyPI servers)
  HTTP_PROXY: "http://127.0.0.1:1111"
  HTTPS_PROXY: "https://127.0.0.1:1111"
  FTP_PROXY: "ftp://127.0.0.1:1111"
  NO_PROXY: "*.test.example.com,.example2.com"

cmds:                           # Additional commands to be inserted into the Dockerfile
                                # To be executed before and after the pip install process
                                # WARNING: This can have unintended consequences.
                                # Only use if you are *absolutely* sure about what you are done.
                                # [Optional]
  pre: "dockercommand"          # Docker command(s) in string format, executed before pip installation
  post: "dockercommand"         # Docker command(s) in string format, executed after pip installation

pip-config:                     # Custom pip configuration values
  global:
    disable-pip-version-check: 1

tag

Docker name/tag to assign to this image after build finishes. This name must obey the official docker image naming convention.

You can override the provided name/tag using the --tag argument from the command line

python

Your desired Python version. The pyATS Image Builder builds from a base image of python:{version}-slim. The default version is 3.6.9.

Make sure your specified version exists at https://hub.docker.com/_/python

platform

Desired platform to build image for. Platform must obey (https://docs.docker.com/build/building/multi-platform/#building-multi-platform-images).

If an existing python-slim image exists on the host machine, this must be removed in order to use this option.

env

Environment variables to be defined in the image. These environment variables will persist in the built image - and visible in your pyATS job runs.

In addition to your custom ones, the builder automatically sets environment variable $WORKSPACE, typically referring /pyats directory. This can be used to dynamically reference files:

# cloning two repositories to workspace, and adding them to your PYTHONPATH
env:
  PYTHONPATH: ${WORKSPACE}/repo1dir/:${WORKSPACE}/repo2dir/
repositories:
  repo1:
    url: "ssh://git@address/path/to/a/repo.git"
  repo2:
    url: "ssh://git@address/path/to/another/repo.git"

files

Section to specify the list of files or folders to copy to /pyats. This section allows you to specify both localhost files and remote files to include in your build image, and as well give you a place to rename them on copy.

# Format
files:
    - <list of files>
    - <in yaml format>
    - <to copy>

List entires under files block supports a few different input formats:

  • /path/to/file: copies a this particular file from your host system to /pyats/file

  • new_name: /path/to/file: copies + rename file, to /pyats/new_name

  • new_dir/new_name: /path/to/file: copies + rename file to /pyats/new_dir/new_name

In addition, in addition to localhost files, you can also specify remote files by URL scheme:

  • scp://[user@]remotehost/path/to/file: SCP this file to /pyats/file
  • ftp://remotehost:2121/path/to/file: same as above, but this time using FTP
  • https://webaddress/path/to/file: same as above, using HTTPS (file-get)

You can also rename remote files, and move them into sub-directories:

files:
    - subdir/new_name: https://webaddress/path/to/file
    - new_ftp_file_name: ftp://remotehost:2121/path/to/file

SCP Limitations

  • pyATS Image Builder does not support any user interaction once building starts, so passwordless ssh authentication must be set up in advance in order to download files with scp.
  • Requires an absolute path to be parsed correctly.
  • Supports recursively copying entire directories.
  • Specifying the user in the URI is optional.
  • Port can be specified in the URI: scp://user@remotehost:23/path/to/file.

FTP Limitations

  • Only single files can be retrieved with ftp.
  • pyATS Image Builder uses the anonymous login for ftp, so the file must be accessible to anonymous users.

packages

List of packages and their corresponding versions to install. Similar to a pip freeze output, but on a line-by-line basis.

# Format
packages:
    - <name>
    - <name>==<version>

# Example
packages:
    - pyats[full]>=20.3
    - netmiko
    - ansible==2.9.7

This list also works with local wheel files, and supports the use of $WORKSPACE environment variable to reference them.

# Example:
# download a wheel file from a remote host, and install it using pip.
files:
  - "scp://[user@]remotehost/path/to/packagename.whl"
packages:
  - ${WORKSPACE}/packagename.whl

repositories

Git repositories to clone to this docker image. By default, each repo will be cloned to the provided name under /pyats. However, you may also specify a new subdirectory to home it in.

If your repository is private you may provide the credentials or ssh_key to gain access. You may also avoid including these private details in the yaml by passing them through host environment variable. Check the Yaml loader section for more details. Once the image is built, it will remove the login information.

# Format

repositories:
    <name>:
        url: <repository url>
        commit_id: <name of branch or commit-id to checkout after clone>

# Example:
repositories:
    #   equivalent to: git clone https://github.com/CiscoTestAutomation/examples /pyats/examples
    examples:
        url: https://github.com/CiscoTestAutomation/examples
        ssh_key: <id_rsa file contents>

    #   equivalent to: mkdir -p /pyats/solutions; git clone https://github.com/CiscoTestAutomation/examples /pyats/solutions/examples
    solutions/examples:
        url: https://github.com/CiscoTestAutomation/solution_examples
        credentials:
            username: <git username>
            password: <git password>

yaml loader

Host environment variables to be loaded into the build yaml. This provides a way to dynamically substitute variables into the yaml.

For example, a substitution of ssh key to a repository:

export example_ssh_key=<private_ssh_key>
# Example:
repositories:
    examples:
        url: https://github.com/CiscoTestAutomation/examples
        ssh_key: '%ENV{ example_ssh_key }'

jobfiles

pyATS Image Builder will attempt to discover all pyATS jobfiles within the image. Only files with a .py extension will be considered as potential jobfiles for discovery. To enable automatic discovery of your jobfile, include the keyword <PYATS_JOBFILE> as part of the module docstring within the first 10 lines of the file.

By default, all file names containing *job*.py will be included as a job file. This behavior is overwritten when you provide any match regex pattern.

For example, a jobfile should look similar to the following:

"""
my_jobfile.py
<PYATS_JOBFILE>

Description of this job.
"""
from pyats.easypy import run

...

You can also define your jobfiles within the build YAML file.

# Format
jobfiles:
  paths:
    - <path-to-jobfile>
    - <another-path-to-jobfile>
  match:
    - <regex-to-match>
    - <another-regex-to-match>

# Example
jobfiles:
  paths:
    - relative/path/to/job.py
    - /pyats/path/to/job.py
  match:
    - .*job.py
    - .*example_job.py

The list of paths also supports the use of the $WORKSPACE environment variable to define paths.

proxy

Proxy variables. Useful if your host server is sitting behind a network proxy, and you need to pull data/packages from public internet.

Supported keys:

  • HTTP_PROXY
  • HTTPS_PROXY
  • FTP_PROXY
  • NO_PROXY

cmds

Any additional docker command(s) in raw text format, to be inserted before/after the pip installation command in the build process. See Dockerfile template for where the pre_cmd and post_cmd are inserted.

Use cases example for cmds block: some packages require non-python dependencies (eg. gcc), which are unlikely to be included in the image since it is minimal. Use the cmds section to invoke apt-get command to install these dependencies.

pip-config

Pip configuration file. The content of this section gets converted to a pip.conf file used to customize your pip installation behavior.

For example, use this section to define your own PyPI server to download packages from.

This section is read as a dictionary, and parsed directly into pip.conf INI format without translation.

# Example
pip-config:
  global:
    trusted-host:
        - pypi.python.org

    index-url: https://pypi.org/simple
    no-cache-dir: True

Alternatively, you can also specify your pip-config block directly in string format - this will be taken directly and stored as the content of pip.conf:

pip-config: |
  [global]
  format = columns
  no-cache-dir = false
  trusted-host = pypi.python.org
  index-url = https://pypi.org/simple
  disable-pip-version-check = 1

  [search]
  index = https://pypi.org/simple

Image Layout

pyATS Docker images created using this package features the following directory structure:

/pyats
    Directory where the Python virtual environment is created. All Python
    packages (including pyATS) specified in the build YAML file are installed
    into here. Additionally acts as the workspace, where all files and
    repositories specified in the YAML build file gets copied to. Set as the
    Docker working directory.

/pyats/installation
    Files related to the building of this docker image is stored under here.
    (for bookkeeping and debugging)

/pyats/installation/build.yaml
    Copy of the input build YAML file.

/pyats/installation/requirements.txt
    Pip packages installed in the virtual environment in pip freeze format.

/pyats/installation/repos.json
    A mapping of all git repos in the image with the current checked out commit
    or branch.

/pyats/installation/jobfiles.txt
    List of all discovered pyATS jobfiles in the image.

/pyats/installation/manifest.json
    A mapping of all discovered manifest files with the contents of that file.

See more about manifest files.

Image Build

The image is built in two main stages.

  1. the builder parsers the input YAML file and sets up the build context on the local build machine in what's called a build context directory.

  2. Generates a Dockerfile, and launches docker build the directory.

Build Context Directory

When the builder starts up, it creates a temporary directory in your file system (eg, /tmp), used for storing artifacts necessary for your pyATS docker image build process:

  • file and git repositories defined in the build YAML file are copied/clones here

  • the pip package dependency list in the build YAML file is converted into a requirements.txt file here

  • if custom pip configuration is provided, a pip.conf file is generated here, customizing the pip installation behavior


Running Built Images

To run the newly generated image, do:

$ docker run [--rm] [-it] [-v LOCAL:CONTAINER] IMAGE [COMMAND]

Where [IMAGE] is the image tag or ID. --rm is an optional flag to remove the container once finished. -it are optional flags that allow the container to run interactively. -v LOCAL:CONTAINER is the optional argument that mounts a file or directory LOCAL to the specified location CONTAINER in the container. Both paths must be absolute. When no command is given, the container will default to starting a bash session.

To run a pyATS job, the command would look like:

$ docker run --rm myimg:latest pyats run job myrepo/myjob.py

In this case, the jobfile in question is $WORKSPACE/myrepo/myjob.py. The starting working directory of the image is $WORKSPACE which is why it does not need to be specified in the command.

It may be beneficial to set an environment variable with the location of a job file or any other information by defining it in the YAML file. These variables cannot be used directly on the command line since the host will attempt to interpolate variables before executing Docker. There are two methods to use environment variables in the Docker container.


Advaned Use Cases

Instructions File

Environment variables can be specified inside a small bash script without any issues. This script can then be mounted and executed inside the container. This has the additional benefit of allowing multiple commands to be passed without running the container interactively

Create a small bash script that has all the commands and variables.

run.sh:

echo "\$JOBFILE is $JOBFILE"
echo "\TESTBEDFILE is $TESTBEDFILE"
pyats run job $JOBFILE --testbed-file $TESTBEDFILE

Then mount and run this file when running the container.

$ docker run --rm -v $(pwd)/run.sh:/mnt/run.sh myimg:latest bash /mnt/run.sh

This mounts the local file run.sh to the location /mnt/run.sh in the container, and executes this file with bash.

A user created python file could also be mounted and run in the exact same manner.

$ docker run --rm -v $(pwd)/run.py:/mnt/run.py myimg:latest python /mnt/run.py

Bash Interpolation

Variables can be preserved when passed from the command line.

$ docker run --rm myimg:latest bash -c 'pyats run job $JOBFILE'

Bash does not interpolate anything within single quotes, so the entire command is preserved as a string in this way when passed to Docker. Then the command bash -c will interpolate and execute the string inside the Docker container, which will resolve variables correctly.

API

pyATS Image Builder can also be used directly from another Python script using the build() function and Image class.

build()

build(config = {}, path = None, tag = None, keep_context = False,
      verbose = False, stream = None, no_cache = False, dry_run = False)
Argument Description
config The mapping typically loaded from the yaml file. Defines Python packages to install, files to copy, etc. Refer to the YAML file section for the schema.
path An alternative path to use as the image context. If this location already exists, it will not be cleaned upon finishing the image build. If this location does not exist, pyATS Image Builder will attempt to create it. If not given, a temporary location will be used.
tag A Docker tag for the completed image that takes precedence over any tag defined inside the config.
keep_context When True prevents the context directory from being cleaned after the build is finished.
verbose When True logs the entire Docker build process to the console.
stream An IO Stream to write the logging output to. This will always include the Docker build logs.
no_cache When True prevents Docker from using cached images when building, forcing intermediate images to be rebuilt.
dry_run When True prevents the Docker build from happening after assembling the context.

Returns

On success, build() will return an Image object representing the Docker Image just built. This can be queried for information or used to push the image to a registry.

from pyatsimagebuilder import build
config = {'tag':'mypyatsimage:latest', 'packages': ['pyats[full]']}
image = build(config=config)

Image Class

The Image class can retrieve information about the newly created image, as well as the push the image to a registry.

Image(image_id, tag = None)
Argument Description
image_id The hash identifier for this image.
tag The main tag to use for this image, since images can have multiple tags. If not provided here, must be given when pushing the image instead.

inspect()

An Image object can be queried with inspect() for a dict with detailed information about the associated Docker image.

image = build(config)
image.inspect()
# {
#   "Id": "sha256:...",
#   "RepoTags": ["myimg:latest"],
#   "Parent": "sha256:...",
#   "Created": "2020-01-01T20:00:00.0000000Z",
#   ...
# }

push()

An Image object has a method for pushing the associate Docker image to a registry. Can add a new tag to the image with the registry address for pushing to a private registry.

push(remote_tag = None, credentials = None)
Argument Description
remote_tag A tag to apply to the image before pushing in order to add the registry.
credentials A dict of username and password to authenticate with instead of the credentials configured in Docker.
image = build(config)
image.push(remote_tag='myregistry.domain.com:5000/myrepo/custom:latest',
           credentials={'username':username, 'password':password})