Batch Shipyard Installation

There are multiple available options for installing Batch Shipyard. Please pick an option that is most suitable for your work environment.

If you wish to install Batch Shipyard into your Azure App Service (e.g., Azure Function App) environment, please see this guide.

Pre-built Binary

Download an appropriate Release binary for your operating system. Pre-built binaries are not available for all platforms and architectures at this time.

Note that for the Linux pre-built binary, it may not work on all distributions. If this is the case, please pick an alternate installation method. After downloading the binary, make sure that the executable bit is set via chmod +x prior to attempting to execute the file.

Installation via Script

Installation is an easy two-step process if using the installers: fetch the code and run the install script to download and setup dependencies. This is typically the most flexible and compatible installation outside of the Docker image for the CLI.

Azure Cloud Shell

Batch Shipyard is now integrated into Azure Cloud Shell with no installation required. Simply request a Cloud Shell session and type shipyard to invoke the CLI. Data stored in your home directory or clouddrive will persist between Cloud Shell sessions.

Note that Azure Cloud Shell may not have the most recent release of Batch Shipyard. You can see the version of Batch Shipyard installed with the command shipyard --version.

If you wish to install Batch Shipyard on your machine, please proceed to the Installation section.

Step 1: Acquire Batch Shipyard

Clone the repository:

git clone https://github.com/Azure/batch-shipyard.git

or download the latest release and unpack the archive.

For the next step (if not using the Batch Shipyard CLI Docker image), refer to your operating system specific installation instructions:

Alternatively, you can install the Batch Shipyard CLI on your machine via Docker or Singularity. If using the Docker image, this is the only step needed and does not require any futher installation steps.

Step 2 [Linux]: Run the install.sh Script

Batch Shipyard includes an installation script to simplify installation on a variety of recent Linux distributions. This installation script can be used regardless of if you obtained Batch Shipyard through git clone or downloading a release package.

Please ensure that your target Python is 3.5+.

The install.sh script supports isolated installation through a virtual environment so that other system-wide or user python dependencies are left unmodified. To perform this style of installation, which is recommended and the default, simply invoke the install.sh script without parameters. If you would like to specify the virtual environment to use, use the -e parameter. If you don't want to use a virtual environment and instead would like to install into your user environment, specify the -u option. Using this option will require modifying your shell rc file for advanced data movement capability provided by Batch Shipyard.

The recommended installation method with a virtual environment:

# Ensure you are NOT root
# Obtain Batch Shipyard through git clone or downloading the archive and unpacking
# Change directory to where Batch Shipyard was cloned or unpacked to
cd batch-shipyard
# Install for Python 3.5+ in the virtual environment ".shipyard"
./install.sh

A helper script named shipyard will be generated with a successful installation. This helper script can be invoked in lieu of shipyard.py which will invoke shipyard with the appropriate interpreter and virtual environment, if created during installation.

Do not delete the virtual environment directory (in the above example, a directory named .shipyard would be created), as this contains the virtual environment required for execution.

Please note that although Anaconda environment installations are supported, there is a larger startup delay for invoking shipyard with Anaconda environments due to the delay in activating a conda environment. Python from python.org (CPython) is recommended as the execution environment.

Alternatively, install directly into your "user" environment:

# Ensure you are not root
# Obtain Batch Shipyard through git clone or downloading the archive and unpacking
# Change directory to where Batch Shipyard was cloned or unpacked to
cd batch-shipyard
# Install in user environment
./install.sh -u
# Add $HOME/.local/bin to your PATH in your shell rc file if it is not present.
# For example, the following line can be added to ~/.bashrc for bash shells:
export PATH=$PATH:$HOME/.local/bin
# Reload .bashrc for bash shells
. ~/.bashrc

Please ensure that you are not invoking the install script as root. sudo will be invoked wherever root access is required for installing system-wide packages in the install.sh script. Python packages required by Batch Shipyard will be installed either in the virtual environment or user context.

Please see the Upgrading section below for information on upgrading to a new release of Batch Shipyard.

Installation on CentOS 6.x / RHEL 6.x / Fedora 13 to 18

The default python interpreter distributed with 6.x series releases is incompatible with Batch Shipyard. To install on these distributions, you must install epel-release package first then the python36 epel package. Once these packages are installed, then invoke the installer in the following manner:

DISTRIB_ID=centos DISTRIB_RELEASE=6.x ./install.sh

Unsupported Linux Distributions

The following distributions will NOT work with the install.sh script:

  • CentOS < 6.0
  • Debian < 8
  • Fedora < 13
  • OpenSUSE < 13.1
  • RHEL < 6.0
  • SLES < 12
  • Ubuntu < 14.04

Please follow the manual installation instructions found later in this document for these distributions.

Step 2 [Mac]: Run the install.sh Script

You will need to follow the steps outlined on this guide to install Batch Shipyard on a Python3 installation rather than the default Python2 that is shipped with Mac OS X.

The install.sh script supports isolated installation through a virtual environment so that other system-wide or user python dependencies are left unmodified. To perform this style of installation, which is recommended, specify the virtual environment to create with the -e parameter. This option also does not require modifying your shell rc file for advanced data movement capability provided by Batch Shipyard.

The recommended installation method with a virtual environment:

# Ensure you are not root
# Obtain Batch Shipyard through git clone or downloading the archive and unpacking
# Change directory to where Batch Shipyard was cloned or unpacked to
cd batch-shipyard
# Install for Python 3.5+ in the virtual environment ".shipyard"
./install.sh

A helper script named shipyard will be generated with a successful installation. This helper script can be invoked in lieu of shipyard.py which will invoke shipyard with the appropriate interpreter and virtual environment, if created during installation.

Do not delete the virtual environment directory (in the above example, a directory named .shipyard would be created), as this contains the virtual environment required for execution.

Please note that although Anaconda environment installations are supported, there is a larger startup delay for invoking shipyard with Anaconda environments due to the delay in activating a conda environment. Python from python.org (CPython) is recommended as the execution environment which can be installed easily with the Homebrew as noted in the aforementioned guide.

Please see the Upgrading section below for information on upgrading to a new release of Batch Shipyard.

Step 2 [Windows]: Run the install.cmd Script

Batch Shipyard includes a installation command file that simplifies installing Batch Shipyard on python.org (CPython) and Anaconda. It is highly recommended to use Python 3.5 or later (or an Anaconda equivalent). The use of the install.cmd script installs Batch Shipyard into a virtual environment. Please ensure that python.exe can be found in your %PATH%. For example:

# Change directory to where Batch Shipyard was cloned or unpacked to
cd batch-shipyard
# Install Batch Shipyard into the virtual environment named "shipyard.venv"
install.cmd shipyard.venv

A helper command file named shipyard.cmd will be generated with a successful installation. This helper script can be invoked in lieu of shipyard.py which will invoke shipyard with the appropriate interpreter and virtual environment, if created during installation.

Do not delete the virtual environment directory (in the above example, a directory named shipyard.venv would be created), as this contains the virtual environment required for execution.

Please note that although Anaconda environment installations are supported, there is a larger startup delay for invoking shipyard.cmd with Anaconda environments due to the delay in activating a conda environment. Python from python.org (CPython) is recommended as the execution environment.

Alternatively you can install Batch Shipyard using the requirements.txt file:

# Install for Windows on Python 3.5+
pip3.exe install --upgrade -r requirements.txt
# Or invoke directly from wherever you installed python
C:\Python36\Scripts\pip3.exe install --upgrade -r requirements.txt

Note that installing directly using the requirements.txt file does not create a shipyard.cmd file that wraps shipyard.py and points it to the proper interpreter.

Please see the Upgrading section below for information on upgrading to a new release of Batch Shipyard.

Step 2 [Windows Subsytem for Linux]: Run the install.sh Script

Windows Subsystem for Linux (using Ubuntu) now comes with Python3 installed out of the box. This is the recommended Python interpreter to use with Batch Shipyard. First, you must enable and install Windows Subsystem for Linux. Please follow the installation guide from Microsoft, if you have not enabled and installed it.

The install.sh script supports isolated installation through a virtual environment so that other system-wide or user python dependencies are left unmodified. To perform this style of installation, which is recommended, specify the virtual environment to create with the -e parameter. This option also does not require modifying your shell rc file for advanced data movement capability provided by Batch Shipyard.

Once you have installed and activated the Windows Subsystem for Linux and Bash, the recommended installation method with a virtual environment is as follows:

# Obtain Batch Shipyard through git clone or downloading the archive and unpacking
# Change directory to where Batch Shipyard was cloned or unpacked to
cd batch-shipyard
# Install for Python3 in the virtual environment ".shipyard"
./install.sh

A helper script named shipyard will be generated with a successful installation. This helper script can be invoked in lieu of shipyard.py which will invoke shipyard with the appropriate interpreter and virtual environment, if created during installation. For the purposes of usage documentation as it refers to either Linux or Windows, Batch Shipyard installations on Windows Subsystem for Linux should be considered as Linux.

Do not delete the virtual environment directory (in the above example, a directory named .shipyard would be created), as this contains the virtual environment required for execution.

Please see the Upgrading section below for information on upgrading to a new release of Batch Shipyard.

Batch Shipyard CLI Installation via Docker

If using the mcr.microsoft.com/azure-batch/shipyard:latest-cli Docker image, then all of the required software is bundled in the image itself, however, you will need an installation of the Docker engine on your machine. More information on installing Docker to your local machine can be found here.

To install:

docker pull mcr.microsoft.com/azure-batch/shipyard:latest-cli

This will pull the CLI Docker image of Batch Shipyard to your local machine. You are now ready to execute it with docker run. Please see the Batch Shipyard Usage guide for more information on how to execute the Batch Shipyard CLI Docker image.

Batch Shipyard CLI Installation via Singularity

If using the library://alfpark/batch/shipyard:latest-cli, Singularity image, then all of the required software is bundled in the image itself, however, you will need an installation of Singularity on your machine. More information on installing Singularity to your local machine can be found here.

To install:

singularity pull library://alfpark/batch/shipyard:cli

This will pull the CLI Singularity image of Batch Shipyard to your local machine (to the current working directory). You are now ready to execute it with singularity run or simply just executing the image. Please see the Batch Shipyard Usage guide for more information on how to execute the Batch Shipyard CLI Singularity image.

Jupyter Notebooks

There are community contributed Jupyter notebooks to help you quickly get started if you prefer that environment instead of a commandline.

Upgrading to New Releases

To upgrade to a new release, simply execute git pull or download a new release archive and unpack. Next, upgrade the dependencies for your respective platform below.

Linux, Mac, and Windows Subsystem for Linux

Rerun the install.sh script with the appropriate parameters for all upgrades. Please ensure that if you specified -u or -e <env name> parameter that these parameters are issued again for upgrades.

Windows

Rerun the install.cmd script with the same virtual environment parameter.

CLI Docker

If using the CLI Docker image, simply re-issue the docker pull command above.

Pre-built Binary

Download a new version of the binary.

Windows Support

Please note that while Batch Shipyard can run on Windows, some functionality may not be supported in Windows out of the box such as SSH, scp, rsync, and credential encryption functionality. To ensure support for these features on Windows, please read the Data Movement and Optional Credential Encryption Support sections below.

Alternatively, you can install Batch Shipyard on the Windows Subsystem for Linux with full support for all functionality as if running on Linux natively.

Note that if you are cloning the repository on Windows, please ensure that git or any text editor does not modify the Unix line-endings (LF) for any file in the scripts or resources directory. The repository's .gitattributes designates line endings for all text files, but can be overridden by your local git configuration. If these files are modified with Windows line-endings (CRLF) then compute nodes will fail to start properly.

Manual Installation

Requirements

The Batch Shipyard tool is written in Python. The client script is compatible with Python 3.5+. You will also need to install dependent Python packages that Batch Shipyard requires. Installation can be performed using the requirements.txt file via the command pip3 install --upgrade --user -r requirements.txt. Note that this pip3 command should be run for every Batch Shipyard upgrade if not using install.sh. The use of install.sh is highly recommended instead of these manual steps below on Linux platforms.

Batch Shipyard has some Python dependencies which require a valid compiler, ssl, ffi, and Python development libraries to be installed due to the cryptography dependency on Linux. For Windows, binary wheels will be installed for most dependencies. If installing on Python 3.5 or later for Windows, no development environment is needed. The following are example commands to execute (as root or with sudo) to install the required dependencies on Linux:

Ubuntu/Debian

apt-get update
apt-get install -y build-essential libssl-dev libffi-dev python3-dev python3-pip
pip3 install --upgrade pip

CentOS/RHEL/Fedora

yum install -y epel-release
yum install -y python36-devel gcc openssl-devel libffi-devel
curl -fSsL https://bootstrap.pypa.io/get-pip.py | python

SLES/OpenSUSE

zypper ref
zypper -n in gcc libopenssl-devel libffi48-devel python3-devel
curl -fSsL https://bootstrap.pypa.io/get-pip.py | python

Data Movement Support

Batch Shipyard contains native support for moving files locally accessible at the point of script execution. The install.sh script ensures that the following programs are installed. The Docker CLI image contains all of the necessary software as well. With a manual installation, the following programs must be installed to take advantage of data movement features of Batch Shipyard:

  1. An SSH client that provides ssh and scp (or ssh.exe and scp.exe on Windows). You can find OpenSSH binaries for Windows released by the PowerShell team. OpenSSH with HPN patches can be used on the client side to further accelerate scp to Azure Batch compute nodes where hpn_server_swap has been set to true in the pool_specification.
  2. rsync if rsync functionality is needed. This is not supported on Windows.
  3. blobxfer if transfering to Azure storage. This is automatically installed if pip install is used with requirements.txt as per above. If installed with --user flag, this is typically placed in ~/.local/bin. This path will need to be added to your PATH environment variable.

Optional Credential Encryption Support

Batch Shipyard supports optionally encrypting credentials that are used by backend components within your pool deployment (please refer to the referenced guide on if your scenario requires this functionality). In order to utilize this feature, you must have openssl (or openssl.exe on Windows) installed. The install.sh script ensures that OpenSSL is installed on Linux. The Docker CLI image also contains OpenSSL binaries. You might be able to find OpenSSL binaries for Windows on the OpenSSL wiki.

Note that all commandlines, environment variables and resource file URLs which are stored by the Azure Batch Service are encrypted by the service. This feature is to prevent credentials from being displayed in the clear when using the Azure Portal, Batch Labs, or other tools to inspect the status of pools, jobs and tasks. If this is not an issue for your scenario, then encrypting credentials is unnecessary. Please review the credential encryption guide for more information.

Next Steps

Either continue on to Batch Shipyard Configuration for a full explanation of all of the Batch Shipyard configuration options within the config files or continue to the Batch Shipyard Quickstart guide.