Skip to content

All of your packages will fit into this van with this one simple trick.

License

Notifications You must be signed in to change notification settings

skyamgarp/vanagon

 
 

Repository files navigation

Build Status

The Vanagon Project

What is Vanagon?

Vanagon is a tool to build a single package out of a project, which can itself contain one or more components. This tooling is being used to develop the puppet-agent package, which contains components such as openssl, ruby, and augeas among others. For a simple example, please see the project in the examples/ directory.

Vanagon builds up a Makefile and packaging files (specfile for RPM, control/rules/etc for DEB) and copies them to a remote host, where make can be invoked to build all of the components and make a package of the contents.

How is it pronounced?

Vanagon (/ˈvænəgɪn/) sounds like "van again." It does not sound like "van wagon" or "van and gone."

Runtime Requirements

Vanagon carries two sets of requirements: requirements for the local host where Vanagon is run, and for the remote target where compilation and packaging happens.

Also, Vanagon ships with a number of engines which may include additional optional dependencies if you wish to use them. These engines are currently considered experimental, and receive less attention than the Hardware or vmpooler engines do. If you find a bug in these engines, please open a ticket and let us know.

Local Host:

  • Ruby (Ruby 2.3.x is the miniumum supported version)
  • fustigit
  • ruby-git
  • docopt
  • The command line tool ssh (homepage) available on the local ${PATH} (any modern version should suffice)
  • The command line tool rsync (homepage) available on the local ${PATH} (At least rsync 2.6.x)
  • The command line tool git (homepage) available on the local ${PATH} (Vanagon is tested against Git version 1.8.x but should work with any newer version)

Optional requirements

Remote Build Target:

Note: package installation & builder configuration for the remote target can be customized in the Platform configuration that defines target provisioning instructions.

  • GNU Make (homepage) (Vanagon specifically targets the feature set provided by GNU Make 3.81 but newer versions are known to work -- older versions are specifically known to not work!)
  • Bash (homepage) is required by the Makefiles that Vanagon generates
  • An ssh server (homepage) is required by most engines
  • The command line tool rsync (homepage) (At least rsync 2.6.x)

Configuration and Usage

Vanagon won't be much use without a project to build. Beyond that, you must define any platforms you want to build for. Vanagon ships with some simple binaries to use, but the one you probably care about is named 'build'.

Overview

Vanagon is broken down into three core ideas: the project, the component and the platform. The project contains one or more components and is built for a platform. As a quick example, if I had a ruby app and wanted to package it, the project would probably contain a component for ruby and a component for my app. If I wanted to build it for debian wheezy, I would define a platform called wheezy and build my project against it.

For more detailed examples of the DSLs available, please see the examples directory and the YARD documentation for Vanagon.

CLI commands

The vanagon command line tool contains these commands:

  • build build a package given a project and platform
  • build_host_info print information about build hosts
  • build_requirements print external packages required to build project
  • completion outputs path to tab completion script
  • inspect a build dry-run, printing lots of information about the build
  • list shows a list of available projects and platforms
  • render create local versions of packaging artifacts for project
  • sign sign a package
  • ship upload a package to a distribution server
  • help print this help

Commands are called with a git-like pattern of vanagon <subcommand>. For example: vanagon list

CLI changes and deprecations (from version 0.16.0)

Prior to 0.16.0, the vanagon command line contained these commands

  • build
  • build_host_info
  • build_requirements
  • inspect
  • render
  • repo
  • ship
  • sign

With the exception of repo, which calls packaging methods, the remaining commands have been moved to a git-like pattern of vanagon <subcommand>. vangon build replaced build, vanagon ship replaced ship and so forth.

The older calling usage is deprecated.

vanagon build usage

The build command has positional arguments and position independent flags.

Positional arguments

project name

The name of the project to build; a file named <project_name>.rb must be present under configs/projects in the working directory.

platform name

The name of the target platform to build <project_name> against; a file named <platform_name>.rb must be present under configs/platforms in the working directory. This can also be a comma separated list of platforms such as platform1,platform2; note that there are no spaces after the comma.

target host

Target host is an optional argument to override the host selection. Instead of using a random VM collected from the pooler (Vanagon's default build engine), the build will attempt connect to the target host over SSH as the root user.

If building on multiple platforms, multiple targets can also be specified using a comma separated list such as host1,host2 (note that there are no spaces after the comma). If less targets are specified than platforms, the default engine (pooler) will be used for platforms without a target. If more targets are specified than platforms, the extra platforms will be ignored.

Build machines should be cleaned between builds.

Flagged arguments

Note: command flags currently can be used anywhere in the command. Recommended usages is to use flagged arguments before the positional arguments.

-w DIR, --workdir DIR

Specifies a directory on the local host where the sources should be placed and builds performed. Defaults to a temporary directory created with Ruby's Dir.mktmpdir method.

-r DIR, --remote-workdir DIR

Explicitly specify a directory on the remote target to place sources and perform builds. Components can then be rebuilt manually on the build host for faster iteration. Sources may not be correctly updated if this directory already exists. Defaults to a temporary directory created by running mktemp -d on the remote target.

-c DIR, --configdir DIR

Specifies where project configuration is found. Defaults to $pwd/configs.

-e ENGINE, --engine ENGINE

Choose a different virtualization engine to use to select the build target. Currently supported engines are:

  • base - Pure ssh backend; no teardown currently defined
  • local - Build on the local machine; platform name must match the local machine
  • docker - Builds in a docker container
  • pooler - Selects a vm from Puppet's internal vmpooler to build on
  • hardware - Build on a specific taget and lock it in redis
  • ec2 - Build on a specific AWS instance.
  • ABS - Selects a vm from Puppet's internal Always-be-scheduling service to build on

Flags

Note: command flags can be used anywhere in the command.

-p, --preserve

Indicates that the host used for building the project should be left intact after the build instead of destroyed. The host is usually destroyed after a successful build, or left after a failed build.

-h, --help

Display command-line help.

Environment variables

VANAGON_SSH_KEY

A full path on disk for a private ssh key to be used in ssh and rsync communications. This will be used instead of whatever defaults are configured in .ssh/config.

VANAGON_SSH_AGENT

When set, Vanagon will forward the ssh authentication agent connection.

VMPOOLER_TOKEN

Used in conjunction with the pooler engine, this is a token to pass to the vmpooler to access the API. Without this token, the default lifetime of vms will be much shorter.

LOCK_MANAGER_HOST

The name of the host where redis is running. Redis is used to handle a lock when using the hardware engine. It defaults to redis, with no domain.

LOCK_MANAGER_PORT

Port of the system where redis is running. Defaults to 6379.

VANAGON_USE_MIRRORS

Controls whether component sources are downloaded directly from upstream URLs or from configured mirrors. Most Puppet projects using Vanagon default to fetching components from internal mirrors. Set this variable to n or false when building outside of the Puppet private network to download directly from upstream sources.

VANAGON_RETRY_COUNT

Some phases of compilation support retries. The default value is 1 but setting to any integer value greater than 1 will causes these components to retry operations on failure until the VANAGON_RETRY_COUNT limit is reached.

VANAGON_TIMEOUT

Some phases of compilation can take an indeterminate (but substantial) amount of time. The default value is 7200 seconds(120 minutes) but setting to any integer value these components to fail after the VANAGON_TIMEOUT count is reached. Note that this value is expected to be in seconds.

VANAGON_FORCE_SIGNING

By default, Vanagon does not fail if extra files signing fails, it just logs an error and continues building the package. This is unwanted behavior in environments where we expect a hard failure when signing cannot proceed. To force Vanagon to fail if extra files signing fails, ensure this variable is set before starting a build.

Example usage

vanagon build --preserve puppet-agent el-6-i386 will build the puppet-agent project on the el-6-i386 platform and leave the host intact afterward.

vanagon build --engine=docker puppet-agent el-6-i386 will build the puppet-agent project on the el-6-i386 platform using the docker engine (the platform must have a docker_image defined in its config).


vanagon inspect usage

The inspect command has positional arguments and position independent flags. It mirrors the build command, but exits with success after loading and interpolating all of the components in the given project. No attempt is made to actually build the given project; instead, a JSON formatted array of hashes is returned and printed to stdout. This JSON array can be further processed by external tooling, such as jq.

Positional arguments

project name

The name of the project to build, and a file named <project_name>.rb must be present in configs/projects in the working directory.

platform name

The name of the platform to build against, and a file named <platform_name>.rb must be present in configs/platforms in the working directory.

Platform can also be a comma separated list of platforms such as platform1,platform2.

Flagged arguments

Note: command flags currently can be used anywhere in the command. Recommended usages is to use flagged arguments before the positional arguments.

-w DIR, --workdir DIR

Specifies a directory where the sources should be placed and builds performed. Defaults to a temporary directory created with Ruby's Dir.mktmpdir.

-c DIR, --configdir DIR

Specifies where project configuration is found. Defaults to $pwd/configs.

-e ENGINE, --engine ENGINE

Choose a different virtualization engine to use to select the build target. Engines are respected, but only insofar as components and projects are rendered -- the vanagon inspect command performs no compilation.

Supported engines are the same as the vanagon build command.

Flags

-h, --help

Display command-line help.

Environment variables

Environment variables are respected, but only insofar as components and projects are rendered -- the vanagon inspect command has no behavior to alter.

Supported environment variables are the same as the vanagon build command.

Example usage

vanagon inspect puppet-agent el-6-i386 will load the puppet-agent project on the el-6-i386 platform and print the resulting list of dependencies, build-time configuration, environment variables, and expected artifacts.


Engines

Amazon Ec2

Note: If you have the aws_ami setup Vanagon will default to the ec2 engine.

To use the ec2 engine you should have your credentials set either via your ~/.aws/credentials or environment variables. After this you can setup your configs/platforms/<platform>.rb to use your ami, instance type, and key_name to setup the instance.

A simple one looks like this

# configs/platforms/el-7-x86_64.rb
platform "el-7-x86_64" do |plat|
    plat.aws_ami "your-ami-id-here" # You must set this
    plat.aws_instance_type "t2.small" # Defaults to t1.micro
    plat.aws_key_name "vanagon" # this is the default but you can use whichever
    plat.aws_user_data <<-eos
#cloud-config
    runcmds:
        - echo #{my_ssh_key} > /root/.ssh/authorized_keys # Most amis block you from logging in as root.
    eos

### Rest of your code here

end

ABS (internal)

When using the ABS engine, there is a variety of ways you can specify your token:

  • the environment variable ABS_TOKEN
  • or vanagon token file ~/.vanagon-token (note this is the same file read by the pooler engine)
  • or vmfloaty's config file ~/.vmfloaty.yml

In order to modify or track the VM via floaty or bit-bar you can optionally add the vmpooler token (different from the ABS token) via

  • VMPOOLER_TOKEN
  • or as a second line to the file ~/.vanagon-token
  • or by using the existing mechanism in floaty using a vmpooler_fallback

Contributing

We'd love to get contributions from you! Once you are up and running, take a look at the Contribution Documents to see how to get your changes merged in.

License

See LICENSE file.

Maintainers

See MAINTAINERS file.

About

All of your packages will fit into this van with this one simple trick.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Ruby 93.4%
  • HTML 4.2%
  • Shell 1.9%
  • Other 0.5%