This is a Vagrant 1.6+ plugin that allows orchestrated deployments
to already provisioned (non-elastic) servers on top of the excellent Vagrant Managed Servers plugin.
It features a powerful templating init
command, support for multiple environments, several deployment strategies
and is designed from the ground up to be cross-platform, with first class support for Windows,
Linux, and Mac.
$ vagrant orchestrate init --servers myserver1.mydomain.com,myserver2.mydomain.com \
--ssh-username USERNAME --ssh-private-key-path PATH \
--shell --shell-inline "echo Hello"
$ ls
Vagrantfile dummy.box
$ vagrant orchestrate push
==> managed-myserver1.mydomain.com: Linking vagrant with managed server myserver1.mydomain.com
==> managed-myserver1.mydomain.com: -- Server: myserver1.mydomain.com
==> managed-myserver1.mydomain.com: Rsyncing folder: ~/dev/demo => /vagrant
==> managed-myserver1.mydomain.com: Running provisioner: shell...
==> managed-myserver1.mydomain.com: Running: inline script
==> managed-myserver1.mydomain.com: Hello
==> managed-myserver1.mydomain.com: Unlinking vagrant from managed server myserver1.mydomain.com
==> managed-myserver1.mydomain.com: -- Server: myserver1.mydomain.com
==> managed-myserver2.mydomain.com: Linking vagrant with managed server myserver2.mydomain.com
==> managed-myserver2.mydomain.com: -- Server: myserver2.mydomain.com
==> managed-myserver2.mydomain.com: Rsyncing folder: ~/dev/demo => /vagrant
==> managed-myserver2.mydomain.com: Running provisioner: shell...
==> managed-myserver2.mydomain.com: Running: inline script
==> managed-myserver2.mydomain.com: Hello
==> managed-myserver2.mydomain.com: Unlinking vagrant from managed server myserver2.mydomain.com
==> managed-myserver2.mydomain.com: -- Server: myserver2.mydomain.com
This also works for Windows with the --winrm --winrm-username USERNAME --wirnm-password PASSWORD
parameters, but must be initiated from a Windows host.
Install using the standard Vagrant plugin installation method:
$ vagrant plugin install vagrant-orchestrate
Initialize a Vagrantfile to orchestrate running a script on multiple managed servers
$ vagrant orchestrate init --shell
Which produces a simple default Vagrantfile that can push to managed servers:
managed_servers = %w( )
Vagrant.configure("2") do |config|
# This disables up, provision, reload, and destroy for managed servers. Use
# `vagrant orchestrate push` to communicate with managed servers.
config.orchestrate.filter_managed_commands = true
config.vm.provision "shell", path: "{{YOUR_SCRIPT_PATH}}"
config.ssh.username = "{{YOUR_SSH_USERNAME}}"
config.ssh.private_key_path = "{{YOUR_SSH_PRIVATE_KEY_PATH}}"
config.vm.define "local", primary: true do |local|
local.vm.box = "ubuntu/trusty64"
end
managed_servers.each do |instance|
config.vm.define instance, autostart: false do |box|
box.vm.box = "managed-server-dummy"
box.vm.box_url = "./dummy.box"
box.vm.provider :managed do |provider|
provider.server = instance
end
end
end
end
You'll need to edit your Vagrantfile and replace some variables, such as ssh username and
private key, and the path to the script to run. Alternatively, you can pass them on the command
line with --ssh-username
and --ssh-private-key-path
. The first line of the file defines an whitespace delimeted
array of managed servers that the push
command will operate on.
managed_servers = %w( myserver1.mydomain.com myserver2.mydomain.com )
This works for Windows managed servers using WinRM as well
$ vagrant orchestrate init --winrm --winrm-username USERNAME --winrm-password PASSWORD
required_plugins = %w( vagrant-managed-servers vagrant-winrm-s )
...
config.vm.communicator = "winrm"
config.winrm.username = "USERNAME"
config.winrm.password = "PASSWORD"
This also supports a portable and repeatable way to install plugins, just list them in the .vagrantplugins file.
version
and source
are the supported values in the options hash, but neither is required.
required_plugins = {}
required_plugins["vagrant-orchestrate"] = {}
required_plugins["vagrant-managed-servers"] = { version: "0.7.0" }
If you are executing in a shared environment, like a build slave, you can create your own
own plugin install directory by setting the VAGRANT_HOME
variable to something relative
to the current directory.
$ VAGRANT_HOME=./.vagrant.d vagrant orchestrate push
or
> SET VAGRANT_HOME=./.vagrant.d
> vagrant orchestrate push
Vagrant Orchestrate offers a way to manage multiple environments using a combination of a single servers.json file and the name of the current git branch as an indicator of the current environment.
To initialize an environment aware Vagrantfile, use
$ vagrant orchestrate init --environments dev,test,prod
You'll need to create git branches with matching names and enter data into the the servers.json file in order for the Vagrantfile to be git branch aware.
Learn more about environments.
Vagrant orchestrate offers the capability to prompt for credentials from the command
line at the time of a push. You can initialize your Vagrantfile to declare this
by passing the --credentials-prompt
flag to the vagrant orchestrate init
command,
or add the following to your Vagrantfile.
config.orchestrate.credentials.prompt = true
The credentials config object can accept one additional parameter, file_path
. Setting
creds.file_path = path/to/username_password.yaml
tells vagrant-orchestrate to
look for a file at the given path, and read from its :username and :password fields
('username' and 'password' are also accepted). Additionally, you can pass the username
and password in using the VAGRANT_ORCHESTRATE_USERNAME
and VAGRANT_ORCHESTRATE_PASSWORD
environment variables. Environment variables take precedence over the file, and the file
takes precedence over the prompting. It is possible to set prompt
to false
, or leave
it unset, in which case only environment variables and the credentials file (if provided)
will be checked.
Experimental puppet templating support is available as well with the --puppet
flag and associated options
See details on the various configuration options here.
Go ahead and push changes to your managed servers, in serial by default.
$ vagrant orchestrate push
The push command is currently limited to vagrant machines that use the :managed
provider. So if you have other, local machines defined in the Vagrantfile, vagrant orchestrate push
will not operate on those.
Similar to the up
and provision
commands in vagrant, the push
command can take in a regular expression for matching only certain machines. This is useful if your Vagrantfile contains all the managed machines, but you only want to push changes to a subset of those.
It can be easy to make mistakes such as rebooting a production server if you have managed long-lived servers as well as local VMs defined in your Vagrantfile. We add some protection with the orchestrate.filter_managed_commands
configuration setting, which will cause up, provision, reload, and destroy commands to be ignored for servers with the managed provider. This can be disabled by setting the variable to false in the Vagrantfile.
config.orchestrate.filter_managed_commands = true
Vagrant Orchestrate supports several deployment strategies including parallel, canary, and half and half.
You can push changes to all of your servers in parallel with
$ vagrant orchestrate push --strategy parallel
The vagrant orchestrate status
command will reach out to each of the defined
managed servers and print information about the last successful push from this
repo, including date, ref, and user that performed the push.
$ vagrant orchestrate status
Current managed server states:
managed-1 2015-04-19 00:46:22 UTC e983dddd8041c5db77494266328f1d266430f57d cbaldauf
managed-2 2015-04-19 00:46:22 UTC e983dddd8041c5db77494266328f1d266430f57d cbaldauf
managed-3 Status unavailable.
managed-4 2015-04-19 00:43:07 UTC e983dddd8041c5db77494266328f1d266430f57d cbaldauf
Vagrant Orchestrate has support for integrating with deployment-tracker, which can collect and record metadata about deployments, including summary records, metrics, and logs. This is great if you're in a distributed deployment environment, but need a central store for operational or compliance reasons.
Add the following configuration option to your Vagrantfile
config.orchestrate.tracker_host = "deploymenttracker.mydomain.com"
See more info on deployment tracker integration.
- When deploying you must have no pasword prompt setup. You can do this by adding the following to your sudoers file. (edit via
sudo visudo
)
UserName ALL=(ALL:ALL) NOPASSWD:ALL
or %GroupName ALL=(ALL:ALL) NOPASSWD:ALL
- Need rsync? Install OpenSSH and then run this script to install rsync. Vagrant managed servers currently only works with cygwin based rsync implementations.
- You MUST have at least PowerShell 3 installed in order to use the SMB folder synch. If you have PowerShell 2 installed and try to execute a folder sync, it will hang with no good error message.
- Libcurl.dll must be installed and on your path. You can find it here
Alternative to the above instructions there is a puppet module that will setup a windows machine for you.
You'll need to bootstrap the target machine. The following script should get you there.
winrm quickconfig -q
winrm set winrm/config/service/auth @{Negotiate="true"}
winrm set winrm/config/service @{AllowUnencrypted="false"}
winrm set winrm/config/winrs @{MaxShellsPerUser="25"}
winrm set winrm/config/winrs @{MaxConcurrentUsers="25"}
sc config winrm start= auto
sc config winrm type= own
- Check out the winrm-s readme for more information
See the Synced Folders section of the Vagrant Managed Servers readme for more detail.
- Fork it ( https://github.com/Cimpress-MCP/vagrant-orchestrate/fork )
- Create your feature branch (
git checkout -b my-new-feature
) - Commit your changes (
git commit -am 'Add some feature'
) - Run locally with
bundle exec vagrant orchestrate [init|push|status]
bundle exec rake build
bundle exec rake acceptance
, which will take a few minutes- Push to the branch (
git push origin my-new-feature
) - Create a new Pull Request
Prerequisites:
- Ruby 2.0 or greater