Feature request: Vagrant Up then Remote Connection

It would be useful!

In the interim I use a strategy to automatically maintain the remote hosts list, which is actually all you need. It solves the problem of the machines coming up on different ports each time.

I thought it worth describing here in case people find it interesting or useful.

Visual Studio Code can parse include directives in .ssh/config, but not wildcards (at the moment).

So I use triggers to regenerate a set of SSH config include files from vagrant ssh-config and store them in a known directory. I then have a method that generates a single index file including each of them, and in turn I include that from .ssh/config.

Basically:

when bringing up/provisioning/reloading/resuming a machine, I dump the output from vagrant ssh-config into a file in a known directory (inside $HOME/.vagrant.d)
when destroying/halting/suspending machines, I remove the file
in each case I rebuild the master include file

Then each trigger action will maintain your remote SSH hosts list for you, without needing to restart vscode.

The triggers in Vagrantfile:

 config.trigger.after [:up, :provision, :reload, :resume] do |trigger|
    trigger.ruby do |env,machine|
      MyClass.up_trigger(env, machine)
    end
  end

  config.trigger.after [:destroy, :halt, :suspend] do |trigger|
    trigger.ruby do |env,machine|
       MyClass.down_trigger(env, machine)
    end
  end

Then in MyClass is the following code.

This function defines an SSH configuration directory within .vagrant, where I will write individual files containing the output of vagrant ssh-config:

    def ssh_config_dir
      cfg_dir = Pathname.new(Vagrant.user_data_path).join('mydir', 'ssh')
      cfg_dir.mkpath unless cfg_dir.exist?
      cfg_dir
    end

This function builds a master include file of all of those files, which gets stored in the same directory (so, err, don't call a vagrant box "include" ;-):

    def build_ssh_include_file(ssh_dir)
      return if !File.directory? ssh_dir

      include_file = [ ]
      Dir.entries(ssh_dir).each do |item| 
        next if File.directory? item  
        next if item == '.' or item == '..'
        next if item !~ /\.config$/i
        next if item == 'include.config'
        path = ssh_dir.join(item)
        include_file << "Include #{path}"
      end

      ssh_dir.join('include.config').write("# all vagrant hosts\n" + include_file.join("\n"))
    end

The up_trigger function uses bash to invoke vagrant ssh-config to write the SSH config to a file. So this is sort of Unix-only for now. I guess I could directly invoke the underlying Vagrant code that produces that output for a more portable approach.

It then rebuilds the index file.

    def up_trigger(env, machine)

      # make our SSH config directory if we need it
      cfg_dir = ssh_config_dir
      return if !File.directory? cfg_dir

      # now dump the vagrant SSH config into a named file
      ssh_config = `bash -c 'vagrant ssh-config'`
      cfg_dir.join(machine.name.to_s + '.config').write("\n" + ssh_config)

      build_ssh_include_file(cfg_dir)
    end

The down trigger removes the appropriate SSH config file and rebuilds the index. So the host will disappear from your vscode remotes.

    def down_trigger(env, machine)
      cfg_dir = ssh_config_dir
      return if !File.directory? cfg_dir

      cfg_file = cfg_dir.join(machine.name.to_s + '.config')
      cfg_file.unlink if File.file? cfg_file
      build_ssh_include_file(cfg_dir)
    end

So you then include the master list in .ssh/config:

Include ~/.vagrant.d/mydir/ssh/include.config

bbenoist / vscode-vagrant

Feature request: Vagrant Up then Remote Connection #18