Need Decision¶

The ssh_remote_execute method is a thin wrapper to Net::SSH.start which is difficult to mock.

Reading online, I’m not sure if we should be mocking API’s we don’t own. The recommendation seems to be implement a thin wrapper and mock that wrapper, which we’ve already done.

I couldn’t quickly figure out how to add coverage for this code:

    # This is the single place to make SSH calls.  It will handle collecting STDOUT
    # in a line oriented manner, printing it to debug log destination and checking the
    # exit code of the remote call.  This should also make it much easier to do unit testing on
    # all of the other methods that need this functionality.  Finally, it should provide
    # one place to swap out the back end SSH implementation if need be.
    def ssh_remote_execute(server, login, command, keyfile = nil)
      Puppet.info "Executing remote command ..."
      Puppet.debug "Command: #{command}"
      buffer = String.new
      exit_code = nil
      Net::SSH.start(server, login, :keys => [ keyfile ]) do |session|
        session.open_channel do |channel|
          channel.on_data do |ch, data|
            buffer << data
            if buffer =~ /\n/
              lines = buffer.split("\n")
              buffer = lines.length > 1 ? lines.pop : String.new
              lines.each do |line|
                Puppet.debug(line)
              end
            end
          end
          channel.on_eof do |ch|
            # Display anything remaining in the buffer
            unless buffer.empty?
              Puppet.debug(buffer)
            end
          end
          channel.on_request("exit-status") do |ch, data|
            exit_code = data.read_long
            Puppet.debug("SSH Command Exit Code: #{exit_code}")
          end
          # Finally execute the command
          channel.exec(command)
        end
      end
      Puppet.info "Executing remote command ... Done"
      exit_code
    end