From: Woody Peterson <woody.peterson@...>
Date: 2008-09-12T19:02:10+09:00
Subject: Re: using net::ssh shell to sudo to another user and execute commands

All,

While I don't have the specific reason your scripts are failing, I have 
a snippet from something I'm writing that might be handy. Note that 
'yall are using Net::SSH v1, and my solution is for v2.

def exec(cmd)
  # the built in ssh.exec wraps some stuff up for us, but to catch sudo 
we
  # have to construct the whole thing ourselves, starting with the 
channel.
  channel = ssh.open_channel do |channel|
    # now we request a "pty" (i.e. interactive) session so we can send 
data
    # back and forth if needed. it WILL NOT WORK without this, and it 
has to
    # be done before any call to exec.
    channel.request_pty do |ch, success|
      raise "Could not obtain pty (i.e. an interactive ssh session)" if 
!success
    end

    channel.exec(cmd) do |ch, success|
      # 'success' isn't related to bash exit codes or anything, but more
      # about ssh internals (i think... not bash related anyways).
      # not sure why it would fail at such a basic level, but it seems 
smart
      # to do something about it.
      abort "could not execute command" unless success

      # on_data is a hook that fires when the loop that this block is 
fired
      # in (see below) returns data. This is what we've been doing all 
this
      # for; now we can check to see if it's a password prompt, and
      # interactively return data if so (see request_pty above).
      channel.on_data do |ch, data|
        if data == "Password:"
          puts "Password request"
          channel.send_data "#{self.password}\n"
        else
          # ssh channels can be treated as a hash for the specific 
purpose of
          # getting values out of the block later
          channel[:result] ||= ""
          channel[:result] << data
        end
      end

      channel.on_extended_data do |ch, type, data|
        raise "SSH command returned on stderr: #{data}"
      end
    end
  end

  # Nothing has actually happened yet. Everything above will respond to 
the
  # server after each execution of the ssh loop until it has nothing 
left
  # to process. For example, if the above recieved a password challenge 
from
  # the server, ssh's exec loop would execute twice - once for the 
password,
  # then again after clearing the password (or twice more and exit if 
the
  # password was bad)
  channel.wait

  return channel[:result].strip # it returns with \r\n at the end
end
-- 
Posted via http://www.ruby-forum.com/.