Question

I want to write code in Ruby witch net::ssh that run commands one by one on remote linux machine and log everything (called command, stdout and stderr on linux machine).

So I write function:

  def rs(ssh,cmds)
    cmds.each do |cmd|
      log.debug "[SSH>] #{cmd}"
      ssh.exec!(cmd) do |ch, stream, data|    
        log.debug "[SSH:#{stream}>] #{data}"            
      end  
    end
  end

For example if I want to create on remote linux new folders and file: “./verylongdirname/anotherlongdirname/a.txt”, and list files in that direcotry, and find firefox there (which is stupid a little :P) so i call above procedure like that:

Net::SSH.start(host, user, :password => pass) do |ssh|  

  cmds=["mkdir verylongdirname", \                                 #1
        "cd verylongdirname; mkdir anotherlongdirname, \           #2
        "cd verylongdirname/anotherlongdirname; touch a.txt", \    #3
        "cd verylongdirname/anotherlongdirname; ls -la", \         #4
        "cd verylongdirname/anotherlongdirname; find ./ firefox"   #5 that command send error to stderr.
        ]

  rs(ssh,cmds)   # HERE we call our function

  ssh.loop
end

After run code above i will have full LOG witch informations about executions commands in line #1,#2,#3,#4,#5. The problem is that state on linux, between execude commands from cmds array, is not saved (so I must repeat “cd” statement before run proper command). And I’m not satisfy with that.

My purpose is to have cmds tables like that:

  cmds=["mkdir verylongdirname", \     #1
        "cd verylongdirname", \        
        "mkdir anotherlongdirname", \  #2
        "cd anotherlongdirname", \
        "touch a.txt", \               #3
        "ls -la", \                    #4
        "find ./ firefox"]             #5

As you see, te state between run each command is save on the linux machine (and we don’t need repeat apropriate “cd” statement before run proper command). How to change “rs(ssh,cmds)” procedure to do it and LOG EVERYTHING (comand,stdout,stdin) like before?

Answer 1

Perhaps try it with an ssh channel instead to open a remote shell. That should preserve state between your commands as the connection will be kept open:

http://net-ssh.github.com/ssh/v1/chapter-5.html

Here’s also an article of doing something similar with a little bit different approach:

http://drnicwilliams.com/2006/09/22/remote-shell-with-ruby/

Edit 1:

Ok. I see what you are saying. SyncShell was removed from Net::SSH 2.0. However I found this, which looks like it does pretty much what SyncShell did:

http://net-ssh-telnet.rubyforge.org/

Example:

s = Net::SSH.start(host, user)
t = Net::SSH::Telnet.new("Session" => s, "Prompt" => %r{^myprompt :})
puts t.cmd("cd /tmp")  
puts t.cmd("ls")       # <- Lists contents of /tmp

I.e. Net::SSH::Telnet is synchronous, and preserves state, because it runs in a pty with your remote shell environment. Remember to set the correct prompt detection, otherwise Net::SSH::Telnet will appear to hang once you call it (it’s trying to find the prompt).